{"aliases": [], "authors": ["Rodrigo Nogueira", "Kyunghyun Cho"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "67f8697c21dabc2c4781eba46bfb60671506b413471efcdb28864c11649b3a2c", "submitted_at": "2016-02-06T14:53:02Z", "updated_at": "2016-05-20T16:26:58Z", "version": 2, "versioned_url": "https://arxiv.org/abs/1602.02261v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:26Z", "role": "primary", "sha256": "3e89566cdb4521a8b3b0d99c514c3401645eda5b5c0de9720eb726b1ffe0a11f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1511.03745%2C1602.02261%2C1809.04860%2C1811.06521%2C1811.07871%2C1812.03426%2C1901.11528%2C1909.00031%2C1912.03098%2C2003.00431%2C2004.14973%2C2006.16241%2C2007.04428%2C2010.09670%2C2010.12083%2C2012.12350%2C2104.08244%2C2105.11941%2C2106.09422%2C2106.13500%2C2107.05911%2C2108.03298%2C2108.05308%2C2110.10661%2C2201.04100%2C2201.11676%2C2203.10897%2C2204.05239%2C2205.10816%2C2206.08853%2C2206.12839%2C2206.15407%2C2208.06032%2C2208.08080%2C2209.14981%2C2210.01379%2C2210.03735%2C2210.08073%2C2210.10275%2C2210.11435%2C2211.07912%2C2211.09527%2C2212.13326%2C2301.10677%2C2301.13779%2C2302.12173%2C2303.00001%2C2303.11366%2C2303.12027%2C2304.03442&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:11:46Z", "role": "primary", "sha256": "5f99bcea7c1585129af46fa5c978002778dedc3d98e3a907e92030f4a5588f00", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/1602.02261v2"}], "id": "arxiv:1602.02261", "identifiers": {"arxiv": "1602.02261"}, "kind": "paper", "last_verified_at": "2026-10-17T09:11:46Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/1602.02261"}], "name": "End-to-End Goal-Driven Web Navigation", "notes": "Reviewed against the exact-version official arXiv abstract page; its normalized abstract SHA-256 matches the retained bibliographic snapshot. The original API response body was not locally available for this review. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2016-02-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebNav for converting websites into goal-directed hyperlink navigation tasks and WikiNav for evaluating language-conditioned navigation and question answering on Wikipedia.", "topics": ["web-navigation", "evaluation", "planning"], "updated_at": "2016-05-20", "url": "https://arxiv.org/abs/1602.02261"}
{"aliases": [], "authors": ["Panupong Pasupat", "Tian-Shun Jiang", "Evan Zheran Liu", "Kelvin Guu", "Percy Liang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f368116f70e987cde111835a05b5b36e0d345dabb4e8351610e2bbfb691b3440", "submitted_at": "2018-08-28T06:09:39Z", "updated_at": "2018-10-01T03:22:58Z", "version": 2, "versioned_url": "https://arxiv.org/abs/1808.09132v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:1808.09132", "identifiers": {"arxiv": "1808.09132"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/1808.09132"}], "name": "Mapping Natural Language Commands to Web Elements", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2018-08-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Defines language grounding on webpages as choosing the UI element targeted by a natural-language command. A collection of more than 50,000 commands and three baseline models study functional references, relationships between elements, and visual position cues.", "topics": ["grounding", "data-generation", "evaluation"], "updated_at": "2018-10-01", "url": "https://arxiv.org/abs/1808.09132"}
{"aliases": [], "authors": ["Jieshan Chen", "Mulong Xie", "Zhenchang Xing", "Chunyang Chen", "Xiwei Xu", "Liming Zhu", "Guoqiang Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d6352985532945fd562ce0382aef0e977ce1ec5909b6872e5fe048167e9732a0", "submitted_at": "2020-08-12T06:36:33Z", "updated_at": "2020-09-07T12:57:33Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2008.05132v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}], "id": "arxiv:2008.05132", "identifiers": {"arxiv": "2008.05132"}, "kind": "paper", "last_verified_at": "2026-10-16T06:42:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2008.05132"}], "name": "Object Detection for Graphical User Interface: Old Fashioned or Deep Learning or a Combination?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2020-08-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares seven traditional and learned GUI element detectors on more than 50,000 interface images. The analysis motivates a coarse-to-fine non-text detector combined with a learned text detector to address GUI-specific localization requirements.", "topics": ["grounding", "screen-understanding", "evaluation"], "updated_at": "2020-09-07", "url": "https://arxiv.org/abs/2008.05132"}
{"aliases": [], "authors": ["Xingyu Chen", "Zihan Zhao", "Lu Chen", "Danyang Zhang", "Jiabao Ji", "Ao Luo", "Yuxuan Xiong", "Kai Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5fdf4bc31edd633fd0858835db0afedeef5dbc2694c4511903976c338bd19b1b", "submitted_at": "2021-01-23T09:43:44Z", "updated_at": "2021-11-08T08:31:44Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2101.09465v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2101.09465", "identifiers": {"arxiv": "2101.09465"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2101.09465"}], "name": "WebSRC: A Dataset for Web-Based Structural Reading Comprehension", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2021-01-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces question answering over webpage text and structure, with aligned HTML, screenshots, and metadata for structural reading comprehension.", "topics": ["screen-understanding", "evaluation"], "updated_at": "2021-11-08", "url": "https://arxiv.org/abs/2101.09465"}
{"aliases": [], "authors": ["Andrea Burns", "Deniz Arsan", "Sanjna Agrawal", "Ranjitha Kumar", "Kate Saenko", "Bryan A. Plummer"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1687369e31b0ad3632381d4a46b1d555c2e2ed1d6504e9c1b63006dd29f3445e", "submitted_at": "2022-02-04T18:51:50Z", "updated_at": "2022-08-15T00:24:24Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2202.02312v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2202.02312", "identifiers": {"arxiv": "2202.02312"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2202.02312"}], "name": "A Dataset for Interactive Vision-Language Navigation with Unknown Command Feasibility", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2022-02-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MoTIF mobile-app interactions with feasible and infeasible commands, feasibility explanations, and follow-up questions for ambiguous requests.", "topics": ["mobile-navigation", "human-in-the-loop", "evaluation"], "updated_at": "2022-08-15", "url": "https://arxiv.org/abs/2202.02312"}
{"aliases": [], "authors": ["Shunyu Yao", "Howard Chen", "John Yang", "Karthik Narasimhan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6de4ab2a89875f1ccd99bbaa075c234d82979e845f27fe9530bdad0539c49570", "submitted_at": "2022-07-04T05:30:22Z", "updated_at": "2023-02-08T01:39:30Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2207.01206v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2207.01206", "identifiers": {"arxiv": "2207.01206"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2207.01206"}], "name": "WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2022-07-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Provides a simulated shopping environment where language agents search, configure, and select products from natural-language requirements.", "topics": ["evaluation", "web-navigation"], "updated_at": "2023-02-08", "url": "https://arxiv.org/abs/2207.01206"}
{"aliases": [], "authors": ["Yu-Chung Hsiao", "Fedir Zubach", "Gilles Baechler", "Srinivas Sunkara", "Victor Carbune", "Jason Lin", "Maria Wang", "Yun Zhu", "Jindong Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "37f7ba5d5b9bbc96fd3a723e49190f296e08d2874dcd545272d3fcef3cc33d24", "submitted_at": "2022-09-16T23:49:00Z", "updated_at": "2025-02-09T21:09:17Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2209.08199v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2209.08199", "identifiers": {"arxiv": "2209.08199"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2209.08199"}], "name": "ScreenQA: Large-Scale Question-Answer Pairs over Mobile App Screenshots", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2022-09-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces screenshot question answering over mobile interfaces, linking answers to screen content and bounding boxes.", "topics": ["screen-understanding", "evaluation"], "updated_at": "2025-02-09", "url": "https://arxiv.org/abs/2209.08199"}
{"aliases": [], "authors": ["Danyang Zhang", "Zhennan Shen", "Rui Xie", "Situo Zhang", "Tianbao Xie", "Zihan Zhao", "Siyuan Chen", "Lu Chen", "Hongshen Xu", "Ruisheng Cao", "Kai Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8e454f547bdf8e01f47eb63c20f0cacfa8bd17a8c4ad37dfd3a1cceb732de178", "submitted_at": "2023-05-14T12:31:03Z", "updated_at": "2024-06-13T11:51:37Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2305.08144v4"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2305.08144", "identifiers": {"arxiv": "2305.08144"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2305.08144"}], "name": "Mobile-Env: Building Qualified Evaluation Benchmarks for LLM-GUI Interaction", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-05-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Provides an Android benchmark toolkit with controlled environments, intermediate feedback, and task sets for reproducible GUI-agent evaluation.", "topics": ["evaluation", "sandbox"], "updated_at": "2024-06-13", "url": "https://arxiv.org/abs/2305.08144"}
{"aliases": [], "authors": ["Haotian Sun", "Yuchen Zhuang", "Lingkai Kong", "Bo Dai", "Chao Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c254a26ebb8e1bcdca70b7581c780aa6b0dd4e5f5ded85e456f7e8fb8b38fcc6", "submitted_at": "2023-05-26T05:52:27Z", "updated_at": "2023-05-26T05:52:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2305.16653v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2305.16653", "identifiers": {"arxiv": "2305.16653"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2305.16653"}], "name": "AdaPlanner: Adaptive Planning from Feedback with Language Models", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. This is a general planning framework; MiniWoB++ supplies the interface-agent evidence, while ALFWorld is a separate environment.", "published_at": "2023-05-26", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces AdaPlanner, which revises code-structured plans in response to environmental feedback and reuses successful plans as exemplars. Experiments include MiniWoB++ web tasks alongside ALFWorld.", "topics": ["planning", "self-reflection", "evaluation"], "updated_at": "2023-05-26", "url": "https://arxiv.org/abs/2305.16653"}
{"aliases": [], "authors": ["Hongxin Li", "Jingran Su", "Yuntao Chen", "Qing Li", "Zhaoxiang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "320ff40872677e9d32efae360dc289ca345076e037e7044188868f765e816008", "submitted_at": "2023-05-30T17:59:30Z", "updated_at": "2023-10-30T06:36:24Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2305.19308v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:56Z", "role": "primary", "sha256": "beb8c811bf2d070845aa6fcc7e94ffd43ca4f9f60ce42c72d958598790fbdffc", "supports": ["environments", "notes"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md"}], "id": "arxiv:2305.19308", "identifiers": {"arxiv": "2305.19308"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2305.19308"}], "name": "SheetCopilot: Bringing Software Productivity to the Next Level through Large Language Models", "notes": "The contribution was reviewed from the retained paper abstract. The original Windows/Excel implementation is identified in the separately retained project README; the later Ubuntu evaluator is not treated as a paper result. No results were independently reproduced.", "published_at": "2023-05-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SheetCopilot, an agent that steers spreadsheet software via atomic virtual APIs, along with a 221-task spreadsheet manipulation benchmark.", "topics": ["workflow", "tool-integration", "evaluation"], "updated_at": "2023-10-30", "url": "https://arxiv.org/abs/2305.19308"}
{"aliases": [], "authors": ["Zhizheng Zhang", "Xiaoyi Zhang", "Wenxuan Xie", "Yan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "631a39d2929c68bb93a7f95af22c17d2ab208668fd97d9082fe0601a73d123dd", "submitted_at": "2023-06-02T02:42:58Z", "updated_at": "2023-12-04T13:36:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2306.01242v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}], "id": "arxiv:2306.01242", "identifiers": {"arxiv": "2306.01242"}, "kind": "paper", "last_verified_at": "2026-10-16T06:42:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2306.01242"}], "name": "Responsible Task Automation: Empowering Large Language Models as Responsible Task Automators", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. UI task automation is stated in the abstract, but its platform and detailed evaluation protocol are not identified.", "published_at": "2023-06-02", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Presents ResponsibleTA, a coordinator-executor framework that checks command feasibility and task completion while using local memory for privacy-related controls. The paper compares prompted and learned implementations and evaluates them on UI task automation.", "topics": ["safety-security", "evaluation", "automation"], "updated_at": "2023-12-04", "url": "https://arxiv.org/abs/2306.01242"}
{"aliases": [], "authors": ["Xiang Deng", "Yu Gu", "Boyuan Zheng", "Shijie Chen", "Samuel Stevens", "Boshi Wang", "Huan Sun", "Yu Su"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "de055dfa452130f3576839c99c007053c62a0c96c9bfbcdd79ecad1a9f74f830", "submitted_at": "2023-06-09T17:44:31Z", "updated_at": "2023-12-09T05:57:46Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2306.06070v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2306.06070", "identifiers": {"arxiv": "2306.06070"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2306.06070"}], "name": "Mind2Web: Towards a Generalist Agent for the Web", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2023-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a web-task dataset spanning diverse websites and evaluates generalization to unseen sites and domains.", "topics": ["data-generation", "web-navigation", "evaluation"], "updated_at": "2023-12-09", "url": "https://arxiv.org/abs/2306.06070"}
{"aliases": [], "authors": ["Christopher Rawles", "Alice Li", "Daniel Rodriguez", "Oriana Riva", "Timothy Lillicrap"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "def4616f7cb13fe839e500c4b3171eafbe37b155b5d14301137806d6a5cde939", "submitted_at": "2023-07-19T15:57:24Z", "updated_at": "2023-10-27T14:24:31Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2307.10088v2"}, "publication": {"authors": ["Christopher Rawles", "Alice Li", "Daniel Rodriguez", "Oriana Riva", "Timothy Lillicrap"], "date_basis": "Explicit publisher citation_publication_date: 2023-12-15. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2023, 12, 15], "doi": "10.52202/075280-2609", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2023/hash/bbbb6308b402fe909c39dd29950c32e0-Abstract-Datasets_and_Benchmarks.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "59708-59728", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2023/file/bbbb6308b402fe909c39dd29950c32e0-Paper-Datasets_and_Benchmarks.pdf", "publisher": null, "series": null, "title": "AndroidInTheWild: A Large-Scale Dataset For Android Device Control", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2023/hash/bbbb6308b402fe909c39dd29950c32e0-Abstract-Datasets_and_Benchmarks.html", "venue": "Advances in Neural Information Processing Systems", "volume": "36"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:02Z", "role": "primary", "sha256": "7fa75eae1a17878ef28414f64e1c46cc83539008f2644331b762248ec96efce5", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2023/hash/bbbb6308b402fe909c39dd29950c32e0-Abstract-Datasets_and_Benchmarks.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2307.10088", "identifiers": {"arxiv": "2307.10088"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2307.10088"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2023/hash/bbbb6308b402fe909c39dd29950c32e0-Abstract-Datasets_and_Benchmarks.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2023/file/bbbb6308b402fe909c39dd29950c32e0-Paper-Datasets_and_Benchmarks.pdf"}], "name": "Android in the Wild: A Large-Scale Dataset for Android Device Control", "notes": "Annotation reviewed against the primary abstract. Original dataset scale and experimental results are not independently reproduced.", "published_at": "2023-07-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Collects Android interaction demonstrations with screens, actions, and instructions across device and platform variations for robustness evaluation.", "topics": ["data-generation", "mobile-navigation", "evaluation"], "updated_at": "2023-10-27", "url": "https://arxiv.org/abs/2307.10088"}
{"aliases": [], "authors": ["Shuyan Zhou", "Frank F. Xu", "Hao Zhu", "Xuhui Zhou", "Robert Lo", "Abishek Sridhar", "Xianyi Cheng", "Tianyue Ou", "Yonatan Bisk", "Daniel Fried", "Uri Alon", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0e9eca0165573ba53b663acbb1c7962fd2d64bc6633f10397f9cad229772a363", "submitted_at": "2023-07-25T22:59:32Z", "updated_at": "2024-04-16T15:13:18Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2307.13854v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2307.13854", "identifiers": {"arxiv": "2307.13854"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2307.13854"}], "name": "WebArena: A Realistic Web Environment for Building Autonomous Agents", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2023-07-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces reproducible web environments with task-completion checks for evaluating autonomous web agents.", "topics": ["evaluation", "web-navigation", "long-horizon"], "updated_at": "2024-04-16", "url": "https://arxiv.org/abs/2307.13854"}
{"aliases": [], "authors": ["Xiao Liu", "Hao Yu", "Hanchen Zhang", "Yifan Xu", "Xuanyu Lei", "Hanyu Lai", "Yu Gu", "Hangliang Ding", "Kaiwen Men", "Kejuan Yang", "Shudan Zhang", "Xiang Deng", "Aohan Zeng", "Zhengxiao Du", "Chenhui Zhang", "Sheng Shen", "Tianjun Zhang", "Yu Su", "Huan Sun", "Minlie Huang", "Yuxiao Dong", "Jie Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d8c4ba4db4aeed3a7b7ce7f77d610d7e34b0e8db26a36463e2c0bdffcad8f22a", "submitted_at": "2023-08-07T16:08:11Z", "updated_at": "2025-10-04T03:54:18Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2308.03688v3"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2308.03688", "identifiers": {"arxiv": "2308.03688"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2308.03688"}], "name": "AgentBench: Evaluating LLMs as Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-08-07", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Evaluates language agents across eight interactive environments and analyzes failures in reasoning, decisions, and instruction following.", "topics": ["evaluation", "planning"], "updated_at": "2025-10-04", "url": "https://arxiv.org/abs/2308.03688"}
{"aliases": [], "authors": ["Justin Payan", "Swaroop Mishra", "Mukul Singh", "Carina Negreanu", "Christian Poelitz", "Chitta Baral", "Subhro Roy", "Rasika Chakravarthy", "Benjamin Van Durme", "Elnaz Nouri"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1648acae2228b8ae2f042d7986585a66bc6a649320d569b12fd093b96c15cb5c", "submitted_at": "2023-10-23T02:00:55Z", "updated_at": "2023-10-23T02:00:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2310.14495v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:30Z", "role": "primary", "sha256": "4856a9c146f74964be623730f3cee6b15fcaad83f7d83a4956e971b541a63070", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2304.08479%2C2305.09699%2C2305.09770%2C2305.11271%2C2305.14343%2C2305.16338%2C2305.16985%2C2306.04529%2C2306.05499%2C2306.07179%2C2306.11134%2C2307.00804%2C2307.02443%2C2308.01552%2C2308.05391%2C2308.05822%2C2308.06595%2C2308.07201%2C2308.09289%2C2309.04561%2C2309.05251%2C2309.07550%2C2309.09958%2C2309.10094%2C2309.13965%2C2310.07433%2C2310.07899%2C2310.09103%2C2310.12670%2C2310.12793%2C2310.12815%2C2310.12945%2C2310.12972%2C2310.13345%2C2310.14495%2C2310.17596%2C2310.19805%2C2311.01011%2C2311.07215%2C2311.09227%2C2311.10538%2C2311.10751%2C2311.11261%2C2311.11855%2C2311.18492%2C2311.18760%2C2312.02976%2C2312.03815%2C2312.07424%2C2312.09238&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:11:47Z", "role": "primary", "sha256": "40fd8a31083963f1cbbcfeb308244494139277e4a6d0e56c2ebe5a7cda821e7c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2310.14495v1"}], "id": "arxiv:2310.14495", "identifiers": {"arxiv": "2310.14495"}, "kind": "paper", "last_verified_at": "2026-10-17T09:11:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2310.14495"}], "name": "InstructExcel: A Benchmark for Natural Language Instruction in Excel", "notes": "Reviewed against the exact-version official arXiv abstract page; its normalized abstract SHA-256 matches the retained bibliographic snapshot. The original API response body was not locally available for this review. Experimental results and artifact availability have not been independently reproduced. This is an application-specific automation benchmark grounded in recorded Excel user actions. Its execution interface is OfficeScripts, so it is adjacent rather than a visual GUI navigation benchmark.", "published_at": "2023-10-23", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces InstructExcel, a benchmark for generating Excel OfficeScripts from natural-language requests, built using scripts recorded from user actions and a collection of existing spreadsheets.", "topics": ["evaluation", "workflow", "automation"], "updated_at": "2023-10-23", "url": "https://arxiv.org/abs/2310.14495"}
{"aliases": [], "authors": ["Jiarun Liu", "Wentao Hu", "Chunhong Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c72eab40cdc28981cf83a753d88f5181b5441d82cd568265457fef96d75b36b6", "submitted_at": "2023-10-20T11:10:14Z", "updated_at": "2023-10-31T06:25:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2310.18331v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2310.18331", "identifiers": {"arxiv": "2310.18331"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2310.18331"}], "name": "AllTogether: Investigating the Efficacy of Spliced Prompt for Web Navigation using Large Language Models", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-10-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "AllTogether standardizes prompts for HTML-based web navigation and compares prompting with instruction fine-tuning. Its analysis examines how HTML context, interaction history, and current environmental feedback affect task execution.", "topics": ["web-navigation", "interaction-history", "evaluation"], "updated_at": "2023-10-31", "url": "https://arxiv.org/abs/2310.18331"}
{"aliases": [], "authors": ["Yiduo Guo", "Zekai Zhang", "Yaobo Liang", "Dongyan Zhao", "Nan Duan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "58dfa75fd82a2cc1d41954f10feb1d987de6469782d11d04c3b8bfffc8a3b745", "submitted_at": "2023-11-03T08:06:35Z", "updated_at": "2023-11-07T10:13:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2311.01767v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2311.01767", "identifiers": {"arxiv": "2311.01767"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2311.01767"}], "name": "PPTC Benchmark: Evaluating Large Language Models for PowerPoint Task Completion", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-11-03", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Evaluates multi-turn PowerPoint creation and editing, scoring the resulting presentation files rather than requiring a fixed API sequence.", "topics": ["evaluation", "tool-integration"], "updated_at": "2023-11-07", "url": "https://arxiv.org/abs/2311.01767"}
{"aliases": [], "authors": ["Grégoire Mialon", "Clémentine Fourrier", "Craig Swift", "Thomas Wolf", "Yann LeCun", "Thomas Scialom"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "29ddf2b7b4d495ee18bbd263cd77f86a71b3014b58a6f140c3bfce7cbae9c922", "submitted_at": "2023-11-21T20:34:47Z", "updated_at": "2023-11-21T20:34:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2311.12983v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2311.12983", "identifiers": {"arxiv": "2311.12983"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2311.12983"}], "name": "GAIA: a benchmark for General AI Assistants", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-11-21", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces questions that test assistant reasoning, multimodal understanding, browsing, and tool use beyond screen-action execution alone.", "topics": ["evaluation", "tool-integration"], "updated_at": "2023-11-21", "url": "https://arxiv.org/abs/2311.12983"}
{"aliases": [], "authors": ["Hiroki Furuta", "Yutaka Matsuo", "Aleksandra Faust", "Izzeddin Gur"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4b00e0a9c74b7ac8819bdda540a1069c33cbac45f0c7fb853dee995f2feb9a59", "submitted_at": "2023-11-30T17:50:47Z", "updated_at": "2024-12-31T04:24:16Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2311.18751v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2311.18751", "identifiers": {"arxiv": "2311.18751"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2311.18751"}], "name": "Exposing Limitations of Language Model Agents in Sequential-Task Compositions on the Web", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-11-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "CompWoB adds 50 composed web-automation tasks to study generalization from simpler interactions. The paper compares prompted and fine-tuned agents, balances task training data, and probes sensitivity to changes in instruction order.", "topics": ["evaluation", "long-horizon", "robustness", "web-navigation"], "updated_at": "2024-12-31", "url": "https://arxiv.org/abs/2311.18751"}
{"aliases": [], "authors": ["Difei Gao", "Lei Ji", "Zechen Bai", "Mingyu Ouyang", "Peiran Li", "Dongxing Mao", "Qinchen Wu", "Weichen Zhang", "Peiyi Wang", "Xiangwu Guo", "Hengxu Wang", "Luowei Zhou", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c771b076044c170728041a4f54662d21493adcc3742d40db08cb2f9bd0614655", "submitted_at": "2023-12-20T15:28:38Z", "updated_at": "2024-01-01T14:26:39Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2312.13108v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2312.13108", "identifiers": {"arxiv": "2312.13108"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2312.13108"}], "name": "ASSISTGUI: Task-Oriented Desktop Graphical User Interface Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-12-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Windows application tasks and an actor-critic agent framework for mouse-and-keyboard automation of desktop workflows.", "topics": ["evaluation", "planning"], "updated_at": "2024-01-01", "url": "https://arxiv.org/abs/2312.13108"}
{"aliases": [], "authors": ["Qi Chen", "Dileepa Pitawela", "Chongyang Zhao", "Gengze Zhou", "Hsiang-Ting Chen", "Qi Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "47dedddced8f06bf436ab5a101d1ec4ecf4380c710e457dc7a35fc86d2e1f230", "submitted_at": "2023-12-25T22:13:26Z", "updated_at": "2023-12-25T22:13:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2312.15820v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2312.15820", "identifiers": {"arxiv": "2312.15820"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2312.15820"}], "name": "WebVLN: Vision-and-Language Navigation on Websites", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-12-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebVLN trains website navigation from question-based instructions using both rendered visual information and underlying HTML. It introduces WebVLN-v1 and a website-aware navigation network to study this digital variant of vision-and-language navigation.", "topics": ["web-navigation", "data-generation", "evaluation"], "updated_at": "2023-12-25", "url": "https://arxiv.org/abs/2312.15820"}
{"aliases": [], "authors": ["Chang Ma", "Junlei Zhang", "Zhihao Zhu", "Cheng Yang", "Yujiu Yang", "Yaohui Jin", "Zhenzhong Lan", "Lingpeng Kong", "Junxian He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "22c5a60f9ed7a8488ec31fbf24d392e9674b0cae79ab1b269fd9d57202025801", "submitted_at": "2024-01-24T01:51:00Z", "updated_at": "2024-12-23T20:12:48Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2401.13178v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "topics", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:15:30Z", "role": "primary", "sha256": "bbfef23bd53d2febb258a91d5a706f630cd597909bdbea56f16ac97dbe4777c0", "supports": ["summary", "environments", "scope", "notes"], "url": "https://arxiv.org/html/2401.13178v2"}], "id": "arxiv:2401.13178", "identifiers": {"arxiv": "2401.13178"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2401.13178"}], "name": "AgentBoard: An Analytical Evaluation Board of Multi-turn LLM Agents", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The full paper documents search/click WebShop actions and 11 browser actions in WebArena, while the overall benchmark also covers non-web domains.", "published_at": "2024-01-24", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces AgentBoard, a multiturn agent evaluation framework with intermediate progress metrics and analysis tools. Its broader task suite includes WebShop and WebArena interactions with stepwise progress measurements.", "topics": ["evaluation", "observability", "web-navigation"], "updated_at": "2024-12-23", "url": "https://arxiv.org/abs/2401.13178"}
{"aliases": [], "authors": ["Jing Yu Koh", "Robert Lo", "Lawrence Jang", "Vikram Duvvur", "Ming Chong Lim", "Po-Yu Huang", "Graham Neubig", "Shuyan Zhou", "Ruslan Salakhutdinov", "Daniel Fried"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3a705b87b43464120ebb01b28b0f6a9e31fad68f3d463c88354f2005dab940c3", "submitted_at": "2024-01-24T18:35:21Z", "updated_at": "2024-06-06T02:01:09Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2401.13649v2"}, "publication": {"authors": ["Jing Yu Koh", "Robert Lo", "Lawrence Jang", "Vikram Duvvur", "Ming Chong Lim", "Po-Yu Huang", "Graham Neubig", "Shuyan Zhou", "Ruslan Salakhutdinov", "Daniel Fried"], "date_basis": "Explicit publisher metadata: 2024/8; precision=month.", "date_parts": [2024, 8], "doi": "10.18653/v1/2024.acl-long.50", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2024.acl-long.50/", "https://aclanthology.org/2024.acl-long.50.pdf"], "identity_basis": "Same work: exact normalized landing title, corresponding author list, and matching specific contribution. Publisher and arXiv versions remain distinct.", "issue": null, "notes": "The formal PDF title adds Visually Grounded; the landing metadata uses Visual Web Tasks. PDF authors include Ming Chong Lim and Ruslan Salakhutdinov; the landing abbreviates them to Ming Lim and Russ Salakhutdinov. Publication title/authors use the PDF; the retained arXiv title remains unchanged. Publisher-hosted proceedings record; distinct from the retained arXiv preprint snapshot.", "pages": "881-905", "pdf_url": "https://aclanthology.org/2024.acl-long.50.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "VisualWebArena: Evaluating Multimodal Agents on Realistic Visually Grounded Web Tasks", "type": "conference-paper", "url": "https://aclanthology.org/2024.acl-long.50/", "venue": "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:48+00:00", "role": "primary", "sha256": "fc908d474b12a6c0f3ff3e48129d29c809e9f3a0365a3cf647b70273f266849a", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2024.acl-long.50/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:56:04+00:00", "role": "primary", "sha256": "6fc85c38c975c2af170c1d1a28f72ba557906d9019d1d9106a15215868bb491e", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.title"], "url": "https://aclanthology.org/2024.acl-long.50.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2401.13649", "identifiers": {"arxiv": "2401.13649"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2401.13649"}, {"rel": "publication", "url": "https://aclanthology.org/2024.acl-long.50/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2024.acl-long.50.pdf"}], "name": "VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-01-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces visually grounded website tasks for evaluating multimodal agents that must interpret both images and text.", "topics": ["evaluation", "grounding", "web-navigation"], "updated_at": "2024-06-06", "url": "https://arxiv.org/abs/2401.13649"}
{"aliases": [], "authors": ["Hongliang He", "Wenlin Yao", "Kaixin Ma", "Wenhao Yu", "Yong Dai", "Hongming Zhang", "Zhenzhong Lan", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dbc7121c9606c5987df0941b849c216ad8731bebcc74a061cf3f154379d7b8da", "submitted_at": "2024-01-25T03:33:18Z", "updated_at": "2024-06-06T18:37:34Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2401.13919v4"}, "publication": {"authors": ["Hongliang He", "Wenlin Yao", "Kaixin Ma", "Wenhao Yu", "Yong Dai", "Hongming Zhang", "Zhenzhong Lan", "Dong Yu"], "date_basis": "Explicit publisher citation_publication_date: 2024/8. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 8], "doi": "10.18653/v1/2024.acl-long.371", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2024.acl-long.371/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity. The normalized publisher abstract matches the retained arXiv abstract SHA-256.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "6864-6890", "pdf_url": "https://aclanthology.org/2024.acl-long.371.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "WebVoyager: Building an End-to-End Web Agent with Large Multimodal Models", "type": "conference-paper", "url": "https://aclanthology.org/2024.acl-long.371/", "venue": "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:48Z", "role": "primary", "sha256": "9772713e428cc47e54a6a236aef9024e44c97c6794a746ade767b51bdba7679a", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2024.acl-long.371/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2401.13919", "identifiers": {"arxiv": "2401.13919"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2401.13919"}, {"rel": "publication", "url": "https://aclanthology.org/2024.acl-long.371/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2024.acl-long.371.pdf"}], "name": "WebVoyager: Building an End-to-End Web Agent with Large Multimodal Models", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-01-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines multimodal web interaction with a real-website task collection and an automated evaluation protocol.", "topics": ["web-navigation", "evaluation"], "updated_at": "2024-06-06", "url": "https://arxiv.org/abs/2401.13919"}
{"aliases": [], "authors": ["Xing Han Lù", "Zdeněk Kasner", "Siva Reddy"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6a7c1a1f260f9287e8999525e1adc5534a70f7b96271d469973fceee1ac23609", "submitted_at": "2024-02-08T18:58:02Z", "updated_at": "2024-09-10T22:22:43Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2402.05930v2"}, "publication": {"authors": ["Xing Han Lù", "Zdeněk Kasner", "Siva Reddy"], "date_basis": "Explicit publisher metadata: 2024/07/08; precision=day.", "date_parts": [2024, 7, 8], "doi": null, "event_dates": "21--27 Jul 2024", "evidence_urls": ["https://proceedings.mlr.press/v235/lu24e.html", "https://raw.githubusercontent.com/mlresearch/v235/main/assets/lu24e/lu24e.pdf"], "identity_basis": "Same work: exact normalized landing title, corresponding author list, and matching specific contribution. Publisher and arXiv versions remain distinct.", "issue": null, "notes": "The PMLR landing renders Xing Han Lu; the paper title page has Xing Han Lù. Publication authors use the PDF. The publisher issue date comes from citation_publication_date; the conference date range is retained separately and does not replace it. Publisher-hosted proceedings record; distinct from the retained arXiv preprint snapshot.", "pages": "33007-33056", "pdf_url": "https://raw.githubusercontent.com/mlresearch/v235/main/assets/lu24e/lu24e.pdf", "publisher": "PMLR", "series": "Proceedings of Machine Learning Research", "title": "WebLINX: Real-World Website Navigation with Multi-Turn Dialogue", "type": "conference-paper", "url": "https://proceedings.mlr.press/v235/lu24e.html", "venue": "Proceedings of the 41st International Conference on Machine Learning", "volume": "235"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:47+00:00", "role": "primary", "sha256": "1b30008f833b3b06238d87d0509a10876a08153a78a06c79bc7590c30c471777", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.event_dates", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.series", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.mlr.press/v235/lu24e.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:59:46+00:00", "role": "primary", "sha256": "c644b1e3f071879aec8ea0e741e31e84193905dd730382ada54f1f24d04298ab", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes"], "url": "https://raw.githubusercontent.com/mlresearch/v235/main/assets/lu24e/lu24e.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2402.05930", "identifiers": {"arxiv": "2402.05930"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2402.05930"}, {"rel": "publication", "url": "https://proceedings.mlr.press/v235/lu24e.html"}, {"rel": "publication_pdf", "url": "https://raw.githubusercontent.com/mlresearch/v235/main/assets/lu24e/lu24e.pdf"}], "name": "WebLINX: Real-World Website Navigation with Multi-Turn Dialogue", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-02-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Collects expert demonstrations of conversational web navigation and studies retrieval-based element selection for processing large pages.", "topics": ["data-generation", "web-navigation", "evaluation"], "updated_at": "2024-09-10", "url": "https://arxiv.org/abs/2402.05930"}
{"aliases": [], "authors": ["Mingzhe Xing", "Rongkai Zhang", "Hui Xue", "Qi Chen", "Fan Yang", "Zhen Xiao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0504cfffc5d46ecf447a4fdeac561aede9482bb8640fcfd9d7f02cf6eaf47597", "submitted_at": "2024-02-09T18:19:25Z", "updated_at": "2024-02-09T18:19:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2402.06596v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2402.06596", "identifiers": {"arxiv": "2402.06596"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2402.06596"}], "name": "Understanding the Weakness of Large Language Model Agents within a Complex Android Environment", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-02-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AndroidArena and analyzes agent limitations in cross-application tasks, user constraints, exploration, and reflection.", "topics": ["evaluation", "cross-app"], "updated_at": "2024-02-09", "url": "https://arxiv.org/abs/2402.06596"}
{"aliases": [], "authors": ["Yang Deng", "Xuan Zhang", "Wenxuan Zhang", "Yifei Yuan", "See-Kiong Ng", "Tat-Seng Chua"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5e4b233e423e9043b6d615c7b185a1eb0d47f646590e4fb9fd43ca156a1a1b16", "submitted_at": "2024-02-23T02:18:12Z", "updated_at": "2024-02-23T02:18:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2402.15057v1"}, "publication": {"authors": ["Yang Deng", "Xuan Zhang", "Wenxuan Zhang", "Yifei Yuan", "See Kiong Ng", "Tat-Seng Chua"], "date_basis": "Explicit publisher citation_publication_date: 2024/8. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 8], "doi": "10.18653/v1/2024.acl-long.477", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2024.acl-long.477/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity. The normalized publisher abstract matches the retained arXiv abstract SHA-256.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "8795-8812", "pdf_url": "https://aclanthology.org/2024.acl-long.477.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "On the Multi-turn Instruction Following for Conversational Web Agents", "type": "conference-paper", "url": "https://aclanthology.org/2024.acl-long.477/", "venue": "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:49Z", "role": "primary", "sha256": "ba56d6d0a422b340672036d5c872cd00d54a0f12cfe1326a3be65e8702786381", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2024.acl-long.477/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2402.15057", "identifiers": {"arxiv": "2402.15057"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2402.15057"}, {"rel": "publication", "url": "https://aclanthology.org/2024.acl-long.477/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2024.acl-long.477.pdf"}], "name": "On the Multi-turn Instruction Following for Conversational Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-02-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "MT-Mind2Web evaluates web navigation through sequences of conversational user instructions. Self-MAP combines memory-augmented planning with reflection to preserve context and dependencies across those interaction turns.", "topics": ["evaluation", "memory", "self-reflection", "web-navigation"], "updated_at": "2024-02-23", "url": "https://arxiv.org/abs/2402.15057"}
{"aliases": [], "authors": ["Fangzhou Wu", "Shutong Wu", "Yulong Cao", "Chaowei Xiao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2cf1458c8c4134b2195394a8acba4919f9b3a5a7c79b1bea017ac4558123f69c", "submitted_at": "2024-02-26T19:01:54Z", "updated_at": "2024-02-26T19:01:54Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2402.16965v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2402.16965", "identifiers": {"arxiv": "2402.16965"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2402.16965"}], "name": "WIPI: A New Web Threat for LLM-Driven Web Agents", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-02-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies WIPI, a black-box indirect-instruction threat in which webpage content redirects LLM-driven web agents. Evaluation spans plugin-based agents, Web GPTs, and open-source web-agent implementations.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2024-02-26", "url": "https://arxiv.org/abs/2402.16965"}
{"aliases": [], "authors": ["Raghav Kapoor", "Yash Parag Butala", "Melisa Russak", "Jing Yu Koh", "Kiran Kamble", "Waseem Alshikh", "Ruslan Salakhutdinov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "54b548ac2bee529d3a769e7ea14b6bfce2270cdf349d832cc8c146e7d779b81a", "submitted_at": "2024-02-27T14:47:53Z", "updated_at": "2024-07-21T23:16:13Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2402.17553v3"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2402.17553", "identifiers": {"arxiv": "2402.17553"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2402.17553"}], "name": "OmniACT: A Dataset and Benchmark for Enabling Multimodal Generalist Autonomous Agents for Desktop and Web", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-02-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a dataset and benchmark for generating executable task scripts from screenshots and visually grounded desktop or web instructions.", "topics": ["evaluation", "grounding"], "updated_at": "2024-07-21", "url": "https://arxiv.org/abs/2402.17553"}
{"aliases": [], "authors": ["Zekai Zhang", "Yiduo Guo", "Yaobo Liang", "Dongyan Zhao", "Nan Duan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "51abf6b165b84c647c9a4870a6441900cac5e9a933199fe8a47d1b7f065eb11e", "submitted_at": "2024-03-06T15:33:32Z", "updated_at": "2024-03-06T15:33:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2403.03788v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2403.03788", "identifiers": {"arxiv": "2403.03788"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2403.03788"}], "name": "PPTC-R benchmark: Towards Evaluating the Robustness of Large Language Models for PowerPoint Task Completion", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The benchmark directly concerns PowerPoint task automation but executes through APIs, so it remains adjacent.", "published_at": "2024-03-06", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Benchmarks robustness of API-based PowerPoint completion under sentence-level, semantic, and multilingual instruction perturbations and changes to the available API set.", "topics": ["evaluation", "robustness", "workflow"], "updated_at": "2024-03-06", "url": "https://arxiv.org/abs/2403.03788"}
{"aliases": [], "authors": ["Alexandre Drouin", "Maxime Gasse", "Massimo Caccia", "Issam H. Laradji", "Manuel Del Verme", "Tom Marty", "Léo Boisvert", "Megh Thakkar", "Quentin Cappart", "David Vazquez", "Nicolas Chapados", "Alexandre Lacoste"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8033f9c26a5f95a49000374a6edebb4dead83e100ebcf5226007f06b7573c305", "submitted_at": "2024-03-12T14:58:45Z", "updated_at": "2024-07-23T06:19:28Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2403.07718v5"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2403.07718", "identifiers": {"arxiv": "2403.07718"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2403.07718"}], "name": "WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-03-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces enterprise web tasks on ServiceNow and BrowserGym interfaces for evaluating browser-based agents.", "topics": ["evaluation", "workflow"], "updated_at": "2024-07-23", "url": "https://arxiv.org/abs/2403.07718"}
{"aliases": [], "authors": ["Kevin Xu", "Yeganeh Kordi", "Tanay Nayak", "Adi Asija", "Yizhong Wang", "Kate Sanders", "Adam Byerly", "Jingyu Zhang", "Benjamin Van Durme", "Daniel Khashabi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d82424b84ebddfeb005ffae38c130c6a9c7dae7adcdc59e2b7d001284b785573", "submitted_at": "2024-03-18T16:06:30Z", "updated_at": "2025-02-22T01:33:54Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2403.11905v4"}, "publication": {"authors": ["Kevin Xu", "Yeganeh Kordi", "Tanay Nayak", "Adi Asija", "Yizhong Wang", "Kate Sanders", "Adam Byerly", "Jingyu Zhang", "Benjamin Van Durme", "Daniel Khashabi"], "date_basis": "Explicit publisher citation_publication_date: 2025/4. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 4], "doi": "10.18653/v1/2025.naacl-long.188", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.naacl-long.188/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "3694-3710", "pdf_url": "https://aclanthology.org/2025.naacl-long.188.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "TurkingBench: A Challenge Benchmark for Web Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.naacl-long.188/", "venue": "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:50Z", "role": "primary", "sha256": "56537fe44b4655949e8c70cc49a4ed7304a8201a12f75ddb3caa5df92d6f8e16", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.naacl-long.188/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2403.11905", "identifiers": {"arxiv": "2403.11905"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2403.11905"}, {"rel": "publication", "url": "https://aclanthology.org/2025.naacl-long.188/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.naacl-long.188.pdf"}], "name": "Tur[k]ingBench: A Challenge Benchmark for Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-03-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates web agents on multimodal annotation tasks adapted from HTML interfaces originally built for crowdsourcing workers.", "topics": ["evaluation", "web-navigation"], "updated_at": "2025-02-22", "url": "https://arxiv.org/abs/2403.11905"}
{"aliases": [], "authors": ["Longtao Zheng", "Zhiyuan Huang", "Zhenghai Xue", "Xinrun Wang", "Bo An", "Shuicheng Yan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fe1432b46c15566182418bf3fa2f918607e8b0e9a383459416d03e28dc7e347c", "submitted_at": "2024-03-26T17:54:15Z", "updated_at": "2025-02-14T08:13:39Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2403.17918v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2403.17918", "identifiers": {"arxiv": "2403.17918"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2403.17918"}], "name": "AgentStudio: A Toolkit for Building General Virtual Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-03-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "AgentStudio combines an interactive GUI/API environment with tools for creating tasks, annotating interface elements, and labeling actions in videos. Its online suite and GroundUI, IDMBench, and CriticBench datasets examine interaction, grounding, learning from video, and success detection.", "topics": ["evaluation", "sandbox", "grounding", "video"], "updated_at": "2025-02-14", "url": "https://arxiv.org/abs/2403.17918"}
{"aliases": [], "authors": ["Junpeng Liu", "Yifan Song", "Bill Yuchen Lin", "Wai Lam", "Graham Neubig", "Yuanzhi Li", "Xiang Yue"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7414bfa289baf86a8f2bb54f3f69f25a6a4cc8516fb0296a1899d21ad3020d4f", "submitted_at": "2024-04-09T02:29:39Z", "updated_at": "2024-04-09T02:29:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2404.05955v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.05955", "identifiers": {"arxiv": "2404.05955"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.05955"}], "name": "VisualWebBench: How Far Have Multimodal LLMs Evolved in Web Page Understanding and Grounding?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-04-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Separately evaluates web-page reading, understanding, and element grounding across tasks drawn from real websites.", "topics": ["evaluation", "screen-understanding", "grounding"], "updated_at": "2024-04-09", "url": "https://arxiv.org/abs/2404.05955"}
{"aliases": [], "authors": ["Jiayi Pan", "Yichi Zhang", "Nicholas Tomlin", "Yifei Zhou", "Sergey Levine", "Alane Suhr"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "abc75ee55c7ac5992656dcec2106e286c670a58130a59b0eef7f5fd74d7496c5", "submitted_at": "2024-04-09T17:25:47Z", "updated_at": "2024-10-07T14:19:37Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2404.06474v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.06474", "identifiers": {"arxiv": "2404.06474"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.06474"}], "name": "Autonomous Evaluation and Refinement of Digital Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2024-04-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares automatic trajectory evaluators for web navigation and device control against oracle judgments. The resulting feedback is used for agent fine-tuning and inference-time guidance, including experiments on WebArena.", "topics": ["evaluation", "reward-models"], "updated_at": "2024-10-07", "url": "https://arxiv.org/abs/2404.06474"}
{"aliases": [], "authors": ["Tianbao Xie", "Danyang Zhang", "Jixuan Chen", "Xiaochuan Li", "Siheng Zhao", "Ruisheng Cao", "Toh Jing Hua", "Zhoujun Cheng", "Dongchan Shin", "Fangyu Lei", "Yitao Liu", "Yiheng Xu", "Shuyan Zhou", "Silvio Savarese", "Caiming Xiong", "Victor Zhong", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a347bfcb4de03a60410392b52d522e5b7ad5dbabb474d60fc1ec097e74c4008d", "submitted_at": "2024-04-11T17:56:05Z", "updated_at": "2024-05-30T08:55:12Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.07972v2"}, "publication": {"authors": ["Tianbao Xie", "Danyang Zhang", "Jixuan Chen", "Xiaochuan Li", "Siheng Zhao", "Ruisheng Cao", "Toh J. Hua", "Zhoujun Cheng", "Dongchan Shin", "Fangyu Lei", "Yitao Liu", "Yiheng Xu", "Shuyan Zhou", "Silvio Savarese", "Caiming Xiong", "Victor Zhong", "Tao Yu"], "date_basis": "Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 12, 16], "doi": "10.52202/079017-1650", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/5d413e48f84dc61244b6be550f1cd8f5-Abstract-Datasets_and_Benchmarks_Track.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "52040-52094", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/5d413e48f84dc61244b6be550f1cd8f5-Paper-Datasets_and_Benchmarks_Track.pdf", "publisher": null, "series": null, "title": "OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/5d413e48f84dc61244b6be550f1cd8f5-Abstract-Datasets_and_Benchmarks_Track.html", "venue": "Advances in Neural Information Processing Systems", "volume": "37"}}, "environments": ["desktop", "linux", "windows", "macos"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:02Z", "role": "primary", "sha256": "f62cbd2e5c92d250cc894a05e46f9052cfd38f0cdbf14902af6a96b96bcc9e8d", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/5d413e48f84dc61244b6be550f1cd8f5-Abstract-Datasets_and_Benchmarks_Track.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.07972", "identifiers": {"arxiv": "2404.07972"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.07972"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/5d413e48f84dc61244b6be550f1cd8f5-Abstract-Datasets_and_Benchmarks_Track.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/5d413e48f84dc61244b6be550f1cd8f5-Paper-Datasets_and_Benchmarks_Track.pdf"}], "name": "OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2024-04-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a real-computer evaluation environment with execution-based checks for tasks spanning desktop applications and operating-system state.", "topics": ["evaluation", "cross-app"], "updated_at": "2024-05-30", "url": "https://arxiv.org/abs/2404.07972"}
{"aliases": [], "authors": ["Lei Ding", "Jeshwanth Bheemanpally", "Yi Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b3a030a53a942eacfabafc64af67591fd0a85af66291c5d1b48baea14556e8aa", "submitted_at": "2024-04-13T00:20:09Z", "updated_at": "2024-07-09T01:14:21Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2404.08860v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.08860", "identifiers": {"arxiv": "2404.08860"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.08860"}], "name": "Enhancing Mobile \"How-to\" Queries with Automated Search Results Verification and Reranking", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-04-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Executes instructions from mobile how-to search results inside a controlled Android environment, then uses the observed success of those instructions to verify and rerank technical-support results.", "topics": ["mobile-navigation", "evaluation", "automation"], "updated_at": "2024-07-09", "url": "https://arxiv.org/abs/2404.08860"}
{"aliases": [], "authors": ["Chenhui Cui", "Tao Li", "Junjie Wang", "Chunyang Chen", "Dave Towey", "Rubing Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a844cdd06a32b8e89f362b7e924b36e27e72c1f1def81ed0b934f761ffcaba78", "submitted_at": "2024-04-13T09:56:50Z", "updated_at": "2026-06-25T14:05:55Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2404.08948v4"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.08948", "identifiers": {"arxiv": "2404.08948"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.08948"}], "name": "Large Language Models for Mobile GUI Text Input Generation: An Empirical Study", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-04-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies language-model generation of mobile text inputs, comparing UI representations, execution feedback, and human assistance for app testing.", "topics": ["automation", "evaluation"], "updated_at": "2026-06-25", "url": "https://arxiv.org/abs/2404.08948"}
{"aliases": [], "authors": ["Shulin Tian", "Ziniu Zhang", "Liangyu Chen", "Ziwei Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "667b1f5053f3ae6ea712999932d5e9ee2f42e402cc72e713048d0790f5931976", "submitted_at": "2024-04-15T17:59:50Z", "updated_at": "2025-06-29T05:24:54Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.09992v2"}, "publication": {"authors": ["Shulin Tian", "Ziniu Zhang", "Liang-Yu Chen", "Ziwei Liu"], "date_basis": "Explicit publisher citation_publication_date: 2025/7. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 7], "doi": "10.18653/v1/2025.findings-acl.703", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.findings-acl.703/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "13682-13697", "pdf_url": "https://aclanthology.org/2025.findings-acl.703.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "MMInA: Benchmarking Multihop Multimodal Internet Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.findings-acl.703/", "venue": "Findings of the Association for Computational Linguistics: ACL 2025", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:51Z", "role": "primary", "sha256": "5e725b77b94b484284adaab9ec70674c9fc13a7fce34b764e59561f9180b915f", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.findings-acl.703/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.09992", "identifiers": {"arxiv": "2404.09992"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.09992"}, {"rel": "publication", "url": "https://aclanthology.org/2025.findings-acl.703/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.findings-acl.703.pdf"}], "name": "MMInA: Benchmarking Multihop Multimodal Internet Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates multimodal agents on compositional tasks across changing websites and studies replaying prior trajectories as memory support.", "topics": ["evaluation", "long-horizon", "memory"], "updated_at": "2025-06-29", "url": "https://arxiv.org/abs/2404.09992"}
{"aliases": [], "authors": ["Li Zhang", "Shihe Wang", "Xianqing Jia", "Zhihan Zheng", "Yunhe Yan", "Longxi Gao", "Yuanchun Li", "Mengwei Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "28c3ad086106ec4199dcc9da2b8aa3b20eac9347d9354b9dfb373e50ea4c3f50", "submitted_at": "2024-04-12T15:39:09Z", "updated_at": "2024-08-02T13:49:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.16054v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.16054", "identifiers": {"arxiv": "2404.16054"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.16054"}], "name": "LlamaTouch: A Faithful and Scalable Testbed for Mobile UI Task Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-04-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "LlamaTouch executes mobile tasks on devices and evaluates whether an agent visits manually annotated essential application or system states. Screenshot and hierarchy annotations, combined with exact and fuzzy state matching, support evaluation of 496 tasks across multiple agents.", "topics": ["evaluation", "mobile-navigation", "state-tracking"], "updated_at": "2024-08-02", "url": "https://arxiv.org/abs/2404.16054"}
{"aliases": [], "authors": ["Juyong Lee", "Taywon Min", "Minyong An", "Dongyoon Hahm", "Haeone Lee", "Changyeon Kim", "Kimin Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "45c13064bbfbcf153566b86c8969f4de5b5e87dc133342c595b3b58bde3bb615", "submitted_at": "2024-04-25T14:56:32Z", "updated_at": "2025-07-21T02:55:09Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2404.16660v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.16660", "identifiers": {"arxiv": "2404.16660"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.16660"}], "name": "Benchmarking Mobile Device Control Agents across Diverse Configurations", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-04-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces B-MoCA, an interactive mobile-control benchmark that varies device settings, layouts, and languages to test generalization.", "topics": ["evaluation", "robustness"], "updated_at": "2025-07-21", "url": "https://arxiv.org/abs/2404.16660"}
{"aliases": [], "authors": ["Zirui Song", "Yaohang Li", "Meng Fang", "Yanda Li", "Zhenhao Chen", "Zecheng Shi", "Yuan Huang", "Xiuying Chen", "Ling Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5e531ec2ef9e5d90586b8f3052e16d41da95114b7382047d0a02fd671a0db355", "submitted_at": "2024-04-28T05:33:15Z", "updated_at": "2025-03-23T13:04:57Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2404.18074v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.18074", "identifiers": {"arxiv": "2404.18074"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.18074"}], "name": "MMAC-Copilot: Multi-modal Agent Collaboration Operating Copilot", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2024-04-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MMAC-Copilot, which coordinates agents with different expertise to operate PC applications. VIBench evaluates visual interaction with applications that lack usable APIs, spanning office, recreation and game tasks.", "topics": ["automation", "evaluation", "planning"], "updated_at": "2025-03-23", "url": "https://arxiv.org/abs/2404.18074"}
{"aliases": [], "authors": ["Lucas-Andreï Thil", "Mirela Popa", "Gerasimos Spanakis"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3498a7bc289c6537ab69780a56a9fd0a2bb697fb74c077949824c4fc018762c6", "submitted_at": "2024-05-01T13:51:45Z", "updated_at": "2024-05-01T13:51:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2405.00516v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2405.00516", "identifiers": {"arxiv": "2405.00516"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2405.00516"}], "name": "Navigating WebAI: Training Agents to Complete Web Tasks with Large Language Models and Reinforcement Learning", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-05-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines supervised and reinforcement learning for MiniWoB web tasks and investigates whether agents understand HTML structure or memorize target elements.", "topics": ["web-navigation", "reinforcement-learning", "evaluation"], "updated_at": "2024-05-01", "url": "https://arxiv.org/abs/2405.00516"}
{"aliases": [], "authors": ["Kihoon Son", "Jinhyeon Kwon", "DaEun Choi", "Tae Soo Kim", "Young-Ho Kim", "Sangdoo Yun", "Juho Kim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a358d417d8bef89a01e799d88564e3ede4678ebda4950e87c96cf1e1954b85e1", "submitted_at": "2024-05-07T17:10:31Z", "updated_at": "2024-05-08T05:44:21Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2405.04497v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2405.04497", "identifiers": {"arxiv": "2405.04497"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2405.04497"}], "name": "Unveiling Disparities in Web Task Handling Between Human and Web Agent", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-05-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares human think-aloud web-task behavior with web-agent planning, action, and reflection. The analysis highlights differences in updating plans, handling ambiguity, and investigating failures.", "topics": ["evaluation", "human-in-the-loop", "planning"], "updated_at": "2024-05-08", "url": "https://arxiv.org/abs/2405.04497"}
{"aliases": [], "authors": ["William E Bishop", "Alice Li", "Christopher Rawles", "Oriana Riva"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "81f62326dddc5c512fb05472eb1f387064c0a5aad77453e9d2e85415bb627fa9", "submitted_at": "2024-05-17T23:27:33Z", "updated_at": "2024-05-17T23:27:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2405.11120v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2405.11120", "identifiers": {"arxiv": "2405.11120"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2405.11120"}], "name": "Latent State Estimation Helps UI Agents to Reason", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-05-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies explicit textual estimates of latent UI state, including whether commanded actions occurred and how far a task has progressed. The experiments compare agents that reason with these estimates against agents that omit them across public and internal task benchmarks.", "topics": ["state-tracking", "planning", "evaluation"], "updated_at": "2024-05-17", "url": "https://arxiv.org/abs/2405.11120"}
{"aliases": [], "authors": ["Christopher Rawles", "Sarah Clinckemaillie", "Yifan Chang", "Jonathan Waltz", "Gabrielle Lau", "Marybeth Fair", "Alice Li", "William Bishop", "Wei Li", "Folawiyo Campbell-Ajala", "Daniel Toyama", "Robert Berry", "Divya Tyamagundlu", "Timothy Lillicrap", "Oriana Riva"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a33b8b0a03c0dc7ac3f699c34193a526d20791bfedfdde2a2fb89604959e08c2", "submitted_at": "2024-05-23T13:48:54Z", "updated_at": "2025-04-06T20:37:50Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2405.14573v5"}, "publication": {"authors": ["Chris Rawles", "Sarah Clinckemaillie", "Yifan Chang", "Jonathan Waltz", "Gabrielle Lau", "Marybeth Fair", "Alice Li", "William Bishop", "Wei Li", "Folawiyo Campbell-Ajala", "Daniel Toyama", "Robert Berry", "Divya Tyamagundlu", "Timothy Lillicrap", "Oriana Riva"], "date_basis": "Explicit publisher citation_publication_date: 2025-05-01. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 5, 1], "doi": null, "event_dates": null, "evidence_urls": ["https://proceedings.iclr.cc/paper_files/paper/2025/hash/01a83bc2f2732a58e6aa731e659e7101-Abstract-Conference.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "406-441", "pdf_url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/01a83bc2f2732a58e6aa731e659e7101-Paper-Conference.pdf", "publisher": null, "series": null, "title": "AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents", "type": "conference-paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/01a83bc2f2732a58e6aa731e659e7101-Abstract-Conference.html", "venue": "International Conference on Learning Representations", "volume": "2025"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:00Z", "role": "primary", "sha256": "e525e5eeedcb08e61765eb0d4034c43381bd3ecc5cc4bf9cc8788cf50e7a2bca", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/01a83bc2f2732a58e6aa731e659e7101-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2405.14573", "identifiers": {"arxiv": "2405.14573"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2405.14573"}, {"rel": "publication", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/01a83bc2f2732a58e6aa731e659e7101-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/01a83bc2f2732a58e6aa731e659e7101-Paper-Conference.pdf"}], "name": "AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2024-05-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces parameterized Android tasks with initialization, success checks, and cleanup for reproducible interactive evaluation.", "topics": ["evaluation", "mobile-navigation"], "updated_at": "2025-04-06", "url": "https://arxiv.org/abs/2405.14573"}
{"aliases": [], "authors": ["Ajay Patel", "Markus Hofmarcher", "Claudiu Leoveanu-Condrei", "Marius-Constantin Dinu", "Chris Callison-Burch", "Sepp Hochreiter"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "77b292052a2ab4e8db61606ffe1da8bd5e63b3ce8c84b2e0e4b7a2fdc2f926ca", "submitted_at": "2024-05-30T17:52:36Z", "updated_at": "2024-10-01T21:28:29Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2405.20309v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2405.20309", "identifiers": {"arxiv": "2405.20309"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2405.20309"}], "name": "Large Language Models Can Self-Improve At Web Agent Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-05-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Examines self-improvement by fine-tuning web agents on three mixtures of model-generated training data. WebArena experiments and trajectory-level evaluation measures study task performance, robustness, and behavior beyond aggregate success rates.", "topics": ["data-generation", "web-navigation", "evaluation"], "updated_at": "2024-10-01", "url": "https://arxiv.org/abs/2405.20309"}
{"aliases": [], "authors": ["Eric Li", "Jim Waldo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b9a728edf0b834adb72abfc50e13d5b2b52f532f444b3e8af34effd3d86beb1b", "submitted_at": "2024-06-01T00:32:26Z", "updated_at": "2024-06-01T00:32:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2406.01623v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2406.01623", "identifiers": {"arxiv": "2406.01623"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.01623"}], "name": "WebSuite: Systematically Evaluating Why Web Agents Fail", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebSuite organizes browser operations into an action taxonomy and tests both isolated controls and complete workflows. Its diagnostic tasks attribute failures to specific actions, enabling comparison of text-based and multimodal web agents.", "topics": ["evaluation", "observability", "web-navigation"], "updated_at": "2024-06-01", "url": "https://arxiv.org/abs/2406.01623"}
{"aliases": [], "authors": ["Wei Li", "William Bishop", "Alice Li", "Chris Rawles", "Folawiyo Campbell-Ajala", "Divya Tyamagundlu", "Oriana Riva"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6eb3209d8e2a431987fe19f01cd0818f9fb490e684d7369e7d3002dc388ff271", "submitted_at": "2024-06-06T01:49:29Z", "updated_at": "2024-11-13T16:42:22Z", "version": 6, "versioned_url": "https://arxiv.org/abs/2406.03679v6"}, "publication": {"authors": ["Wei Li", "William Bishop", "Alice Li", "Chris Rawles", "Folawiyo Campbell-Ajala", "Divya Tyamagundlu", "Oriana Riva"], "date_basis": "Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 12, 16], "doi": "10.52202/079017-2925", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/a79f3ef3b445fd4659f44648f7ea8ffd-Abstract-Datasets_and_Benchmarks_Track.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "92130-92154", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/a79f3ef3b445fd4659f44648f7ea8ffd-Paper-Datasets_and_Benchmarks_Track.pdf", "publisher": null, "series": null, "title": "On the Effects of Data Scale on UI Control Agents", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/a79f3ef3b445fd4659f44648f7ea8ffd-Abstract-Datasets_and_Benchmarks_Track.html", "venue": "Advances in Neural Information Processing Systems", "volume": "37"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:03Z", "role": "primary", "sha256": "0f499dc380fcd787579aab846e3d1e39b720b672fd9916b4a016688218ca48dd", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/a79f3ef3b445fd4659f44648f7ea8ffd-Abstract-Datasets_and_Benchmarks_Track.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2406.03679", "identifiers": {"arxiv": "2406.03679"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.03679"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/a79f3ef3b445fd4659f44648f7ea8ffd-Abstract-Datasets_and_Benchmarks_Track.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/a79f3ef3b445fd4659f44648f7ea8ffd-Paper-Datasets_and_Benchmarks_Track.pdf"}], "name": "On the Effects of Data Scale on UI Control Agents", "notes": "Annotation reviewed against the primary abstract. Original dataset scale and experimental results are not independently reproduced.", "published_at": "2024-06-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies how Android task performance changes with demonstration scale and releases AndroidControl with high- and low-level instructions.", "topics": ["data-generation", "mobile-navigation", "evaluation"], "updated_at": "2024-11-13", "url": "https://arxiv.org/abs/2406.03679"}
{"aliases": [], "authors": ["Luyuan Wang", "Yongyu Deng", "Yiwei Zha", "Guodong Mao", "Qinmin Wang", "Tianchen Min", "Wei Chen", "Shoufa Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "637843b1b8634e608bed71fc0f3df205ae218d1b0b5550ce480046c67bb14b0f", "submitted_at": "2024-06-12T13:14:50Z", "updated_at": "2024-06-12T13:14:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2406.08184v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2406.08184", "identifiers": {"arxiv": "2406.08184"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.08184"}], "name": "MobileAgentBench: An Efficient and User-Friendly Benchmark for Mobile LLM Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a mobile-agent benchmark with tasks across open-source applications, organized by difficulty to reduce manual evaluation effort.", "topics": ["evaluation", "mobile-navigation"], "updated_at": "2024-06-12", "url": "https://arxiv.org/abs/2406.08184"}
{"aliases": [], "authors": ["Shengcheng Yu", "Chunrong Fang", "Mingzhe Du", "Zimin Ding", "Zhenyu Chen", "Zhendong Su"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0061e9004afe11bf188164ea37de1acd15e1b95ae2fcad22044264a89e4636b3", "submitted_at": "2024-06-12T15:48:39Z", "updated_at": "2024-06-12T15:48:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2406.08340v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2406.08340", "identifiers": {"arxiv": "2406.08340"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.08340"}], "name": "Practical, Automated Scenario-based Mobile App Testing", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "ScenTest builds scenario-specific event knowledge graphs from crowdsourced test reports containing GUI and action information. It uses the current app state and testing context to generate scenario tests across applications.", "topics": ["automation", "cross-app", "screen-understanding", "evaluation"], "updated_at": "2024-06-12", "url": "https://arxiv.org/abs/2406.08340"}
{"aliases": [], "authors": ["Zhen Xiang", "Linzhi Zheng", "Yanjie Li", "Junyuan Hong", "Qinbin Li", "Han Xie", "Jiawei Zhang", "Zidi Xiong", "Chulin Xie", "Carl Yang", "Dawn Song", "Bo Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "50a8ea492d3f1c2ff2bd00ebfb18e19077bacc96b44c751d985bfe0f7bef0247", "submitted_at": "2024-06-13T14:49:26Z", "updated_at": "2025-05-29T03:09:05Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2406.09187v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}], "id": "arxiv:2406.09187", "identifiers": {"arxiv": "2406.09187"}, "kind": "paper", "last_verified_at": "2026-10-16T06:42:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.09187"}], "name": "GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoning", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The web-agent evidence is Mind2Web-SC; the healthcare benchmark does not independently establish GUI control.", "published_at": "2024-06-13", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces GuardAgent, which translates safety requests into executable guardrail code using plans and retrieved demonstrations. Mind2Web-SC supplies a web-agent safety-policy evaluation alongside a separate healthcare access-control benchmark.", "topics": ["safety-security", "evaluation", "tool-integration"], "updated_at": "2025-05-29", "url": "https://arxiv.org/abs/2406.09187"}
{"aliases": [], "authors": ["Kevin Qinghong Lin", "Linjie Li", "Difei Gao", "Qinchen WU", "Mingyi Yan", "Zhengyuan Yang", "Lijuan Wang", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "66b5fb1e839490df80e9750fdc049f7fe3d2d4686c4090d78ea01bc542fe55b0", "submitted_at": "2024-06-14T17:59:08Z", "updated_at": "2024-06-14T17:59:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2406.10227v1"}, "publication": {"authors": ["Kevin Q. Lin", "Linjie Li", "Difei Gao", "Qinchen Wu", "Mingyi Yan", "Zhengyuan Yang", "Lijuan Wang", "Mike Z. Shou"], "date_basis": "Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 12, 16], "doi": "10.52202/079017-2214", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/804e757b7d7043c26701c3a313032101-Abstract-Datasets_and_Benchmarks_Track.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "69329-69360", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/804e757b7d7043c26701c3a313032101-Paper-Datasets_and_Benchmarks_Track.pdf", "publisher": null, "series": null, "title": "VideoGUI: A Benchmark for GUI Automation from Instructional Videos", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/804e757b7d7043c26701c3a313032101-Abstract-Datasets_and_Benchmarks_Track.html", "venue": "Advances in Neural Information Processing Systems", "volume": "37"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:03Z", "role": "primary", "sha256": "5731b34872eb84e83338aa851d7f3922e304162869753bf6100ffdc1183c96cf", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/804e757b7d7043c26701c3a313032101-Abstract-Datasets_and_Benchmarks_Track.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2406.10227", "identifiers": {"arxiv": "2406.10227"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.10227"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/804e757b7d7043c26701c3a313032101-Abstract-Datasets_and_Benchmarks_Track.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/804e757b7d7043c26701c3a313032101-Paper-Datasets_and_Benchmarks_Track.pdf"}], "name": "VideoGUI: A Benchmark for GUI Automation from Instructional Videos", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates GUI automation from instructional videos at separate levels of high-level planning, action narration, and primitive execution.", "topics": ["evaluation", "video", "planning"], "updated_at": "2024-06-14", "url": "https://arxiv.org/abs/2406.10227"}
{"aliases": [], "authors": ["Dongping Chen", "Yue Huang", "Siyuan Wu", "Jingyu Tang", "Liuyi Chen", "Yilin Bai", "Zhigang He", "Chenlong Wang", "Huichi Zhou", "Yiqiang Li", "Tianshuo Zhou", "Yue Yu", "Chujie Gao", "Qihui Zhang", "Yi Gui", "Zhen Li", "Yao Wan", "Pan Zhou", "Jianfeng Gao", "Lichao Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9997c3ab43f0e3e69ad4dcbe15257ea4392d36fa3c82f90c6b7730667a067f99", "submitted_at": "2024-06-16T06:56:53Z", "updated_at": "2025-03-24T11:46:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2406.10819v2"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2406.10819", "identifiers": {"arxiv": "2406.10819"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.10819"}], "name": "GUI-World: A Video Benchmark and Dataset for Multimodal GUI-oriented Understanding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces video-based GUI questions and annotations for studying temporal interface understanding across varied interaction settings.", "topics": ["evaluation", "video", "screen-understanding"], "updated_at": "2025-03-24", "url": "https://arxiv.org/abs/2406.10819"}
{"aliases": [], "authors": ["Yichen Pan", "Dehan Kong", "Sida Zhou", "Cheng Cui", "Yifei Leng", "Bing Jiang", "Hangyu Liu", "Yanyi Shang", "Shuyan Zhou", "Tongshuang Wu", "Zhengyang Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c5d0fbea7a4813a1da85c9ecb237c2115cc183b4c9e6fd0cbfd63764b0d418e3", "submitted_at": "2024-06-18T07:58:33Z", "updated_at": "2024-07-16T06:19:21Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2406.12373v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2406.12373", "identifiers": {"arxiv": "2406.12373"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.12373"}], "name": "WebCanvas: Benchmarking Web Agents in Online Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebCanvas and Mind2Web-Live, evaluating online task progress through meaningful intermediate states on changing websites.", "topics": ["evaluation", "web-navigation"], "updated_at": "2024-07-16", "url": "https://arxiv.org/abs/2406.12373"}
{"aliases": [], "authors": ["Chen Henry Wu", "Rishi Shah", "Jing Yu Koh", "Ruslan Salakhutdinov", "Daniel Fried", "Aditi Raghunathan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7cc1c1ac93f8e328d33accf7cdaa1d65487c4f7abec15234b9e596a501947550", "submitted_at": "2024-06-18T17:32:48Z", "updated_at": "2025-02-04T20:02:17Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2406.12814v3"}, "publication": {"authors": ["Chen Wu", "Rishi Shah", "Jing Yu Koh", "Russ Salakhutdinov", "Daniel Fried", "Aditi Raghunathan"], "date_basis": "Explicit publisher citation_publication_date: 2025-05-01. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 5, 1], "doi": null, "event_dates": null, "evidence_urls": ["https://proceedings.iclr.cc/paper_files/paper/2025/hash/460a1d8eac34125dad453b28d6d64446-Abstract-Conference.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity. The normalized publisher abstract matches the retained arXiv abstract SHA-256.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "28362-28383", "pdf_url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/460a1d8eac34125dad453b28d6d64446-Paper-Conference.pdf", "publisher": null, "series": null, "title": "Dissecting Adversarial Robustness of Multimodal LM Agents", "type": "conference-paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/460a1d8eac34125dad453b28d6d64446-Abstract-Conference.html", "venue": "International Conference on Learning Representations", "volume": "2025"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:01Z", "role": "primary", "sha256": "fa23fda1b80dac019741737fc6f611fccf9234a4d73ffbb8e24e54d8caf4dac3", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/460a1d8eac34125dad453b28d6d64446-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2406.12814", "identifiers": {"arxiv": "2406.12814"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.12814"}, {"rel": "publication", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/460a1d8eac34125dad453b28d6d64446-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/460a1d8eac34125dad453b28d6d64446-Paper-Conference.pdf"}], "name": "Dissecting Adversarial Robustness of Multimodal LM Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2024-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces adversarial tasks on VisualWebArena and an evaluation framework that tracks how manipulated information passes between agent components. The analysis examines vulnerabilities introduced by reflection and tree-search evaluators as well as direct action generation.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": "2025-02-04", "url": "https://arxiv.org/abs/2406.12814"}
{"aliases": [], "authors": ["Michael Wornow", "Avanika Narayan", "Ben Viggiano", "Ishan S. Khare", "Tathagat Verma", "Tibor Thompson", "Miguel Angel Fuentes Hernandez", "Sudharsan Sundar", "Chloe Trujillo", "Krrish Chawla", "Rongfei Lu", "Justin Shen", "Divya Nagaraj", "Joshua Martinez", "Vardhan Agrawal", "Althea Hudson", "Nigam H. Shah", "Christopher Re"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a7e7d00030ea8c77b9ed6f950d62f2bd417d3870f24b2f4589d8068c9484e328", "submitted_at": "2024-06-19T06:50:15Z", "updated_at": "2024-10-11T00:06:31Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2406.13264v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:37Z", "role": "primary", "sha256": "a75a630c4d4403233df3cb296d7e6bb60ca2f30705b5fc7ed4bd53b188c394ca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.10290%2C2406.10707%2C2406.11342%2C2406.11775%2C2406.12045%2C2406.12275%2C2406.12934%2C2406.13264%2C2406.13352%2C2406.14884%2C2406.14991%2C2406.18820%2C2406.20041%2C2406.20092%2C2407.01502%2C2407.03502%2C2407.04363%2C2407.04856%2C2407.06048%2C2407.08898%2C2407.09025%2C2407.10839%2C2407.12821%2C2407.13696%2C2407.17789%2C2407.20143%2C2407.20727%2C2408.00545%2C2408.02248%2C2408.04168%2C2408.07720%2C2408.09559%2C2408.10215%2C2408.10605%2C2408.11791%2C2408.16119%2C2409.00608%2C2409.03402%2C2409.04999%2C2409.05655%2C2409.07703%2C2409.08273%2C2409.08831%2C2409.10033%2C2409.17652%2C2409.19715%2C2410.01039%2C2410.01606%2C2410.05340%2C2410.05656&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:15:30Z", "role": "primary", "sha256": "95c08b3a28f77a248d6f990345ef4b2b16cbf196492920bfa1ce0ec7e398345e", "supports": ["name", "kind", "summary", "topics", "notes"], "url": "https://arxiv.org/abs/2406.13264v2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:16:02Z", "role": "primary", "sha256": "07de7eac76877364a2d9e9d31f6c65f5f2cace2f97f7821862a1441f83d9f95c", "supports": ["summary", "environments", "scope", "notes"], "url": "https://arxiv.org/html/2406.13264v2"}], "id": "arxiv:2406.13264", "identifiers": {"arxiv": "2406.13264"}, "kind": "paper", "last_verified_at": "2026-10-17T09:15:30Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.13264"}], "name": "WONDERBREAD: A Benchmark for Evaluating Multimodal Foundation Models on Business Process Management Tasks", "notes": "Reviewed against the exact-version official arXiv abstract page; its normalized abstract SHA-256 matches the retained bibliographic snapshot. The original API response body was not locally available for this review. Experimental results and artifact availability have not been independently reproduced. The demonstrations were rehearsed, so their completion rate is not a zero-shot human performance estimate.", "published_at": "2024-06-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WONDERBREAD, a benchmark pairing human WebArena workflow demonstrations with screen recordings, action logs, and written procedures. Six evaluation tasks cover documentation, knowledge transfer, and process improvement beyond end-to-end execution.", "topics": ["evaluation", "workflow", "video"], "updated_at": "2024-10-11", "url": "https://arxiv.org/abs/2406.13264"}
{"aliases": [], "authors": ["Ke Wang", "Tianyu Xia", "Zhangxuan Gu", "Yi Zhao", "Shuheng Shen", "Changhua Meng", "Weiqiang Wang", "Ke Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e4d5a46d2141ca59fcf221bbe618601cb7cce29c1b067ae60712e53e907095bd", "submitted_at": "2024-06-20T12:22:05Z", "updated_at": "2024-07-01T08:37:14Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2406.14250v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2406.14250", "identifiers": {"arxiv": "2406.14250"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.14250"}], "name": "E-ANT: A Large-Scale Dataset for Efficient Automatic GUI NavigaTion", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Chinese-language GUI navigation traces and annotated screenshots collected from human interactions with mini-applications.", "topics": ["mobile-navigation", "evaluation"], "updated_at": "2024-07-01", "url": "https://arxiv.org/abs/2406.14250"}
{"aliases": [], "authors": ["Omri Berkovitch", "Sapir Caduri", "Noam Kahlon", "Anatoly Efros", "Avi Caciularu", "Ido Dagan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a0bc798b69e59d1ec0ca8ecb0fab7a7ee6594f4d656ca4212b19be7e3085c0c5", "submitted_at": "2024-06-20T13:46:10Z", "updated_at": "2025-03-03T15:47:10Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2406.14314v3"}}, "environments": ["android", "web", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2406.14314", "identifiers": {"arxiv": "2406.14314"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.14314"}], "name": "Identifying User Goals from UI Trajectories", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Defines goal identification from observed UI trajectories and an evaluation method for comparing intent descriptions within the same interface context. Experiments reuse Android and web automation data to compare human and model inference of users’ detailed goals.", "topics": ["evaluation", "interaction-history", "screen-understanding"], "updated_at": "2025-03-03", "url": "https://arxiv.org/abs/2406.14314"}
{"aliases": [], "authors": ["Shihan Deng", "Weikai Xu", "Hongda Sun", "Wei Liu", "Tao Tan", "Jianfeng Liu", "Ang Li", "Jian Luan", "Bin Wang", "Rui Yan", "Shuo Shang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "90ce9900644c21004e0ba9e749e10d94238d2447d0e84bd7205ff55dcbd243f4", "submitted_at": "2024-07-01T06:10:01Z", "updated_at": "2024-07-01T06:10:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.00993v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.00993", "identifiers": {"arxiv": "2407.00993"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.00993"}], "name": "Mobile-Bench: An Evaluation Benchmark for LLM-based Mobile Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates mobile agents using combined UI and API actions, tasks with varying app scope, and checkpoints for intermediate progress.", "topics": ["evaluation", "cross-app", "tool-integration"], "updated_at": "2024-07-01", "url": "https://arxiv.org/abs/2407.00993"}
{"aliases": [], "authors": ["Jing Yu Koh", "Stephen McAleer", "Daniel Fried", "Ruslan Salakhutdinov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ec7aaf2a34e38f72da652bb2d8350c951229c154e2bc368a927a79287b36c439", "submitted_at": "2024-07-01T17:07:55Z", "updated_at": "2026-02-08T15:06:40Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2407.01476v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.01476", "identifiers": {"arxiv": "2407.01476"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.01476"}], "name": "Tree Search for Language Model Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2024-07-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Applies best-first search to agent actions within interactive web environments, allowing exploration and multistep planning at inference time. WebArena and VisualWebArena experiments examine search behavior and test-time computation.", "topics": ["planning", "web-navigation", "evaluation"], "updated_at": "2026-02-08", "url": "https://arxiv.org/abs/2407.01476"}
{"aliases": [], "authors": ["Tianqi Xu", "Linyao Chen", "Dai-Jie Wu", "Yanjun Chen", "Zecheng Zhang", "Xiang Yao", "Zhiqiang Xie", "Yongchao Chen", "Shilong Liu", "Bochen Qian", "Anjie Yang", "Zhaoxuan Jin", "Jianbo Deng", "Philip Torr", "Bernard Ghanem", "Guohao Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "12e5464ea0fc62bd566f19d301b887826b7ff317cfc27c86fa0cda8d3de271d9", "submitted_at": "2024-07-01T17:55:04Z", "updated_at": "2025-07-20T13:42:07Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2407.01511v4"}, "publication": {"authors": ["Tianqi Xu", "Linyao Chen", "Dai-Jie Wu", "Yanjun Chen", "Zecheng Zhang", "Xiang Yao", "Zhiqiang Xie (谢志强)", "Yongchao Chen", "Shilong Liu", "Bochen Qian", "Anjie Yang", "Zhaoxuan Jin", "Jianbo Deng", "Philip Torr", "Bernard Ghanem", "Guohao Li"], "date_basis": "Explicit publisher citation_publication_date: 2025/7. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 7], "doi": "10.18653/v1/2025.findings-acl.1113", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.findings-acl.1113/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "21607-21647", "pdf_url": "https://aclanthology.org/2025.findings-acl.1113.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "CRAB: Cross-environment Agent Benchmark for Multimodal Language Model Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.findings-acl.1113/", "venue": "Findings of the Association for Computational Linguistics: ACL 2025", "volume": null}}, "environments": ["desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:51Z", "role": "primary", "sha256": "a8cf178f2faa088fd94b77d03785b8c6e2908db34b411287984066b4e5d1ef3f", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.findings-acl.1113/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.01511", "identifiers": {"arxiv": "2407.01511"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.01511"}, {"rel": "publication", "url": "https://aclanthology.org/2025.findings-acl.1113/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.findings-acl.1113.pdf"}], "name": "CRAB: Cross-environment Agent Benchmark for Multimodal Language Model Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CRAB, a framework for cross-environment agent tasks with graph-based evaluation and a benchmark spanning desktop and phone interactions.", "topics": ["evaluation", "cross-app"], "updated_at": "2025-07-20", "url": "https://arxiv.org/abs/2407.01511"}
{"aliases": [], "authors": ["El Hassane Ettifouri", "Jessica López Espejel", "Laura Minkova", "Tassnim Dardouri", "Walid Dahhane"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "01d3b3a013f3b10249b8197a8e18aa695bdc20febaa34f07128791f36afdcbac", "submitted_at": "2024-05-05T19:10:19Z", "updated_at": "2025-07-18T11:35:01Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2407.01558v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.01558", "identifiers": {"arxiv": "2407.01558"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.01558"}], "name": "Visual Grounding Methods for Efficient Interaction with Desktop Graphical User Interfaces", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-05-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares modular OCR-based and end-to-end methods for locating desktop GUI targets from natural-language instructions.", "topics": ["grounding", "evaluation"], "updated_at": "2025-07-18", "url": "https://arxiv.org/abs/2407.01558"}
{"aliases": [], "authors": ["Zhe Liu", "Cheng Li", "Chunyang Chen", "Junjie Wang", "Mengzhuo Chen", "Boyu Wu", "Yawen Wang", "Jun Hu", "Qing Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73b706ac6089a7d19836268437f3fea58d10b9d170ac596ac0d5a12f0d3a9668", "submitted_at": "2024-07-03T11:58:09Z", "updated_at": "2024-12-04T08:00:42Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2407.03037v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.03037", "identifiers": {"arxiv": "2407.03037"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.03037"}], "name": "Seeing is Believing: Vision-driven Non-crash Functional Bug Detection for Mobile Apps", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trident combines exploration, monitoring, and detection agents to identify functional bugs that do not crash a mobile app. It reasons over screenshot sequences and interface transitions to construct test oracles, evaluated on 590 non-crash bugs and additional live apps.", "topics": ["automation", "evaluation", "screen-understanding", "state-tracking"], "updated_at": "2024-12-04", "url": "https://arxiv.org/abs/2407.03037"}
{"aliases": [], "authors": ["Léo Boisvert", "Megh Thakkar", "Maxime Gasse", "Massimo Caccia", "Thibault Le Sellier De Chezelles", "Quentin Cappart", "Nicolas Chapados", "Alexandre Lacoste", "Alexandre Drouin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2b56a6208e4fa3523cae3fbae673affcd5fe3d6a9c19df482d27fe71d43ffef5", "submitted_at": "2024-07-07T07:15:49Z", "updated_at": "2025-02-05T21:50:07Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2407.05291v2"}, "publication": {"authors": ["Léo Boisvert", "Megh Thakkar", "Maxime Gasse", "Massimo Caccia", "Thibault L. De Chezelles", "Quentin Cappart", "Nicolas Chapados", "Alexandre Lacoste", "Alexandre Drouin"], "date_basis": "Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 12, 16], "doi": "10.52202/079017-0195", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/0b82662b6c32e887bb252a74d8cb2d5e-Abstract-Datasets_and_Benchmarks_Track.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "5996-6051", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/0b82662b6c32e887bb252a74d8cb2d5e-Paper-Datasets_and_Benchmarks_Track.pdf", "publisher": null, "series": null, "title": "WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/0b82662b6c32e887bb252a74d8cb2d5e-Abstract-Datasets_and_Benchmarks_Track.html", "venue": "Advances in Neural Information Processing Systems", "volume": "37"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:04Z", "role": "primary", "sha256": "4fe01789f93df840699d604de87c068557fc02aa78d22d53c6b3ae7c9db69cc4", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/0b82662b6c32e887bb252a74d8cb2d5e-Abstract-Datasets_and_Benchmarks_Track.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.05291", "identifiers": {"arxiv": "2407.05291"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.05291"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/0b82662b6c32e887bb252a74d8cb2d5e-Abstract-Datasets_and_Benchmarks_Track.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/0b82662b6c32e887bb252a74d8cb2d5e-Paper-Datasets_and_Benchmarks_Track.pdf"}], "name": "WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-07-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Defines compositional enterprise workflows to evaluate planning and reasoning, with generation of ground-truth interaction traces.", "topics": ["evaluation", "planning", "workflow"], "updated_at": "2025-02-05", "url": "https://arxiv.org/abs/2407.05291"}
{"aliases": [], "authors": ["Peitong Duan", "Chin-yi Chen", "Gang Li", "Bjoern Hartmann", "Yang Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "390a7415ad5f665de43f7d294e0a575785ef7e2810dead3363632174b2986af4", "submitted_at": "2024-07-11T20:18:19Z", "updated_at": "2024-08-13T23:41:43Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2407.08850v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.08850", "identifiers": {"arxiv": "2407.08850"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.08850"}], "name": "UICrit: Enhancing Automated Design Evaluation with a UICritique Dataset", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-11", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces designer-authored mobile UI critiques and quality ratings, studying how targeted examples improve automated design feedback.", "topics": ["screen-understanding", "evaluation"], "updated_at": "2024-08-13", "url": "https://arxiv.org/abs/2407.08850"}
{"aliases": [], "authors": ["Yongxiang Hu", "Xuan Wang", "Yingchuan Wang", "Yu Zhang", "Shiyu Guo", "Chaoyi Chen", "Xin Wang", "Yangfan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a594fa7e150c9f5c22ed4d88f0551305f801f5780851ecf844802f2d2cd07c48", "submitted_at": "2024-07-12T06:14:46Z", "updated_at": "2024-07-12T06:14:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.09018v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.09018", "identifiers": {"arxiv": "2407.09018"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.09018"}], "name": "AUITestAgent: Automatic Requirements Oriented GUI Function Testing", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Automates mobile GUI tests from natural-language requirements, combining action generation with verification based on interaction traces.", "topics": ["automation", "evaluation"], "updated_at": "2024-07-12", "url": "https://arxiv.org/abs/2407.09018"}
{"aliases": [], "authors": ["Yulong Yang", "Xinshan Yang", "Shuaidong Li", "Chenhao Lin", "Zhengyu Zhao", "Chao Shen", "Tianwei Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "88576325385d3c361f0666311eae8858ea0d4a0098d73221f789550bebf1e5db", "submitted_at": "2024-07-12T14:30:05Z", "updated_at": "2025-03-16T07:13:53Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2407.09295v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2407.09295", "identifiers": {"arxiv": "2407.09295"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.09295"}], "name": "Systematic Categorization, Construction and Evaluation of New Attacks against Multi-modal Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Develops a threat taxonomy and evaluation framework for attacks against multimodal mobile GUI agents, with case studies and dataset experiments.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-03-16", "url": "https://arxiv.org/abs/2407.09295"}
{"aliases": [], "authors": ["Ruisheng Cao", "Fangyu Lei", "Haoyuan Wu", "Jixuan Chen", "Yeqiao Fu", "Hongcheng Gao", "Xinzhuang Xiong", "Hanchong Zhang", "Yuchen Mao", "Wenjing Hu", "Tianbao Xie", "Hongshen Xu", "Danyang Zhang", "Sida Wang", "Ruoxi Sun", "Pengcheng Yin", "Caiming Xiong", "Ansong Ni", "Qian Liu", "Victor Zhong", "Lu Chen", "Kai Yu", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7922a723142ac62018064fe5fc9357fb77c08e316ac4d6d6ba728f1b337df690", "submitted_at": "2024-07-15T17:54:37Z", "updated_at": "2024-07-15T17:54:37Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.10956v1"}, "publication": {"authors": ["Ruisheng Cao", "Fangyu Lei", "Haoyuan Wu", "Jixuan Chen", "Yeqiao Fu", "Hongcheng Gao", "Xinzhuang Xiong", "Hanchong Zhang", "Yuchen Mao", "Wenjing Hu", "Tianbao Xie", "Hongshen Xu", "Danyang Zhang", "Sida Wang", "Ruoxi Sun", "Pengcheng Yin", "Caiming Xiong", "Ansong Ni", "Qian Liu", "Victor Zhong", "Lu Chen", "Kai Yu", "Tao Yu"], "date_basis": "Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 12, 16], "doi": "10.52202/079017-3421", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity. The normalized publisher abstract matches the retained arXiv abstract SHA-256.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "107703-107744", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/c2f71567cd53464161cab3336e8fc865-Paper-Datasets_and_Benchmarks_Track.pdf", "publisher": null, "series": null, "title": "Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html", "venue": "Advances in Neural Information Processing Systems", "volume": "37"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:04Z", "role": "primary", "sha256": "872286a0210616be9737de716cc6069a1e9cbee1df3182d45c718582e185f62f", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.10956", "identifiers": {"arxiv": "2407.10956"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.10956"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/c2f71567cd53464161cab3336e8fc865-Paper-Datasets_and_Benchmarks_Track.pdf"}], "name": "Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Benchmarks multimodal agents on data-science and engineering workflows involving enterprise applications, code, and GUI operations in real computer environments.", "topics": ["evaluation", "science", "workflow"], "updated_at": "2024-07-15", "url": "https://arxiv.org/abs/2407.10956"}
{"aliases": [], "authors": ["Ori Yoran", "Samuel Joseph Amouyal", "Chaitanya Malaviya", "Ben Bogin", "Ofir Press", "Jonathan Berant"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2f028fb7f4d64fd4c29886d49b5d6f0fabddba309b669e019edf13586deb9c90", "submitted_at": "2024-07-22T15:18:45Z", "updated_at": "2024-10-21T15:45:31Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2407.15711v2"}, "publication": {"authors": ["Ori Yoran", "Samuel Joseph Amouyal", "Chaitanya Malaviya", "Ben Bogin", "Ofir Press", "Jonathan Berant"], "date_basis": "Explicit publisher citation_publication_date: 2024/11. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 11], "doi": "10.18653/v1/2024.emnlp-main.505", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2024.emnlp-main.505/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "8938-8968", "pdf_url": "https://aclanthology.org/2024.emnlp-main.505.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "AssistantBench: Can Web Agents Solve Realistic and Time-Consuming Tasks?", "type": "conference-paper", "url": "https://aclanthology.org/2024.emnlp-main.505/", "venue": "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:51Z", "role": "primary", "sha256": "a18f0ea465d83796bcd83b95f48c3c13cbdc0aa778ea0dd454aa61f8082920b4", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2024.emnlp-main.505/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2407.15711", "identifiers": {"arxiv": "2407.15711"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.15711"}, {"rel": "publication", "url": "https://aclanthology.org/2024.emnlp-main.505/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2024.emnlp-main.505.pdf"}], "name": "AssistantBench: Can Web Agents Solve Realistic and Time-Consuming Tasks?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces realistic open-web tasks and the SeePlanAct agent, analyzing the limitations of browsing and retrieval in time-consuming assistant work.", "topics": ["evaluation", "web-navigation", "planning"], "updated_at": "2024-10-21", "url": "https://arxiv.org/abs/2407.15711"}
{"aliases": [], "authors": ["Harsh Trivedi", "Tushar Khot", "Mareike Hartmann", "Ruskin Manku", "Vinty Dong", "Edward Li", "Shashank Gupta", "Ashish Sabharwal", "Niranjan Balasubramanian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e59087ecf28d07baf81b60e4166a866d4bea4d8cd934bea06c2149dd61f38f1a", "submitted_at": "2024-07-26T17:55:45Z", "updated_at": "2024-07-26T17:55:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.18901v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.18901", "identifiers": {"arxiv": "2407.18901"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.18901"}], "name": "AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-26", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces simulated application APIs and state-based task evaluation for interactive coding agents performing everyday digital workflows.", "topics": ["evaluation", "tool-integration", "cross-app"], "updated_at": "2024-07-26", "url": "https://arxiv.org/abs/2407.18901"}
{"aliases": [], "authors": ["Zilong Wang", "Yuedong Cui", "Li Zhong", "Zimin Zhang", "Da Yin", "Bill Yuchen Lin", "Jingbo Shang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c443e6a508c438332ac5990cf254a4441f5bdcc24354fd8aeca36bcbf4604a29", "submitted_at": "2024-07-26T19:27:17Z", "updated_at": "2024-07-26T19:27:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.19056v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.19056", "identifiers": {"arxiv": "2407.19056"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.19056"}], "name": "OfficeBench: Benchmarking Language Agents across Multiple Applications for Office Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates language agents on office workflows requiring information integration, long-term planning, and switching between applications.", "topics": ["evaluation", "cross-app", "long-horizon"], "updated_at": "2024-07-26", "url": "https://arxiv.org/abs/2407.19056"}
{"aliases": [], "authors": ["Xinbei Ma", "Yiting Wang", "Yao Yao", "Tongxin Yuan", "Aston Zhang", "Zhuosheng Zhang", "Hai Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d66c0c2102dd2734d5334faeb12b4971d50a2d2e50a6671ede2fa78a84a0c03f", "submitted_at": "2024-08-05T15:16:22Z", "updated_at": "2025-09-05T09:21:10Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2408.02544v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2408.02544", "identifiers": {"arxiv": "2408.02544"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2408.02544"}], "name": "Caution for the Environment: Multimodal LLM Agents are Susceptible to Environmental Distractions", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The abstract describes a simulated dataset; real-browser deployment is not asserted.", "published_at": "2024-08-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Measures how irrelevant GUI content affects multimodal agents using a simulated dataset and three perception patterns. The study also examines adversarial environment injection and possible ways to improve instruction faithfulness.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": "2025-09-05", "url": "https://arxiv.org/abs/2408.02544"}
{"aliases": [], "authors": ["Xiao Liu", "Tianjie Zhang", "Yu Gu", "Iat Long Iong", "Yifan Xu", "Xixuan Song", "Shudan Zhang", "Hanyu Lai", "Xinyi Liu", "Hanlin Zhao", "Jiadai Sun", "Xinyue Yang", "Yu Yang", "Zehan Qi", "Shuntian Yao", "Xueqiao Sun", "Siyi Cheng", "Qinkai Zheng", "Hao Yu", "Hanchen Zhang", "Wenyi Hong", "Ming Ding", "Lihang Pan", "Xiaotao Gu", "Aohan Zeng", "Zhengxiao Du", "Chan Hee Song", "Yu Su", "Yuxiao Dong", "Jie Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "603bfa4357617bfeb7a718065e77c71d77614f7240258684938dd4e6fd1a20df", "submitted_at": "2024-08-12T17:44:17Z", "updated_at": "2024-08-12T17:44:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2408.06327v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2408.06327", "identifiers": {"arxiv": "2408.06327"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2408.06327"}], "name": "VisualAgentBench: Towards Large Multimodal Models as Visual Foundation Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-08-12", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Benchmarks visual agents across GUI, embodied, and design tasks, with training trajectories collected through solvers, agents, and human demonstrations.", "topics": ["evaluation", "data-generation"], "updated_at": "2024-08-12", "url": "https://arxiv.org/abs/2408.06327"}
{"aliases": [], "authors": ["Segev Shlomov", "Ben wiesel", "Aviad Sela", "Ido Levy", "Liane Galanti", "Roy Abitbol"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cdb228e652218d7f8981d9d16ca29b465004675c510d1e9c5c7b543bec9dc454", "submitted_at": "2024-09-03T14:17:09Z", "updated_at": "2024-09-03T14:17:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2409.01927v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2409.01927", "identifiers": {"arxiv": "2409.01927"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2409.01927"}], "name": "From Grounding to Planning: Benchmarking Bottlenecks in Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-09-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Separates web-agent planning and grounding evaluations on Mind2Web to analyze which components limit task performance.", "topics": ["evaluation", "grounding", "planning"], "updated_at": "2024-09-03", "url": "https://arxiv.org/abs/2409.01927"}
{"aliases": [], "authors": ["Rogerio Bonatti", "Dan Zhao", "Francesco Bonacci", "Dillon Dupont", "Sara Abdali", "Yinheng Li", "Yadong Lu", "Justin Wagle", "Kazuhito Koishida", "Arthur Bucker", "Lawrence Jang", "Zack Hui"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fe77f7f8ffd7490192ab54b7f6d49c0cb5b10be6df0a645bfb181b3be724118a", "submitted_at": "2024-09-12T17:56:43Z", "updated_at": "2024-09-13T20:17:13Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2409.08264v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2409.08264", "identifiers": {"arxiv": "2409.08264"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2409.08264"}], "name": "Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-09-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Adapts real-desktop evaluation to Windows applications and supports parallel execution of benchmark tasks.", "topics": ["evaluation", "cross-app"], "updated_at": "2024-09-13", "url": "https://arxiv.org/abs/2409.08264"}
{"aliases": [], "authors": ["Zeyi Liao", "Lingbo Mo", "Chejian Xu", "Mintong Kang", "Jiawei Zhang", "Chaowei Xiao", "Yuan Tian", "Bo Li", "Huan Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "897cdbb6aa52478b260ecba830e0a13527c2614454f7c766c9f5c8c9bdbfd4e3", "submitted_at": "2024-09-17T15:49:44Z", "updated_at": "2025-03-12T20:54:00Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2409.11295v5"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2409.11295", "identifiers": {"arxiv": "2409.11295"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2409.11295"}], "name": "EIA: Environmental Injection Attack on Generalist Web Agents for Privacy Leakage", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-09-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates privacy-directed webpage instruction injection against generalist web agents using PII-related interaction steps from Mind2Web, including attempts to expose specific user information or entire requests.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2025-03-12", "url": "https://arxiv.org/abs/2409.11295"}
{"aliases": [], "authors": ["Wayne Chi", "Ameet Talwalkar", "Chris Donahue"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a2f59976cec7adb9f9c88bd79bdfbeed040e4874b6d9ab0c229136f98101aa98", "submitted_at": "2024-09-18T16:04:10Z", "updated_at": "2024-10-06T21:25:34Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2409.12089v3"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}], "id": "arxiv:2409.12089", "identifiers": {"arxiv": "2409.12089"}, "kind": "paper", "last_verified_at": "2026-10-16T06:42:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2409.12089"}], "name": "The Impact of Element Ordering on LM Agent Performance", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2024-09-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies how the ordering of UI elements in an agent observation affects web and desktop performance. A pixel-based element detector and dimensionality-reduction ordering are evaluated on OmniACT when document hierarchy is unavailable.", "topics": ["screen-understanding", "grounding", "evaluation"], "updated_at": "2024-10-06", "url": "https://arxiv.org/abs/2409.12089"}
{"aliases": [], "authors": ["Maria Wang", "Srinivas Sunkara", "Gilles Baechler", "Jason Lin", "Yun Zhu", "Fedir Zubach", "Lei Shu", "Jindong Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1ca44e6c98c09a2261b2cf8f3b6d18584b2d2471934b2dcac8d5b72c17b32e61", "submitted_at": "2024-09-06T18:44:25Z", "updated_at": "2024-09-24T18:38:02Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2409.13711v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2409.13711", "identifiers": {"arxiv": "2409.13711"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2409.13711"}], "name": "WebQuest: A Benchmark for Multimodal QA on Web Page Sequences", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-09-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces question answering over individual screens, multiple web pages, and navigation traces to test information synthesis across views.", "topics": ["evaluation", "screen-understanding"], "updated_at": "2024-09-24", "url": "https://arxiv.org/abs/2409.13711"}
{"aliases": [], "authors": ["Sabrina Haque", "Christoph Csallner"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2d927b7644f0c6b3a49fbdf1aa976a00f2409adf9a72e1c04333438f66f1014b", "submitted_at": "2024-09-26T17:01:33Z", "updated_at": "2024-10-07T19:04:38Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2409.18060v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}], "id": "arxiv:2409.18060", "identifiers": {"arxiv": "2409.18060"}, "kind": "paper", "last_verified_at": "2026-10-16T06:45:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2409.18060"}], "name": "Inferring Alt-text For UI Icons With Large Language Models During App Development", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. This is a UI accessibility and element-description method; autonomous GUI control is not evaluated.", "published_at": "2024-09-26", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces IconDesc, which generates mobile icon alt-text from partial UI context such as element identifiers, bounds, nearby text and hierarchy. Model evaluation and a user study examine the resulting accessibility descriptions.", "topics": ["screen-understanding", "ui-design", "evaluation"], "updated_at": "2024-10-07", "url": "https://arxiv.org/abs/2409.18060"}
{"aliases": [], "authors": ["Ruhana Azam", "Tamer Abuelsaad", "Aditya Vempaty", "Ashish Jagmohan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "69febc7aabfe38a6da439f1b4f673e0c36bc34c6afd1e8b06481433c1f43b1aa", "submitted_at": "2024-10-01T13:43:55Z", "updated_at": "2024-10-11T15:42:52Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2410.00689v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2410.00689", "identifiers": {"arxiv": "2410.00689"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.00689"}], "name": "Multimodal Auto Validation For Self-Refinement in Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies text and visual validation for web agents and uses the resulting validator to detect failures and guide workflow self-refinement.", "topics": ["evaluation", "robustness", "web-navigation"], "updated_at": "2024-10-11", "url": "https://arxiv.org/abs/2410.00689"}
{"aliases": [], "authors": ["Ido Levy", "Ben Wiesel", "Sami Marreed", "Alon Oved", "Avi Yaeli", "Nir Mashkif", "Segev Shlomov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ba2b599c2b3f09974ee41d329545b69773a3b408b6db0124c5b1073d5fe8f1cf", "submitted_at": "2024-10-09T09:13:38Z", "updated_at": "2026-06-04T09:59:28Z", "version": 7, "versioned_url": "https://arxiv.org/abs/2410.06703v7"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2410.06703", "identifiers": {"arxiv": "2410.06703"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.06703"}], "name": "ST-WebAgentBench: A Benchmark for Evaluating Safety and Trustworthiness in Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates web-task completion under explicit enterprise policies, measuring policy-compliant completion and safety violations separately.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-06-04", "url": "https://arxiv.org/abs/2410.06703"}
{"aliases": [], "authors": ["Zhuohao Jerry Zhang", "Eldon Schoop", "Jeffrey Nichols", "Anuj Mahajan", "Amanda Swearngin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ff649a0792864bd40b7bf2f825f6a8a815eb60cc795ed4dc0c6b55a942477758", "submitted_at": "2024-10-11T17:24:00Z", "updated_at": "2025-03-22T18:01:37Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2410.09006v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.09006", "identifiers": {"arxiv": "2410.09006"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.09006"}], "name": "From Interaction to Impact: Towards Safer AI Agents Through Understanding and Evaluating Mobile UI Operation Impacts", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Develops a taxonomy of mobile-action consequences and annotated interaction data to evaluate whether models recognize impactful or risky UI operations.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2025-03-22", "url": "https://arxiv.org/abs/2410.09006"}
{"aliases": [], "authors": ["Priyanshu Kumar", "Elaine Lau", "Saranya Vijayakumar", "Tu Trinh", "Scale Red Team", "Elaine Chang", "Vaughn Robinson", "Sean Hendryx", "Shuyan Zhou", "Matt Fredrikson", "Summer Yue", "Zifan Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4922d9362b23bc0e41d47dacf590c6e9bd6e88cfa6418cf3517814131295270c", "submitted_at": "2024-10-11T06:54:12Z", "updated_at": "2024-10-21T18:06:58Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2410.13886v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2410.13886", "identifiers": {"arxiv": "2410.13886"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.13886"}], "name": "Refusal-Trained LLMs Are Easily Jailbroken As Browser Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces BrowserART and studies whether chat-model refusal behavior transfers to browser agents carrying out potentially harmful instructions.", "topics": ["safety-security", "evaluation"], "updated_at": "2024-10-21", "url": "https://arxiv.org/abs/2410.13886"}
{"aliases": [], "authors": ["Jingxuan Chen", "Derek Yuen", "Bin Xie", "Yuhao Yang", "Gongwei Chen", "Zhihao Wu", "Li Yixing", "Xurui Zhou", "Weiwen Liu", "Shuai Wang", "Kaiwen Zhou", "Rui Shao", "Liqiang Nie", "Yasheng Wang", "Jianye Hao", "Jun Wang", "Kun Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "206059754652205f1e036fac8ee117f71dd356bfe2d35fd84aff8d2935ea242d", "submitted_at": "2024-10-19T17:28:48Z", "updated_at": "2025-03-31T20:39:17Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2410.15164v3"}, "publication": {"authors": ["Jingxuan Chen", "Derek Yuen", "Bin Xie", "Yuhao Yang", "Gongwei Chen", "Zhihao Wu", "Li Yixing", "Xurui Zhou", "Weiwen Liu", "Shuai Wang", "Kaiwen Zhou", "Rui Shao", "Liqiang Nie", "Yasheng Wang", "Jianye HAO", "Jun Wang", "Kun Shao"], "date_basis": "Explicit publisher citation_publication_date: 2025-05-01. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 5, 1], "doi": null, "event_dates": null, "evidence_urls": ["https://proceedings.iclr.cc/paper_files/paper/2025/hash/9a75f4dd9679aa4ff80a4e6f0a1dc700-Abstract-Conference.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "61769-61806", "pdf_url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/9a75f4dd9679aa4ff80a4e6f0a1dc700-Paper-Conference.pdf", "publisher": null, "series": null, "title": "SPA-BENCH: A COMPREHENSIVE BENCHMARK FOR SMARTPHONE AGENT EVALUATION", "type": "conference-paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/9a75f4dd9679aa4ff80a4e6f0a1dc700-Abstract-Conference.html", "venue": "International Conference on Learning Representations", "volume": "2025"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:02Z", "role": "primary", "sha256": "56efdfea12b96edd94ad366f941d57668cdb75f6889fc03297e749497ff03e92", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/9a75f4dd9679aa4ff80a4e6f0a1dc700-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.15164", "identifiers": {"arxiv": "2410.15164"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.15164"}, {"rel": "publication", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/9a75f4dd9679aa4ff80a4e6f0a1dc700-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/9a75f4dd9679aa4ff80a4e6f0a1dc700-Paper-Conference.pdf"}], "name": "SPA-Bench: A Comprehensive Benchmark for SmartPhone Agent Evaluation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a smartphone-agent benchmark with English and Chinese tasks, real-time Android interaction, and completion and resource-use metrics.", "topics": ["evaluation", "efficiency", "mobile-navigation"], "updated_at": "2025-03-31", "url": "https://arxiv.org/abs/2410.15164"}
{"aliases": [], "authors": ["Yueqi Song", "Frank Xu", "Shuyan Zhou", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0944387dd3c65ea7b2f5ca52721496cf10b3c9860b303063201e7811318f4950", "submitted_at": "2024-10-21T19:46:06Z", "updated_at": "2025-06-16T22:23:37Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2410.16464v3"}, "publication": {"authors": ["Yueqi Song", "Frank F. Xu", "Shuyan Zhou", "Graham Neubig"], "date_basis": "Explicit publisher citation_publication_date: 2025/7. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 7], "doi": "10.18653/v1/2025.findings-acl.577", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.findings-acl.577/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "11066-11085", "pdf_url": "https://aclanthology.org/2025.findings-acl.577.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "Beyond Browsing: API-Based Web Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.findings-acl.577/", "venue": "Findings of the Association for Computational Linguistics: ACL 2025", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:55Z", "role": "primary", "sha256": "12c4aaa6eb73e55e905fb507db47a14cc245c933504453365b06242c39e92481", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.findings-acl.577/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2410.16464", "identifiers": {"arxiv": "2410.16464"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.16464"}, {"rel": "publication", "url": "https://aclanthology.org/2025.findings-acl.577/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.findings-acl.577.pdf"}], "name": "Beyond Browsing: API-Based Web Agents", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-10-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares browsing-only, API-only, and hybrid web agents on WebArena, examining how access to application APIs changes task execution alongside browser interaction.", "topics": ["web-navigation", "tool-integration", "evaluation"], "updated_at": "2025-06-16", "url": "https://arxiv.org/abs/2410.16464"}
{"aliases": [], "authors": ["Chejian Xu", "Mintong Kang", "Jiawei Zhang", "Zeyi Liao", "Lingbo Mo", "Mengqi Yuan", "Huan Sun", "Bo Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "226a4acbbaab42c8cbc44a897651d8c745119426ea1223f1b849558af32fab02", "submitted_at": "2024-10-22T20:18:26Z", "updated_at": "2025-05-31T18:34:01Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2410.17401v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2410.17401", "identifiers": {"arxiv": "2410.17401"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.17401"}], "name": "AdvAgent: Controllable Blackbox Red-teaming on Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a black-box web-agent red-teaming framework that learns adversarial prompts from agent feedback to study security weaknesses.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-05-31", "url": "https://arxiv.org/abs/2410.17401"}
{"aliases": [], "authors": ["Juyong Lee", "Dongyoon Hahm", "June Suk Choi", "W. Bradley Knox", "Kimin Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "362cbf7ea267ddc4a64505aa95c4f16a9f3631f95319ce10eef9981e2c85937b", "submitted_at": "2024-10-23T02:51:43Z", "updated_at": "2026-01-27T01:51:55Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2410.17520v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.17520", "identifiers": {"arxiv": "2410.17520"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.17520"}], "name": "MobileSafetyBench: Evaluating Safety of Autonomous Agents in Mobile Device Control", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-10-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates mobile agents on misuse, side effects, and indirect prompt-injection risks in an Android emulator environment.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-01-27", "url": "https://arxiv.org/abs/2410.17520"}
{"aliases": [], "authors": ["Lawrence Jang", "Yinheng Li", "Dan Zhao", "Charles Ding", "Justin Lin", "Paul Pu Liang", "Rogerio Bonatti", "Kazuhito Koishida"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6f413943bd5c2a395ddb5cf38a1a5ad61754b35a9eb5de43e8d980bd83e4adff", "submitted_at": "2024-10-24T19:03:01Z", "updated_at": "2025-02-15T05:19:38Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2410.19100v3"}, "publication": {"authors": ["Lawrence Jang", "Yinheng Li", "Dan Zhao", "Charles Ding", "Justin Lin", "Paul Pu Liang", "Rogerio Bonatti", "Kazuhito Koishida"], "date_basis": "Explicit publisher citation_publication_date: 2025-05-01. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 5, 1], "doi": null, "event_dates": null, "evidence_urls": ["https://proceedings.iclr.cc/paper_files/paper/2025/hash/5b555804d495321df2e3208cc27f4fbc-Abstract-Conference.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "36934-36958", "pdf_url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/5b555804d495321df2e3208cc27f4fbc-Paper-Conference.pdf", "publisher": null, "series": null, "title": "VideoWebArena:  Evaluating Long Context Multimodal Agents with Video Understanding Web Tasks", "type": "conference-paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/5b555804d495321df2e3208cc27f4fbc-Abstract-Conference.html", "venue": "International Conference on Learning Representations", "volume": "2025"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:02Z", "role": "primary", "sha256": "fdb42289bd00ad80cb51b66a01b91e13e60f134e809bb6c46181b4024ccabad2", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/5b555804d495321df2e3208cc27f4fbc-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.19100", "identifiers": {"arxiv": "2410.19100"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.19100"}, {"rel": "publication", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/5b555804d495321df2e3208cc27f4fbc-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/5b555804d495321df2e3208cc27f4fbc-Paper-Conference.pdf"}], "name": "VideoWebArena: Evaluating Long Context Multimodal Agents with Video Understanding Web Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates whether web agents can retain procedural skills and factual information from long video tutorials and use them during tasks.", "topics": ["evaluation", "video", "memory"], "updated_at": "2025-02-15", "url": "https://arxiv.org/abs/2410.19100"}
{"aliases": [], "authors": ["Haoyi Qiu", "Alexander R. Fabbri", "Divyansh Agarwal", "Kung-Hsiang Huang", "Sarah Tan", "Nanyun Peng", "Chien-Sheng Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "26fac89c78ab03fcbaf25bbfe93ebce090dec9adbf5c6dd5440641f154cffac9", "submitted_at": "2024-10-30T17:35:44Z", "updated_at": "2025-03-08T23:37:49Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2410.23252v3"}, "publication": {"authors": ["Haoyi Qiu", "Alexander Richard Fabbri", "Divyansh Agarwal", "Kung-Hsiang Huang", "Sarah Tan", "Nanyun Peng", "Chien-Sheng Wu"], "date_basis": "Explicit publisher citation_publication_date: 2025/4. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 4], "doi": "10.18653/v1/2025.findings-naacl.222", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.findings-naacl.222/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "3978-4005", "pdf_url": "https://aclanthology.org/2025.findings-naacl.222.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "Evaluating Cultural and Social Awareness of LLM Web Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.findings-naacl.222/", "venue": "Findings of the Association for Computational Linguistics: NAACL 2025", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:55Z", "role": "primary", "sha256": "6563b807356c30b69a1c5bb9328358c139d171777cb893079ddc123cb82a282e", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.findings-naacl.222/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2410.23252", "identifiers": {"arxiv": "2410.23252"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.23252"}, {"rel": "publication", "url": "https://aclanthology.org/2025.findings-naacl.222/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.findings-naacl.222.pdf"}], "name": "Evaluating Cultural and Social Awareness of LLM Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CASA to assess web agents' responses to cultural and social norms in shopping and forum tasks, including misleading observations.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-03-08", "url": "https://arxiv.org/abs/2410.23252"}
{"aliases": [], "authors": ["Yifan Xu", "Xiao Liu", "Xueqiao Sun", "Siyi Cheng", "Hao Yu", "Hanyu Lai", "Shudan Zhang", "Dan Zhang", "Jie Tang", "Yuxiao Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d29fe5262adf8677bbd5fead628c5fc28707af263aa7e800d1dee1f7e499784c", "submitted_at": "2024-10-31T15:25:20Z", "updated_at": "2024-11-04T05:57:31Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2410.24024v2"}, "publication": {"authors": ["Yifan Xu", "Xiao Liu", "Xueqiao Sun", "Siyi Cheng", "Hao Yu (余浩)", "Hanyu Lai", "Shudan Zhang", "Dan Zhang", "Jie Tang", "Yuxiao Dong"], "date_basis": "Explicit publisher citation_publication_date: 2025/7. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 7], "doi": "10.18653/v1/2025.acl-long.107", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.acl-long.107/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.", "pages": "2144-2166", "pdf_url": "https://aclanthology.org/2025.acl-long.107.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "AndroidLab: Training and Systematic Benchmarking of Android Autonomous Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.acl-long.107/", "venue": "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)", "volume": null}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:56Z", "role": "primary", "sha256": "52b82854b5f4703e6e7f3cd3e9fc49adcd011ca3ba0a8fd1aced77214c452043", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.acl-long.107/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.24024", "identifiers": {"arxiv": "2410.24024"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.24024"}, {"rel": "publication", "url": "https://aclanthology.org/2025.acl-long.107/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.acl-long.107.pdf"}], "name": "AndroidLab: Training and Systematic Benchmarking of Android Autonomous Agents", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2024-10-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Provides a shared Android action environment, reproducible tasks, and instruction data for training and comparing language and multimodal agents.", "topics": ["evaluation", "data-generation", "mobile-navigation"], "updated_at": "2024-11-04", "url": "https://arxiv.org/abs/2410.24024"}
{"aliases": [], "authors": ["Biao Wu", "Yanda Li", "Zhiwei Zhang", "Yunchao Wei", "Meng Fang", "Ling Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "de860a29f6c54317586c2218688436898f1d7a8b61207b1d5f4a65b05b3a4e6e", "submitted_at": "2024-11-04T11:50:58Z", "updated_at": "2025-09-14T09:13:39Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2411.02006v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:41Z", "role": "primary", "sha256": "0ba893415123ecc1763614ef5fc97e5c8bb7835f10b26b43f1b921ea860484cf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.06212%2C2410.06215%2C2410.07283%2C2410.07331%2C2410.07869%2C2410.08209%2C2410.08368%2C2410.10570%2C2410.12481%2C2410.12669%2C2410.15549%2C2410.16950%2C2410.21909%2C2410.22584%2C2410.22770%2C2410.23022%2C2410.23555%2C2411.02006%2C2411.02393%2C2411.02479%2C2411.03312%2C2411.03817%2C2411.05451%2C2411.11145%2C2411.11904%2C2411.11916%2C2411.12990%2C2411.14228%2C2411.15397%2C2412.01857%2C2412.04144%2C2412.04383%2C2412.05819%2C2412.06224%2C2412.06771%2C2412.09530%2C2412.09613%2C2412.10047%2C2412.10380%2C2412.10673%2C2412.12004%2C2412.13178%2C2412.13810%2C2412.14203%2C2412.14222%2C2412.16359%2C2412.19638%2C2412.20206%2C2501.03895%2C2501.05179&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:13:27Z", "role": "primary", "sha256": "bda47fcb119c581a798dbc791b4929af91832d5053ac073a02dab8fd4e10ace3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2411.02006v3"}], "id": "arxiv:2411.02006", "identifiers": {"arxiv": "2411.02006"}, "kind": "survey", "last_verified_at": "2026-10-17T09:13:27Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.02006"}], "name": "Foundations and Recent Trends in Multimodal Mobile Agents: A Survey", "notes": "Reviewed against the exact-version official arXiv abstract page; its normalized abstract SHA-256 matches the retained bibliographic snapshot. The original API response body was not locally available for this review. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-11-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Surveys multimodal mobile agents, organizing prompt-based and training-based approaches alongside evaluation benchmarks, supporting technologies, and open challenges in mobile task execution.", "topics": ["survey", "mobile-navigation", "evaluation"], "updated_at": "2025-09-14", "url": "https://arxiv.org/abs/2411.02006"}
{"aliases": [], "authors": ["Yanzhe Zhang", "Tao Yu", "Diyi Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "085c64595365cc6439c8c5fa8a479259a3d59b7e9740e0f9d43e48d00e1439c0", "submitted_at": "2024-11-04T18:56:42Z", "updated_at": "2025-05-24T16:15:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2411.02391v2"}, "publication": {"authors": ["Yanzhe Zhang", "Tao Yu", "Diyi Yang"], "date_basis": "Explicit publisher citation_publication_date: 2025/7. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 7], "doi": "10.18653/v1/2025.acl-long.411", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.acl-long.411/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "8387-8401", "pdf_url": "https://aclanthology.org/2025.acl-long.411.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "Attacking Vision-Language Computer Agents via Pop-ups", "type": "conference-paper", "url": "https://aclanthology.org/2025.acl-long.411/", "venue": "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)", "volume": null}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:56Z", "role": "primary", "sha256": "a28a603d74e08954a05265f54c8765bd874a6cb2fc0c5a68a35e15e68692b0a1", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.acl-long.411/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2411.02391", "identifiers": {"arxiv": "2411.02391"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.02391"}, {"rel": "publication", "url": "https://aclanthology.org/2025.acl-long.411/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.acl-long.411.pdf"}], "name": "Attacking Vision-Language Computer Agents via Pop-ups", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-11-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Tests whether adversarial visual pop-ups divert computer agents from intended tasks in desktop and web benchmark environments.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-05-24", "url": "https://arxiv.org/abs/2411.02391"}
{"aliases": [], "authors": ["Siyuan Hu", "Mingyu Ouyang", "Difei Gao", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "255e101e263c4e3a0cb491d4c34b3e75a8f3660cec2141e3a507a30cf5514dbd", "submitted_at": "2024-11-15T16:23:52Z", "updated_at": "2024-11-15T16:23:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2411.10323v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2411.10323", "identifiers": {"arxiv": "2411.10323"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.10323"}], "name": "The Dawn of GUI Agent: A Preliminary Case Study with Claude 3.5 Computer Use", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-11-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents a preliminary study of Claude 3.5 Computer Use across desktop tasks and software, analyzing planning, action, and critique behavior and providing a framework for deploying GUI-automation models.", "topics": ["evaluation", "automation", "planning"], "updated_at": "2024-11-15", "url": "https://arxiv.org/abs/2411.10323"}
{"aliases": [], "authors": ["Minghe Gao", "Wendong Bu", "Bingchen Miao", "Yang Wu", "Yunfei Li", "Juncheng Li", "Siliang Tang", "Qi Wu", "Yueting Zhuang", "Meng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "daa9b7092616e94a1d0124d36d8b75d0401b9647ec762c963524616a92def9a5", "submitted_at": "2024-11-17T02:44:56Z", "updated_at": "2024-11-17T02:44:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2411.10943v1"}}, "environments": ["web", "desktop", "android", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:29:13.577033+00:00", "role": "primary", "sha256": "92f18020e481d6f42a6d35e96dff820594e9370597f5392d038e9ee908c49f1e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2411.10943v1"}], "id": "arxiv:2411.10943", "identifiers": {"arxiv": "2411.10943"}, "kind": "survey", "last_verified_at": "2026-10-18T04:29:13.577033+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.10943"}], "name": "Generalist Virtual Agents: A Survey on Autonomous Agents Across Digital Platforms", "notes": "Reviewed the complete retained arXiv abstract and exact-version official HTML sections on browser/application/OS environments, screen observations, input actions and evaluation. The HTML abstract fingerprint and ordered authors match the retained Atom entry. Literature claims and experiments were not independently reproduced.", "published_at": "2024-11-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Surveys virtual agents for browsers, applications and operating systems, organizing prior work by task type, observation and action spaces, model strategies and evaluation methods. It examines screen-based interaction and keyboard, mouse and touch control, alongside transferability and long-sequence decision-making challenges.", "topics": ["survey", "web-navigation", "mobile-navigation", "evaluation", "planning"], "updated_at": "2024-11-17", "url": "https://arxiv.org/abs/2411.10943"}
{"aliases": [], "authors": ["Thibault Le Sellier De Chezelles", "Maxime Gasse", "Alexandre Drouin", "Massimo Caccia", "Léo Boisvert", "Megh Thakkar", "Tom Marty", "Rim Assouel", "Sahar Omidi Shayegan", "Lawrence Keunho Jang", "Xing Han Lù", "Ori Yoran", "Dehan Kong", "Frank F. Xu", "Siva Reddy", "Quentin Cappart", "Graham Neubig", "Ruslan Salakhutdinov", "Nicolas Chapados", "Alexandre Lacoste"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "572a993cd37a83bf7100dca71f0a44473ed6547f521409abf0f4ee22733a28cc", "submitted_at": "2024-12-06T23:43:59Z", "updated_at": "2025-02-28T16:02:27Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2412.05467v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2412.05467", "identifiers": {"arxiv": "2412.05467"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.05467"}], "name": "The BrowserGym Ecosystem for Web Agent Research", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-12-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes BrowserGym's shared web interaction environment and AgentLab tooling for consistent experiments across multiple agent benchmarks.", "topics": ["evaluation", "sandbox"], "updated_at": "2025-02-28", "url": "https://arxiv.org/abs/2412.05467"}
{"aliases": [], "authors": ["Lu Wang", "Fangkai Yang", "Chaoyun Zhang", "Junting Lu", "Jiaxu Qian", "Shilin He", "Pu Zhao", "Bo Qiao", "Ray Huang", "Si Qin", "Qisheng Su", "Jiayi Ye", "Yudi Zhang", "Jian-Guang Lou", "Qingwei Lin", "Saravan Rajmohan", "Dongmei Zhang", "Qi Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "54051d974e388ed2a26f1421b8c1fc79472bb941810519039ef48e90bf4761b6", "submitted_at": "2024-12-13T11:19:56Z", "updated_at": "2025-01-13T06:47:21Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2412.10047v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:41Z", "role": "primary", "sha256": "0ba893415123ecc1763614ef5fc97e5c8bb7835f10b26b43f1b921ea860484cf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.06212%2C2410.06215%2C2410.07283%2C2410.07331%2C2410.07869%2C2410.08209%2C2410.08368%2C2410.10570%2C2410.12481%2C2410.12669%2C2410.15549%2C2410.16950%2C2410.21909%2C2410.22584%2C2410.22770%2C2410.23022%2C2410.23555%2C2411.02006%2C2411.02393%2C2411.02479%2C2411.03312%2C2411.03817%2C2411.05451%2C2411.11145%2C2411.11904%2C2411.11916%2C2411.12990%2C2411.14228%2C2411.15397%2C2412.01857%2C2412.04144%2C2412.04383%2C2412.05819%2C2412.06224%2C2412.06771%2C2412.09530%2C2412.09613%2C2412.10047%2C2412.10380%2C2412.10673%2C2412.12004%2C2412.13178%2C2412.13810%2C2412.14203%2C2412.14222%2C2412.16359%2C2412.19638%2C2412.20206%2C2501.03895%2C2501.05179&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:13:28Z", "role": "primary", "sha256": "aaddc3cae2fcadc9780390fceb663c0200ecb042639e2c69bccea19607cd78ea", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2412.10047v2"}], "id": "arxiv:2412.10047", "identifiers": {"arxiv": "2412.10047"}, "kind": "paper", "last_verified_at": "2026-10-17T09:13:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.10047"}], "name": "Large Action Models: From Inception to Implementation", "notes": "Reviewed against the exact-version official arXiv abstract page; its normalized abstract SHA-256 matches the retained bibliographic snapshot. The original API response body was not locally available for this review. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-12-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses a Windows-agent case study to describe a development workflow for large action models, covering data collection, training, grounding, environment integration, and evaluation.", "topics": ["automation", "data-generation", "evaluation"], "updated_at": "2025-01-13", "url": "https://arxiv.org/abs/2412.10047"}
{"aliases": [], "authors": ["Frank F. Xu", "Yufan Song", "Boxuan Li", "Yuxuan Tang", "Kritanjali Jain", "Mengxue Bao", "Zora Z. Wang", "Xuhui Zhou", "Zhitong Guo", "Murong Cao", "Mingyang Yang", "Hao Yang Lu", "Amaad Martin", "Zhe Su", "Leander Maben", "Raj Mehta", "Wayne Chi", "Lawrence Jang", "Yiqing Xie", "Shuyan Zhou", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "77f91dd9ec9f50fb5526a78cfe270a7a1705d5bd83a2aef26644453740c005d5", "submitted_at": "2024-12-18T18:55:40Z", "updated_at": "2025-09-10T08:35:19Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2412.14161v3"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2412.14161", "identifiers": {"arxiv": "2412.14161"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.14161"}], "name": "TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-12-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Simulates a software-company workplace to evaluate agents on professional tasks involving websites, programs, files, and coworker communication.", "topics": ["evaluation", "long-horizon", "workflow"], "updated_at": "2025-09-10", "url": "https://arxiv.org/abs/2412.14161"}
{"aliases": [], "authors": ["Kangjia Zhao", "Jiahui Song", "Leigang Sha", "Haozhan Shen", "Zhi Chen", "Tiancheng Zhao", "Xiubo Liang", "Jianwei Yin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9e509f015068bcf1294a97236ff3bbde5e3ceb167f3cb032fcfd84fc4a5b01d0", "submitted_at": "2024-12-24T13:41:47Z", "updated_at": "2024-12-24T13:41:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2412.18426v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2412.18426", "identifiers": {"arxiv": "2412.18426"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.18426"}], "name": "GUI Testing Arena: A Unified Benchmark for Advancing Autonomous GUI Testing Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-12-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates GUI testing through test-intent generation, execution, and defect detection using real, modified, and synthetic mobile application data.", "topics": ["evaluation", "automation", "mobile-navigation"], "updated_at": "2024-12-24", "url": "https://arxiv.org/abs/2412.18426"}
{"aliases": [], "authors": ["Yuxiang Chai", "Shunye Tang", "Han Xiao", "Weifeng Lin", "Hanhao Li", "Jiayu Zhang", "Liang Liu", "Pengxiang Zhao", "Guangyi Liu", "Guozhi Wang", "Shuai Ren", "Rongduo Han", "Haining Zhang", "Siyuan Huang", "Hongsheng Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8ac7100a0f3cc3f004672eb6923b1f0ecb43c3d98da6959f0ff7514d0fb2f442", "submitted_at": "2025-01-02T09:03:56Z", "updated_at": "2026-01-12T07:01:09Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2501.01149v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2501.01149", "identifiers": {"arxiv": "2501.01149"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2501.01149"}], "name": "A3: Android Agent Arena for Mobile GUI Agents with Essential-State Procedural Evaluation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-01-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Android Agent Arena with online application tasks and procedural evaluation based on reaching essential intermediate states.", "topics": ["evaluation", "mobile-navigation"], "updated_at": "2026-01-12", "url": "https://arxiv.org/abs/2501.01149"}
{"aliases": [], "authors": ["Jialong Wu", "Wenbiao Yin", "Yong Jiang", "Zhenglin Wang", "Zekun Xi", "Runnan Fang", "Linhai Zhang", "Yulan He", "Deyu Zhou", "Pengjun Xie", "Fei Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "32cf550b1e1d5b671b3a4c834787cb67c32c01f508494c5610dac91e95bd92ba", "submitted_at": "2025-01-13T18:58:07Z", "updated_at": "2025-08-10T05:59:20Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2501.07572v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2501.07572", "identifiers": {"arxiv": "2501.07572"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2501.07572"}], "name": "WebWalker: Benchmarking LLMs in Web Traversal", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-01-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebWalkerQA and an exploration-and-critique framework for gathering information by traversing multiple levels of a website.", "topics": ["evaluation", "web-navigation"], "updated_at": "2025-08-10", "url": "https://arxiv.org/abs/2501.07572"}
{"aliases": [], "authors": ["Faria Huq", "Zora Zhiruo Wang", "Frank F. Xu", "Tianyue Ou", "Shuyan Zhou", "Jeffrey P. Bigham", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b14b88273749ac54ebbe3694cbe4196640da8f81ba2faf77b0e6bbd79a060f6", "submitted_at": "2025-01-28T00:56:53Z", "updated_at": "2026-02-27T05:35:06Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2501.16609v4"}, "publication": {"authors": ["Faria Huq", "Zora Zhiruo Wang", "Frank F. Xu", "Tianyue Ou", "Shuyan Zhou", "Jeffrey P. Bigham", "Graham Neubig"], "date_basis": "The ACL Anthology citation_publication_date is 2025/4, and its visible Month/Year fields state April 2025. Only month precision is supplied; no day is inferred from conference dates.", "date_parts": [2025, 4], "doi": "10.18653/v1/2025.naacl-demo.17", "event_dates": null, "evidence_urls": ["https://arxiv.org/abs/2501.16609v4", "https://aclanthology.org/2025.naacl-demo.17/"], "identity_basis": "The retained arXiv v4 abstract page explicitly links this DOI. The ACL publication has the same complete title and the same seven authors in the same order, and describes the same CowPilot human-agent navigation framework. This attaches the proceedings edition to the existing arXiv identity.", "issue": null, "notes": "The April 2025 proceedings edition is distinct from the retained arXiv v4, last revised on 2026-02-27. The publisher abstract has minor wording and typesetting differences from the retained arXiv abstract; neither its text nor its date replaces the original arXiv snapshot or annotation. Exact publication day, numeric volume, issue and series are not established.", "pages": "163-172", "pdf_url": "https://aclanthology.org/2025.naacl-demo.17.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "CowPilot: A Framework for Autonomous and Human-Agent Collaborative Web Navigation", "type": "conference-paper", "url": "https://aclanthology.org/2025.naacl-demo.17/", "venue": "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (System Demonstrations)", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:44Z", "role": "primary", "sha256": "9ca057ee6538002c739e58a7b3f86fc51a434d085a33e5590f0720e4f27c6b24", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2501.06497%2C2501.06994%2C2501.07834%2C2501.07927%2C2501.11233%2C2501.15570%2C2501.16403%2C2501.16609%2C2501.17391%2C2502.00372%2C2502.00392%2C2502.00964%2C2502.00988%2C2502.01562%2C2502.01839%2C2502.03723%2C2502.05174%2C2502.05227%2C2502.08489%2C2502.11221%2C2502.12130%2C2502.14496%2C2502.16587%2C2502.16786%2C2502.16796%2C2502.18356%2C2502.18512%2C2502.20383%2C2502.20390%2C2502.20582%2C2503.06378%2C2503.08026%2C2503.10613%2C2503.10809%2C2503.11579%2C2503.12228%2C2503.12461%2C2503.13843%2C2503.14495%2C2503.15426%2C2503.16036%2C2503.16248%2C2503.16975%2C2503.17671%2C2503.18102%2C2503.21889%2C2503.21949%2C2503.23803%2C2504.01328%2C2504.01377&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:19:57Z", "role": "primary", "sha256": "8c39a7773a9f2b58950aa22d6924b00ed32ab5b2b42302590dcec88da255a71f", "supports": ["details.publication.doi", "details.publication.identity_basis", "details.publication.notes"], "url": "https://arxiv.org/abs/2501.16609v4"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:44:12Z", "role": "primary", "sha256": "410da7f4b7a79e56efc5f88014a078da0feb8d2ae3ef3d36a04dce5be29d4250", "supports": ["details.publication.type", "details.publication.title", "details.publication.authors", "details.publication.date_parts", "details.publication.date_basis", "details.publication.url", "details.publication.pdf_url", "details.publication.doi", "details.publication.venue", "details.publication.publisher", "details.publication.pages", "details.publication.identity_basis", "details.publication.notes"], "url": "https://aclanthology.org/2025.naacl-demo.17/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:19:57Z", "role": "primary", "sha256": "8c39a7773a9f2b58950aa22d6924b00ed32ab5b2b42302590dcec88da255a71f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2501.16609v4"}], "id": "arxiv:2501.16609", "identifiers": {"arxiv": "2501.16609"}, "kind": "paper", "last_verified_at": "2026-10-17T11:19:57Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2501.16609"}, {"rel": "publication", "url": "https://aclanthology.org/2025.naacl-demo.17/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.naacl-demo.17.pdf"}], "name": "CowPilot: A Framework for Autonomous and Human-Agent Collaborative Web Navigation", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CowPilot, a web-navigation framework where agents suggest actions and users can pause, reject, override, or resume them. It evaluates autonomous and collaborative use through task success and human effort.", "topics": ["web-navigation", "human-in-the-loop", "evaluation"], "updated_at": "2026-02-27", "url": "https://arxiv.org/abs/2501.16609"}
{"aliases": [], "authors": ["Henry Hengyuan Zhao", "Kaiming Yang", "Wendi Yu", "Difei Gao", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "68e6ecfcdff0163debe96f6440fb33529356bf9fd2c8743b5953ae52ae71c9ef", "submitted_at": "2025-02-12T01:06:10Z", "updated_at": "2026-05-23T20:19:43Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2502.08047v5"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2502.08047", "identifiers": {"arxiv": "2502.08047"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.08047"}], "name": "WorldGUI: An Interactive Benchmark for Desktop GUI Automation from Any Starting Point", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces desktop and web tasks with varied initial states to test planning recovery and adaptability beyond canonical starting configurations.", "topics": ["evaluation", "robustness", "planning"], "updated_at": "2026-05-23", "url": "https://arxiv.org/abs/2502.08047"}
{"aliases": [], "authors": ["Yurun Chen", "Xavier Hu", "Keting Yin", "Juncheng Li", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5af8d3a8b4db150bab0630f6040a0f3bced612f68b11d31985053f1f25c0f6f6", "submitted_at": "2025-02-18T17:01:28Z", "updated_at": "2025-08-06T07:04:43Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2502.13053v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2502.13053", "identifiers": {"arxiv": "2502.13053"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.13053"}], "name": "Evaluating the Robustness of Multimodal Agents Against Active Environmental Injection Attacks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Examines active environmental injection against mobile agents, testing whether adversarial interface content redirects decisions during task execution.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": "2025-08-06", "url": "https://arxiv.org/abs/2502.13053"}
{"aliases": [], "authors": ["Lars Krupp", "Daniel Geißler", "Paul Lukowicz", "Jakob Karolus"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3903e97e6db6bd470699f0d7b2785e53e15291234f1e5f717bae8d2ccd851ca4", "submitted_at": "2025-02-25T06:58:40Z", "updated_at": "2025-02-25T06:58:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2502.17903v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2502.17903", "identifiers": {"arxiv": "2502.17903"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.17903"}], "name": "Towards Sustainable Web Agents: A Plea for Transparency and Dedicated Metrics for Energy Consumption", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Examines energy and carbon-accounting limitations in web-agent evaluation and argues for transparent, dedicated sustainability measurements.", "topics": ["evaluation", "efficiency"], "updated_at": "2025-02-25", "url": "https://arxiv.org/abs/2502.17903"}
{"aliases": [], "authors": ["Pranjal Aggarwal", "Sean Welleck"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9e4091fc740249b6406dd9e96f2692d7fac9323759ec6fad6ba5a05f3e445844", "submitted_at": "2025-02-24T18:41:33Z", "updated_at": "2025-10-03T02:09:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2502.18525v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2502.18525", "identifiers": {"arxiv": "2502.18525"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.18525"}], "name": "Programming with Pixels: Can Computer-Use Agents do Software Engineering?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a visual IDE environment and benchmark for software-engineering tasks, comparing screenshot-based control with access to editing and shell tools.", "topics": ["evaluation", "tool-integration", "automation"], "updated_at": "2025-10-03", "url": "https://arxiv.org/abs/2502.18525"}
{"aliases": [], "authors": ["Jeffrey Yang Fan Chiang", "Seungjae Lee", "Jia-Bin Huang", "Furong Huang", "Yizheng Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "27eacc2668411e2b7f79423940aa12293fa57aa81390c64587d3074c141dcb4f", "submitted_at": "2025-02-27T18:56:26Z", "updated_at": "2025-09-21T22:29:53Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2502.20383v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:44Z", "role": "primary", "sha256": "9ca057ee6538002c739e58a7b3f86fc51a434d085a33e5590f0720e4f27c6b24", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2501.06497%2C2501.06994%2C2501.07834%2C2501.07927%2C2501.11233%2C2501.15570%2C2501.16403%2C2501.16609%2C2501.17391%2C2502.00372%2C2502.00392%2C2502.00964%2C2502.00988%2C2502.01562%2C2502.01839%2C2502.03723%2C2502.05174%2C2502.05227%2C2502.08489%2C2502.11221%2C2502.12130%2C2502.14496%2C2502.16587%2C2502.16786%2C2502.16796%2C2502.18356%2C2502.18512%2C2502.20383%2C2502.20390%2C2502.20582%2C2503.06378%2C2503.08026%2C2503.10613%2C2503.10809%2C2503.11579%2C2503.12228%2C2503.12461%2C2503.13843%2C2503.14495%2C2503.15426%2C2503.16036%2C2503.16248%2C2503.16975%2C2503.17671%2C2503.18102%2C2503.21889%2C2503.21949%2C2503.23803%2C2504.01328%2C2504.01377&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:04Z", "role": "primary", "sha256": "0118540c95d316ed30aabf783c072d413677aaf89539316e08e1d81ddab2c5e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2502.20383v3"}], "id": "arxiv:2502.20383", "identifiers": {"arxiv": "2502.20383"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.20383"}], "name": "Why Are Web AI Agents More Vulnerable Than Standalone LLMs? A Security Analysis", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-02-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Analyzes web-agent vulnerabilities at the component level, examining goal placement in prompts, multistep action generation, and observations as sources of increased susceptibility.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2025-09-21", "url": "https://arxiv.org/abs/2502.20383"}
{"aliases": [], "authors": ["Sami Marreed", "Alon Oved", "Avi Yaeli", "Segev Shlomov", "Ido Levy", "Offer Akrabi", "Aviad Sela", "Asaf Adi", "Nir Mashkif"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2248d423c954ee6e406e3f46f6da43fabd6e832ffacc05cfa32083f2ff0c2700", "submitted_at": "2025-02-24T09:31:56Z", "updated_at": "2025-07-09T13:52:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2503.01861v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.01861", "identifiers": {"arxiv": "2503.01861"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.01861"}], "name": "Towards Enterprise-Ready Computer Using Generalist Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes the CUGA enterprise-agent development process, emphasizing iterative evaluation, failure analysis, and refinement on software-task benchmarks.", "topics": ["automation", "evaluation", "workflow"], "updated_at": "2025-07-09", "url": "https://arxiv.org/abs/2503.01861"}
{"aliases": [], "authors": ["Jiahui Sun", "Zhichao Hua", "Yubin Xia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a0c393d109ab0f8da2bf25bb39581236babd50f7554f2ec99679a99fb7215171", "submitted_at": "2025-03-04T08:44:30Z", "updated_at": "2025-09-24T05:11:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2503.02403v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}], "id": "arxiv:2503.02403", "identifiers": {"arxiv": "2503.02403"}, "kind": "paper", "last_verified_at": "2026-10-16T06:45:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.02403"}], "name": "AutoEval: A Practical Framework for Autonomous Evaluation of Mobile Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Generates reward signals from UI-state changes and uses a judging system to evaluate mobile agents with reduced manual evaluator construction.", "topics": ["evaluation", "mobile-navigation"], "updated_at": "2025-09-24", "url": "https://arxiv.org/abs/2503.02403"}
{"aliases": [], "authors": ["Zheng Hui", "Yinheng Li", "Dan zhao", "Tianyi Chen", "Colby Banbury", "Kazuhito Koishida"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cc499e361f0bc46c3c5ee5a8abebd911396e383a004655de56db1ec7cec8e231", "submitted_at": "2025-01-27T08:29:17Z", "updated_at": "2025-01-27T08:29:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2503.04730v1"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.04730", "identifiers": {"arxiv": "2503.04730"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.04730"}], "name": "WinClick: GUI Grounding with Multimodal Large Language Models", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-01-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a screenshot-based Windows GUI agent and the WinSpot benchmark for evaluating localization of actionable interface regions.", "topics": ["grounding", "evaluation"], "updated_at": "2025-01-27", "url": "https://arxiv.org/abs/2503.04730"}
{"aliases": [], "authors": ["Ada Defne Tur", "Nicholas Meade", "Xing Han Lù", "Alejandra Zambrano", "Arkil Patel", "Esin Durmus", "Spandana Gella", "Karolina Stańczak", "Siva Reddy"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e4588022dd47624058db8a16d7f7763c591241298e94594b56bfeb4ab2e01737", "submitted_at": "2025-03-06T20:43:14Z", "updated_at": "2025-03-06T20:43:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2503.04957v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.04957", "identifiers": {"arxiv": "2503.04957"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.04957"}], "name": "SafeArena: Evaluating the Safety of Autonomous Web Agents", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2025-03-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Pairs benign and harmful web tasks to assess deliberate misuse risks and categorize agent behavior by risk level.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-03-06", "url": "https://arxiv.org/abs/2503.04957"}
{"aliases": [], "authors": ["Wenhao Wang", "Zijie Yu", "Rui Ye", "Jianqing Zhang", "Siheng Chen", "Yanfeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b408c850ae83e603a1b3d8303164b9f2d7be51580be99027275b0734ea8f56f2", "submitted_at": "2025-03-07T04:52:20Z", "updated_at": "2025-03-07T04:52:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2503.05143v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2503.05143", "identifiers": {"arxiv": "2503.05143"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.05143"}], "name": "FedMABench: Benchmarking Mobile Agents on Decentralized Heterogeneous User Data", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a benchmark for federated mobile-agent training under heterogeneous user and application distributions, combining datasets, algorithms, and base models.", "topics": ["evaluation", "mobile-navigation"], "updated_at": "2025-03-07", "url": "https://arxiv.org/abs/2503.05143"}
{"aliases": [], "authors": ["Yixiao Song", "Katherine Thai", "Chau Minh Pham", "Yapei Chang", "Mazin Nadaf", "Mohit Iyyer"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e1cd6327559d11982fb4a97e739cb6b5234c509859692084ed78130241f6d252", "submitted_at": "2025-03-10T23:50:30Z", "updated_at": "2025-07-24T17:45:05Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2503.07919v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.07919", "identifiers": {"arxiv": "2503.07919"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.07919"}], "name": "BEARCUBS: A benchmark for computer-using web agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "BEARCUBS provides 111 live-web information-seeking questions with short answers and human-validated browsing trajectories. Its tasks require multimodal computer interaction, including video understanding and navigation of interactive 3D content.", "topics": ["evaluation", "web-navigation", "video"], "updated_at": "2025-07-24", "url": "https://arxiv.org/abs/2503.07919"}
{"aliases": [], "authors": ["Pei Yang", "Hai Ci", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7d4423ee1e7047a6228f5b29684618fc4108af0694484a7ceddf6c3b53fbb06c", "submitted_at": "2025-03-12T10:38:15Z", "updated_at": "2025-03-12T10:38:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2503.09241v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2503.09241", "identifiers": {"arxiv": "2503.09241"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.09241"}], "name": "In-Context Defense in Computer Agents: An Empirical Study", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-03-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies in-context defenses for computer agents exposed to misleading pop-ups and other environmental instructions. It examines defensive reasoning before action planning and how a small set of examples changes resistance to these attacks.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-03-12", "url": "https://arxiv.org/abs/2503.09241"}
{"aliases": [], "authors": ["Arman Zharmagambetov", "Chuan Guo", "Ivan Evtimov", "Maya Pavlova", "Ruslan Salakhutdinov", "Kamalika Chaudhuri"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bbb0726446c9679781b8e5930da0d3f406989a4b47759ce3440c9ad56d5839d8", "submitted_at": "2025-03-12T19:30:31Z", "updated_at": "2025-10-01T23:55:58Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2503.09780v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.09780", "identifiers": {"arxiv": "2503.09780"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.09780"}], "name": "AgentDAM: Privacy Leakage Evaluation for Autonomous Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates whether autonomous web agents minimize use of sensitive personal information when completing tasks that expose unnecessary private data.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-10-01", "url": "https://arxiv.org/abs/2503.09780"}
{"aliases": [], "authors": ["Fanbin Lu", "Zhisheng Zhong", "Ziqin Wei", "Shu Liu", "Chi-Wing Fu", "Jiaya Jia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e883289bafee01759a0f4f3f7051cbbebdcd8059e7520d68c25ba313b8da69a6", "submitted_at": "2025-03-16T14:53:43Z", "updated_at": "2025-03-24T16:33:28Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2503.12532v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.12532", "identifiers": {"arxiv": "2503.12532"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.12532"}], "name": "STEVE: A Step Verification Pipeline for Computer-use Agent Training", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Labels individual computer-use actions by comparing surrounding screenshots, then trains agents from both successful and unsuccessful steps.", "topics": ["data-generation", "evaluation", "reinforcement-learning"], "updated_at": "2025-03-24", "url": "https://arxiv.org/abs/2503.12532"}
{"aliases": [], "authors": ["Shravan Nayak", "Xiangru Jian", "Kevin Qinghong Lin", "Juan A. Rodriguez", "Montek Kalsi", "Rabiul Awal", "Nicolas Chapados", "M. Tamer Özsu", "Aishwarya Agrawal", "David Vazquez", "Christopher Pal", "Perouz Taslakian", "Spandana Gella", "Sai Rajeswar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5abf3c7b08a4331d4f634d8cc63a7f8ff40713f7344231887dac1b5e061a8eae", "submitted_at": "2025-03-19T19:26:17Z", "updated_at": "2025-05-06T17:43:11Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2503.15661v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2503.15661", "identifiers": {"arxiv": "2503.15661"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.15661"}], "name": "UI-Vision: A Desktop-centric GUI Benchmark for Visual Perception and Interaction", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2025-03-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Provides annotated desktop demonstrations for evaluating element grounding, layout grounding, and action prediction.", "topics": ["grounding", "evaluation", "screen-understanding"], "updated_at": "2025-05-06", "url": "https://arxiv.org/abs/2503.15661"}
{"aliases": [], "authors": ["Li Zhang", "Longxi Gao", "Mengwei Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e450216f8f802d8f3d16c5886097cef801d051951af0c57fdf195cc176ac744f", "submitted_at": "2025-03-21T01:52:43Z", "updated_at": "2025-03-21T01:52:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2503.16788v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.16788", "identifiers": {"arxiv": "2503.16788"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.16788"}], "name": "Does Chain-of-Thought Reasoning Help Mobile GUI Agent? An Empirical Study", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-03-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study compares matched reasoning and non-reasoning vision-language models on static mobile GUI tasks and interactive AndroidWorld navigation. It finds that chain-of-thought effects vary across tasks and model families.", "topics": ["evaluation", "mobile-navigation", "planning"], "updated_at": "2025-03-21", "url": "https://arxiv.org/abs/2503.16788"}
{"aliases": [], "authors": ["Noam Kahlon", "Guy Rom", "Anatoly Efros", "Filippo Galgani", "Omri Berkovitch", "Sapir Caduri", "William E. Bishop", "Oriana Riva", "Ido Dagan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8a49d662eddf84690395b025e50f3378c3be2a3fc8abe02bd7202c6ff879bbe3", "submitted_at": "2025-03-25T10:46:08Z", "updated_at": "2025-03-25T10:46:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2503.19537v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2503.19537", "identifiers": {"arxiv": "2503.19537"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.19537"}], "name": "Agent-Initiated Interaction in Phone UI Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Defines when phone agents should request user interaction and introduces AndroidInteraction annotations for deciding whether and how to engage the user.", "topics": ["human-in-the-loop", "mobile-navigation", "evaluation"], "updated_at": "2025-03-25", "url": "https://arxiv.org/abs/2503.19537"}
{"aliases": [], "authors": ["Sejin Lee", "Jian Kim", "Haon Park", "Ashkan Yousefpour", "Sangyoon Yu", "Min Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5e90cbc547703a2673cfb30bbe9079496ce4c076b51f39e0557f5db9048947c8", "submitted_at": "2025-03-26T07:08:15Z", "updated_at": "2025-06-08T05:39:56Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2503.20279v3"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2503.20279", "identifiers": {"arxiv": "2503.20279"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.20279"}], "name": "sudo rm -rf agentic_security", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-03-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates security failures of computer-use agents under adversarial requests and screen-context manipulation. The study examines refusal behavior while agents perform tasks in desktop and web environments.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-06-08", "url": "https://arxiv.org/abs/2503.20279"}
{"aliases": [], "authors": ["Ziming Cheng", "Zhiyuan Huang", "Junting Pan", "Zhaohui Hou", "Mingjie Zhan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1bb09b54379d2fbc5b489f24a27f6311b91bb0821eefe5187aa160ffb00639d9", "submitted_at": "2025-03-31T14:56:24Z", "updated_at": "2025-08-24T08:39:18Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2503.24180v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}], "id": "arxiv:2503.24180", "identifiers": {"arxiv": "2503.24180"}, "kind": "paper", "last_verified_at": "2026-10-16T06:45:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.24180"}], "name": "Navi-plus: Managing Ambiguous GUI Navigation Tasks with Follow-up Questions", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Navi-plus evaluates GUI agents that ask follow-up questions to complete ambiguous user instructions. It provides GUI clarification question-answer pairs and Dual-Stream Trajectory Evaluation for this interactive setting.", "topics": ["human-in-the-loop", "evaluation", "planning"], "updated_at": "2025-08-24", "url": "https://arxiv.org/abs/2503.24180"}
{"aliases": [], "authors": ["Tianci Xue", "Weijian Qi", "Tianneng Shi", "Chan Hee Song", "Boyu Gou", "Dawn Song", "Huan Sun", "Yu Su"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ccad667f7c5bb75e58d75d5eab5cc24c7c3610be9eed1a09079de5e23c85af04", "submitted_at": "2025-04-02T05:51:29Z", "updated_at": "2025-10-08T02:22:45Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2504.01382v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.01382", "identifiers": {"arxiv": "2504.01382"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.01382"}], "name": "An Illusion of Progress? Assessing the Current State of Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Online-Mind2Web, with 300 tasks across 136 live websites, and examines how benchmark design affects assessments of web agents. It also develops an LLM judge for task completion and compares its judgments with human evaluation.", "topics": ["evaluation", "web-navigation"], "updated_at": "2025-10-08", "url": "https://arxiv.org/abs/2504.01382"}
{"aliases": [], "authors": ["Antoine Chevrot", "Alexandre Vernotte", "Jean-Rémy Falleri", "Xavier Blanc", "Bruno Legeard"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "76d0ffb55478570499c8cddf92a27d346cf11749f8ae5f8e128c87653b2bdbdc", "submitted_at": "2025-04-02T08:48:01Z", "updated_at": "2025-04-02T08:48:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.01495v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.01495", "identifiers": {"arxiv": "2504.01495"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.01495"}], "name": "Are Autonomous Web Agents Good Testers?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Adapts web agents to execute natural-language test steps, verify assertions, and issue test verdicts. The study introduces 113 passing and failing test cases across three offline web applications and compares two agent implementations with qualitative failure analysis.", "topics": ["automation", "evaluation", "web-navigation"], "updated_at": "2025-04-02", "url": "https://arxiv.org/abs/2504.01495"}
{"aliases": [], "authors": ["Xiaolei Li", "Jialun Cao", "Yepang Liu", "Shing-Chi Cheung", "Hailong Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a646bb2136cdaf47a0d8b3f485d3e1973fc15ac4f844e1beddedf6021b56ffe4", "submitted_at": "2025-04-03T07:45:09Z", "updated_at": "2025-04-03T07:45:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.02357v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:12:09Z", "role": "primary", "sha256": "054a5a305d86c7fb9107d467d194fc4a64aa4f5af23d214bea7f52c2f2c9fd35", "supports": ["environments"], "url": "https://arxiv.org/html/2504.02357"}], "id": "arxiv:2504.02357", "identifiers": {"arxiv": "2504.02357"}, "kind": "paper", "last_verified_at": "2026-10-16T06:45:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.02357"}], "name": "ReuseDroid: A VLM-empowered Android UI Test Migrator Boosted by Active Feedback", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced. The platform or GUI evaluation setting was additionally checked in the primary full text.", "published_at": "2025-04-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "ReuseDroid migrates Android GUI tests between functionally similar apps using VLM agents that separate shared test intent from app-specific interaction flows. It introduces LinPro, a dataset of 578 migration tasks across 39 apps, to evaluate test transfer.", "topics": ["automation", "evaluation", "cross-app"], "updated_at": "2025-04-03", "url": "https://arxiv.org/abs/2504.02357"}
{"aliases": [], "authors": ["Haoren Zhao", "Tianyi Chen", "Zhen Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ededfa864150f1b7b4ba53b265b5cbaa60bcf42690a8e22f5c13f915b6cd0108", "submitted_at": "2025-04-07T03:58:45Z", "updated_at": "2025-04-07T03:58:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.04716v1"}}, "environments": ["web", "desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.04716", "identifiers": {"arxiv": "2504.04716"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.04716"}], "name": "On the Robustness of GUI Grounding Models Against Image Attacks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates GUI grounding models under natural image noise, untargeted adversarial perturbations, and targeted image attacks. Experiments cover mobile, desktop, and web interfaces and analyze sensitivity to image resolution and adversarial changes.", "topics": ["grounding", "robustness", "safety-security", "evaluation"], "updated_at": "2025-04-07", "url": "https://arxiv.org/abs/2504.04716"}
{"aliases": [], "authors": ["Kaixin Li", "Ziyang Meng", "Hongzhan Lin", "Ziyang Luo", "Yuchen Tian", "Jing Ma", "Zhiyong Huang", "Tat-Seng Chua"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "679bfcac736544e9a241c579adfda0cacbc40838452ea56ad631f8086868415a", "submitted_at": "2025-04-04T14:25:17Z", "updated_at": "2025-04-04T14:25:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.07981v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.07981", "identifiers": {"arxiv": "2504.07981"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.07981"}], "name": "ScreenSpot-Pro: GUI Grounding for Professional High-Resolution Computer Use", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2025-04-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates GUI target localization in high-resolution professional software and studies cascaded visual search to narrow the target region.", "topics": ["grounding", "evaluation"], "updated_at": "2025-04-04", "url": "https://arxiv.org/abs/2504.07981"}
{"aliases": [], "authors": ["Xing Han Lù", "Amirhossein Kazemnejad", "Nicholas Meade", "Arkil Patel", "Dongchan Shin", "Alejandra Zambrano", "Karolina Stańczak", "Peter Shaw", "Christopher J. Pal", "Siva Reddy"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "53cba694577319ac1992c853be79735d654d9d32c12edcc543e6738ce010050d", "submitted_at": "2025-04-11T19:49:22Z", "updated_at": "2025-10-06T15:46:30Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.08942v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.08942", "identifiers": {"arxiv": "2504.08942"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.08942"}], "name": "AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "AgentRewardBench evaluates automated web-agent judges against expert reviews of 1,302 trajectories from five benchmarks. The annotations cover completion, side effects, and repetition, enabling comparison of LLM judges and rule-based task evaluators.", "topics": ["evaluation", "web-navigation"], "updated_at": "2025-10-06", "url": "https://arxiv.org/abs/2504.08942"}
{"aliases": [], "authors": ["Junlei Zhang", "Zichen Ding", "Chang Ma", "Zijie Chen", "Qiushi Sun", "Zhenzhong Lan", "Junxian He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "171517999164a3581f71cf363ec6eed357b28ca86908cb97e8498c6958762a5f", "submitted_at": "2025-04-14T11:35:02Z", "updated_at": "2025-04-15T17:13:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.10127v2"}}, "environments": ["web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.10127", "identifiers": {"arxiv": "2504.10127"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.10127"}], "name": "Breaking the Data Barrier -- Building GUI Agents Through Task Generalization", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies whether mid-training on GUI perception, multimodal reasoning, and text reasoning transfers to GUI-agent planning. Experiments compare 11 task sources and curated mixtures using WebArena and AndroidWorld to investigate training-data choices when GUI trajectories are scarce.", "topics": ["data-generation", "planning", "evaluation"], "updated_at": "2025-04-15", "url": "https://arxiv.org/abs/2504.10127"}
{"aliases": [], "authors": ["Suyu Ye", "Haojun Shi", "Darren Shih", "Hyokun Yun", "Tanya Roosta", "Tianmin Shu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "268adf347e1f224f434e6dba838873c99a4050686cae1495643d9638d840d809", "submitted_at": "2025-04-14T17:36:46Z", "updated_at": "2025-12-01T15:22:40Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.10445v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2504.10445", "identifiers": {"arxiv": "2504.10445"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.10445"}], "name": "RealWebAssist: A Benchmark for Long-Horizon Web Assistance with Real-World Users", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "RealWebAssist evaluates web assistance through sequences of instructions collected from real users across multiple websites. It tests intent interpretation, user-specific routines, long-term context, and visual grounding when instructions are ambiguous or evolve during interaction.", "topics": ["evaluation", "long-horizon", "human-in-the-loop", "grounding"], "updated_at": "2025-12-01", "url": "https://arxiv.org/abs/2504.10445"}
{"aliases": [], "authors": ["Xinyi Liu", "Xiaoyi Zhang", "Ziyun Zhang", "Yan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "363e3c3e882fa93a19df74bbf77aa98525cf4639c8f62100c3ddc3e0a438997f", "submitted_at": "2025-04-15T14:56:21Z", "updated_at": "2025-07-30T13:57:49Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2504.11257v4"}, "publication": {"authors": ["Xinyi Liu", "Xiaoyi Zhang", "Ziyun Zhang", "Yan Lu"], "date_basis": "Explicit publisher citation_publication_date: 2025/7. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 7], "doi": "10.18653/v1/2025.findings-acl.809", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.findings-acl.809/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "15668-15684", "pdf_url": "https://aclanthology.org/2025.findings-acl.809.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "UI-E2I-Synth: Advancing GUI Grounding with Large-Scale Instruction Synthesis", "type": "conference-paper", "url": "https://aclanthology.org/2025.findings-acl.809/", "venue": "Findings of the Association for Computational Linguistics: ACL 2025", "volume": null}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:59Z", "role": "primary", "sha256": "f7bcea2d3656ebc3ac305d65271000e411b735eb10556d23ee0d6157593a109f", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.findings-acl.809/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.11257", "identifiers": {"arxiv": "2504.11257"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.11257"}, {"rel": "publication", "url": "https://aclanthology.org/2025.findings-acl.809/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.findings-acl.809.pdf"}], "name": "UI-E2I-Synth: Advancing GUI Grounding with Large-Scale Instruction Synthesis", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-E2I-Synth uses GPT-4o to synthesize GUI grounding instructions with varied complexity, addressing small elements, imbalanced element types, and implicit requests. The accompanying UI-I2E-Bench evaluates instruction-to-element localization across these annotation challenges.", "topics": ["data-generation", "grounding", "evaluation"], "updated_at": "2025-07-30", "url": "https://arxiv.org/abs/2504.11257"}
{"aliases": [], "authors": ["Chaoran Chen", "Zhiping Zhang", "Bingcan Guo", "Shang Ma", "Ibrahim Khalilov", "Simret A Gebreegziabher", "Yanfang Ye", "Ziang Xiao", "Yaxing Yao", "Tianshi Li", "Toby Jia-Jun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "48acf3279b854d45ec076e630e5a84601879fc9ff894c5d368adb59d03bd3de6", "submitted_at": "2025-04-15T15:21:09Z", "updated_at": "2025-04-15T15:21:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.11281v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.11281", "identifiers": {"arxiv": "2504.11281"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.11281"}], "name": "The Obvious Invisible Threat: LLM-Powered GUI Agents' Vulnerability to Fine-Print Injections", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Characterizes six forms of GUI-content injection and studies their effects on GUI agents and human users using adversarial webpages. The analysis examines privacy failures and the limits of human oversight.", "topics": ["safety-security", "human-in-the-loop", "evaluation"], "updated_at": "2025-04-15", "url": "https://arxiv.org/abs/2504.11281"}
{"aliases": [], "authors": ["Biniam Fisseha Demissie", "Yan Naing Tun", "Lwin Khin Shar", "Mariano Ceccato"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1a25cb3ca680732a08cf2c461a65caa0c6a70404224bdfd919ba5bec7760ba72", "submitted_at": "2025-04-16T00:19:31Z", "updated_at": "2025-04-16T00:19:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.11675v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}], "id": "arxiv:2504.11675", "identifiers": {"arxiv": "2504.11675"}, "kind": "paper", "last_verified_at": "2026-10-16T06:45:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.11675"}], "name": "VLM-Fuzz: Vision Language Model Assisted Recursive Depth-first Search Exploration for Effective UI Testing of Android Apps", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "VLM-Fuzz combines depth-first UI exploration with Android manifest analysis, runtime widget hierarchies, and on-demand visual reasoning to generate app test inputs. It evaluates coverage on 59 benchmark apps and uses additional Google Play apps to study crash discovery.", "topics": ["automation", "evaluation", "mobile-navigation"], "updated_at": "2025-04-16", "url": "https://arxiv.org/abs/2504.11675"}
{"aliases": [], "authors": ["Guangyi Liu", "Pengxiang Zhao", "Liang Liu", "Zhiming Chen", "Yuxiang Chai", "Shuai Ren", "Hao Wang", "Shibo He", "Wenchao Meng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2bbe6c4969a9557546752235809d25912bff96a97001fade30102d8d2c751054", "submitted_at": "2025-04-18T17:13:34Z", "updated_at": "2025-04-18T17:13:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.13805v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.13805", "identifiers": {"arxiv": "2504.13805"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.13805"}], "name": "LearnAct: Few-Shot Mobile GUI Agent with a Unified Demonstration Benchmark", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "LearnAct parses human demonstrations into reusable knowledge, retrieves relevant examples, and uses them during mobile GUI task execution. The accompanying LearnGUI benchmark includes 2,252 offline tasks and 101 online tasks for studying adaptation from demonstrations.", "topics": ["mobile-navigation", "memory", "human-in-the-loop", "evaluation"], "updated_at": "2025-04-18", "url": "https://arxiv.org/abs/2504.13805"}
{"aliases": [], "authors": ["Leo Boisvert", "Mihir Bansal", "Chandra Kiran Reddy Evuru", "Gabriel Huang", "Abhay Puri", "Avinandan Bose", "Maryam Fazel", "Quentin Cappart", "Jason Stanley", "Alexandre Lacoste", "Alexandre Drouin", "Krishnamurthy Dvijotham"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c5bdf5fefa1ee31126d99de08236991df559896212f60e0b3058fcd8cc8a134b", "submitted_at": "2025-04-18T20:36:10Z", "updated_at": "2025-10-07T13:10:30Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2504.14064v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}], "id": "arxiv:2504.14064", "identifiers": {"arxiv": "2504.14064"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:08Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.14064"}], "name": "DoomArena: A framework for Testing AI Agents Against Evolving Security Threats", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents a modular security testing framework that plugs into BrowserGym and tool-use benchmarks to evaluate agents under configurable threat models.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-10-07", "url": "https://arxiv.org/abs/2504.14064"}
{"aliases": [], "authors": ["Chaoran Chen", "Zhiping Zhang", "Ibrahim Khalilov", "Bingcan Guo", "Simret A Gebreegziabher", "Yanfang Ye", "Ziang Xiao", "Yaxing Yao", "Tianshi Li", "Toby Jia-Jun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d2a55e347bc10e5571464924d9301ff2c32d5e9f1f990ca91725ec8c9f6da594", "submitted_at": "2025-04-24T20:51:20Z", "updated_at": "2025-06-05T12:40:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.17934v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.17934", "identifiers": {"arxiv": "2504.17934"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.17934"}], "name": "Toward a Human-Centered Evaluation Framework for Trustworthy LLM-Powered GUI Agents", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. This is a position paper proposing an evaluation framework; the abstract does not report a new benchmark or empirical validation of the proposed framework.", "published_at": "2025-04-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents a position paper on privacy and security evaluation for GUI agents, identifying risk categories and challenges in involving human evaluators. It proposes risk-aware assessment and in-context consent as design directions.", "topics": ["evaluation", "safety-security", "human-in-the-loop"], "updated_at": "2025-06-05", "url": "https://arxiv.org/abs/2504.17934"}
{"aliases": [], "authors": ["Ivan Evtimov", "Arman Zharmagambetov", "Aaron Grattafiori", "Chuan Guo", "Kamalika Chaudhuri"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3a104d9076797555795766749f6b4eeec9428d0efb9340f7e55cffecd5d727b0", "submitted_at": "2025-04-22T17:51:03Z", "updated_at": "2025-05-16T22:42:29Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2504.18575v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.18575", "identifiers": {"arxiv": "2504.18575"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.18575"}], "name": "WASP: Benchmarking Web Agent Security Against Prompt Injection Attacks", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-04-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WASP, an end-to-end benchmark for prompt-injection security in web agents. Its evaluation separates partial attack progress from completion of the attacker’s full objective.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2025-05-16", "url": "https://arxiv.org/abs/2504.18575"}
{"aliases": [], "authors": ["Mariya Davydova", "Daniel Jeffries", "Patrick Barker", "Arturo Márquez Flores", "Sinéad Ryan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e8bb4c6f8080ed1ee91e1089233cae6f231ab3256b38ac7d6403ebea3ae56fe0", "submitted_at": "2025-05-06T14:29:47Z", "updated_at": "2025-05-06T14:29:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.03570v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:47Z", "role": "primary", "sha256": "27144160e8981bb9f36b60c391689774913ff3f026eeeffc2923154c404a6529", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2504.03206%2C2504.03601%2C2504.06821%2C2504.09723%2C2504.11301%2C2504.11543%2C2504.12609%2C2504.14233%2C2504.14348%2C2504.14603%2C2504.16073%2C2504.16419%2C2504.17892%2C2504.18580%2C2504.20965%2C2505.00019%2C2505.00693%2C2505.01578%2C2505.02094%2C2505.03570%2C2505.03574%2C2505.05026%2C2505.05849%2C2505.06507%2C2505.07003%2C2505.07473%2C2505.10468%2C2505.10887%2C2505.12053%2C2505.12490%2C2505.12728%2C2505.13291%2C2505.14030%2C2505.14454%2C2505.14646%2C2505.15801%2C2505.15816%2C2505.17020%2C2505.18746%2C2505.19095%2C2505.19713%2C2505.20023%2C2505.20732%2C2505.20795%2C2505.21055%2C2505.22304%2C2505.22429%2C2505.23450%2C2505.23559%2C2506.00641&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:17Z", "role": "primary", "sha256": "39de8f4a9aee74cc0eae64e24efc08ed070714eaaea4a30dbc20cb08397eb9bd", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2505.03570v1"}], "id": "arxiv:2505.03570", "identifiers": {"arxiv": "2505.03570"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:17Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.03570"}], "name": "OSUniverse: Benchmark for Multimodal GUI-navigation AI Agents", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OSUniverse, a desktop GUI benchmark spanning precise clicks through multistep and multiapplication tasks. It provides both manual scoring and an automated validation mechanism.", "topics": ["evaluation", "cross-app", "long-horizon"], "updated_at": "2025-05-06", "url": "https://arxiv.org/abs/2505.03570"}
{"aliases": [], "authors": ["Jaehyun Jeon", "Min Soo Kim", "Jang Han Yoon", "Sumin Shim", "Yejin Choi", "Hanbin Kim", "Dae Hyun Kim", "Youngjae Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3f65cd2856c554584baab9b1cbf9c5250b5c01d4f31980d69aefab727427ded2", "submitted_at": "2025-05-08T08:00:32Z", "updated_at": "2026-06-04T06:58:13Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2505.05026v5"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:47Z", "role": "primary", "sha256": "27144160e8981bb9f36b60c391689774913ff3f026eeeffc2923154c404a6529", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2504.03206%2C2504.03601%2C2504.06821%2C2504.09723%2C2504.11301%2C2504.11543%2C2504.12609%2C2504.14233%2C2504.14348%2C2504.14603%2C2504.16073%2C2504.16419%2C2504.17892%2C2504.18580%2C2504.20965%2C2505.00019%2C2505.00693%2C2505.01578%2C2505.02094%2C2505.03570%2C2505.03574%2C2505.05026%2C2505.05849%2C2505.06507%2C2505.07003%2C2505.07473%2C2505.10468%2C2505.10887%2C2505.12053%2C2505.12490%2C2505.12728%2C2505.13291%2C2505.14030%2C2505.14454%2C2505.14646%2C2505.15801%2C2505.15816%2C2505.17020%2C2505.18746%2C2505.19095%2C2505.19713%2C2505.20023%2C2505.20732%2C2505.20795%2C2505.21055%2C2505.22304%2C2505.22429%2C2505.23450%2C2505.23559%2C2506.00641&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:21Z", "role": "primary", "sha256": "1220bb1498a7f9c68cff0fbbd8eee209dfed138e771dfd6eb946fc0639c49cd1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2505.05026v5"}], "id": "arxiv:2505.05026", "identifiers": {"arxiv": "2505.05026"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:21Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.05026"}], "name": "Do MLLMs Capture How Interfaces Guide User Behavior? A Benchmark for Multimodal UI/UX Design Understanding", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced. This benchmark measures UI/UX design understanding from static image pairs; it does not measure autonomous GUI action execution.", "published_at": "2025-05-08", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces WiserUI-Bench, pairing real UI variants from industry A/B tests with observed behavioral winners and expert explanations. It evaluates models’ ability to predict the more effective interface and explain its effect on user behavior.", "topics": ["screen-understanding", "ui-design", "evaluation"], "updated_at": "2026-06-04", "url": "https://arxiv.org/abs/2505.05026"}
{"aliases": [], "authors": ["Shuning Zhang", "Jingruo Chen", "Zhiqi Gao", "Jiajing Gao", "Xin Yi", "Hewu Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "63f4a505645fec16eac6b59923a6c1979c53376f19c6a4ea53c9a5d32a45e3d4", "submitted_at": "2025-05-15T00:42:51Z", "updated_at": "2025-05-16T07:01:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.09875v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.09875", "identifiers": {"arxiv": "2505.09875"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.09875"}], "name": "Characterizing Unintended Consequences in Human-GUI Agent Collaboration for Web Browsing", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies unintended consequences of collaborative web browsing with GUI agents through social-media analysis and interviews. It organizes observed failures, user impacts, and mitigation practices around planning, interface interaction, feedback, privacy, and trust.", "topics": ["human-in-the-loop", "evaluation", "safety-security"], "updated_at": "2025-05-16", "url": "https://arxiv.org/abs/2505.09875"}
{"aliases": [], "authors": ["Weikai Xu", "Zhizheng Jiang", "Yuxuan Liu", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Yuanchun Li", "Yunxin Liu", "Bin Wang", "Bo An"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5ca1ebfdd5a96e146d37539294462c7fb67bf20deaf77307de37aa3bb1dde082", "submitted_at": "2025-05-17T07:58:34Z", "updated_at": "2026-02-02T06:55:10Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2505.11891v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2505.11891", "identifiers": {"arxiv": "2505.11891"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.11891"}], "name": "Mobile-Bench-v2: A More Realistic and Comprehensive Benchmark for VLM-based Mobile Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Mobile-Bench-v2 evaluates mobile GUI agents using multiple acceptable offline action paths, noisy interfaces with ads or pop-ups, and ambiguous instructions with scripted clarification exchanges. These splits separate step-level execution, robustness, and proactive interaction.", "topics": ["evaluation", "mobile-navigation", "robustness", "human-in-the-loop"], "updated_at": "2026-02-02", "url": "https://arxiv.org/abs/2505.11891"}
{"aliases": [], "authors": ["Zheng Wu", "Pengzhou Cheng", "Zongru Wu", "Lingzhong Dong", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ce9570eb6a52feb220cdd2eb4c7f37a2ea63082febe48888a2654c974138a82d", "submitted_at": "2025-05-19T08:29:05Z", "updated_at": "2025-11-22T15:07:27Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2505.12842v4"}}, "environments": ["desktop", "web", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.12842", "identifiers": {"arxiv": "2505.12842"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.12842"}], "name": "GEM: Gaussian Embedding Modeling for Out-of-Distribution Detection in GUI Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-05-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "GEM detects instructions outside a GUI agent’s capabilities by modeling distances in its input-embedding space with a Gaussian mixture. Experiments across mobile, computer and browser data also examine selective escalation to a cloud model when an input is detected as unfamiliar.", "topics": ["robustness", "safety-security", "evaluation", "efficiency"], "updated_at": "2025-11-22", "url": "https://arxiv.org/abs/2505.12842"}
{"aliases": [], "authors": ["Liangxuan Wu", "Chao Wang", "Tianming Liu", "Yanjie Zhao", "Haoyu Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9657a66aebe46151ed56bf76131ffe45f52459c1cf0e000907113776c91259d9", "submitted_at": "2025-05-19T11:17:46Z", "updated_at": "2025-05-20T07:02:05Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.12981v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}], "id": "arxiv:2505.12981", "identifiers": {"arxiv": "2505.12981"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:08Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.12981"}], "name": "From Assistants to Adversaries: Exploring the Security Risks of Mobile LLM Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Analyzes security vulnerabilities across OEM system-level, third-party, and research mobile LLM agents, covering GUI and cross-application attack surfaces.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2025-05-20", "url": "https://arxiv.org/abs/2505.12981"}
{"aliases": [], "authors": ["Mykyta Mudryi", "Markiyan Chaklosh", "Grzegorz Wójcik"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "709ee511ffa39e805fcc58abe066cf2cec6c1707a4a9c64d9cfd107bef7544af", "submitted_at": "2025-05-19T13:10:29Z", "updated_at": "2025-05-19T13:10:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.13076v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}], "id": "arxiv:2505.13076", "identifiers": {"arxiv": "2505.13076"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:08Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.13076"}], "name": "The Hidden Dangers of Browsing AI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Outlines an end-to-end threat model and empirical security evaluation across architectural layers of autonomous browser agents.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2025-05-19", "url": "https://arxiv.org/abs/2505.13076"}
{"aliases": [], "authors": ["Tianbao Xie", "Jiaqi Deng", "Xiaochuan Li", "Junlin Yang", "Haoyuan Wu", "Jixuan Chen", "Wenjing Hu", "Xinyuan Wang", "Yuhui Xu", "Zekun Wang", "Yiheng Xu", "Junli Wang", "Doyen Sahoo", "Tao Yu", "Caiming Xiong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1bec4ad578c01ebdfb58bd269562d3ca7a2bd8d2687cde07cb867db0b78cb105", "submitted_at": "2025-05-19T15:09:23Z", "updated_at": "2025-10-24T19:08:03Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2505.13227v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.13227", "identifiers": {"arxiv": "2505.13227"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.13227"}], "name": "Scaling Computer-Use Grounding via User Interface Decomposition and Synthesis", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2025-05-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OSWorld-G grounding evaluation and the Jedi synthesis pipeline for examples spanning layout reasoning and fine-grained interaction.", "topics": ["grounding", "data-generation", "evaluation"], "updated_at": "2025-10-24", "url": "https://arxiv.org/abs/2505.13227"}
{"aliases": [], "authors": ["Yanheng He", "Jiahe Jin", "Pengfei Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "026a64046926912a54e6a7c6cf62a17740a6f89a94713615f9750f8bf8302c78", "submitted_at": "2025-05-20T04:20:18Z", "updated_at": "2026-03-03T03:55:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.13909v2"}}, "environments": ["windows", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.13909", "identifiers": {"arxiv": "2505.13909"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.13909"}], "name": "Efficient Agent Training for Computer Use", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "PC Agent-E augments 312 human computer-use trajectories with alternative action decisions synthesized by Claude 3.7 Sonnet. It studies training with this combination and introduces WindowsAgentArena-V2 as an updated evaluation benchmark.", "topics": ["data-generation", "evaluation"], "updated_at": "2026-03-03", "url": "https://arxiv.org/abs/2505.13909"}
{"aliases": [], "authors": ["Yijie Lu", "Manman Zhao", "Tianjie Ju", "Zihe Yan", "Xinbei Ma", "Yuan Guo", "Daizong Ding", "Gongshen Liu", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2c256d8b0d774efdeee896860f8c504c0d1593215508fdf0881d3fd3cdc3968c", "submitted_at": "2025-05-20T12:41:05Z", "updated_at": "2026-06-05T10:24:33Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.14289v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.14289", "identifiers": {"arxiv": "2505.14289"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.14289"}], "name": "EVA: Evolving Semantic Adversaries for Red-Teaming GUI Agents Against Environmental Injection Attacks", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces EVA, a red-teaming framework that evolves semantic payloads for GUI environmental-injection tests. It studies the contribution of semantic versus visual deception and distills discovered patterns into reusable evaluation rules.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-06-05", "url": "https://arxiv.org/abs/2505.14289"}
{"aliases": [], "authors": ["Hyungjoo Chae", "Sunghwan Kim", "Junhee Cho", "Seungone Kim", "Seungjun Moon", "Gyeom Hwangbo", "Dongha Lim", "Minjin Kim", "Yeonjun Hwang", "Minju Gwak", "Dongwook Choi", "Minseok Kang", "Gwanhoon Im", "ByeongUng Cho", "Hyojun Kim", "Jun Hee Han", "Taeyoon Kwon", "Minju Kim", "Beong-woo Kwak", "Dongjin Kang", "Jinyoung Yeo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "440abdc75c0a51478cfedc6ee0a9196ccff641ec5a23849ba56be43448bab15f", "submitted_at": "2025-05-21T08:56:55Z", "updated_at": "2025-11-25T21:34:57Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.15277v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.15277", "identifiers": {"arxiv": "2505.15277"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.15277"}], "name": "Web-Shepherd: Advancing PRMs for Reinforcing Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Web-Shepherd is a process reward model for assessing individual steps of web-navigation trajectories. It introduces a collection of 40,000 step-level preference pairs with checklists and a separate benchmark for comparing reward models, then evaluates verification during WebArena-lite navigation.", "topics": ["reinforcement-learning", "evaluation", "data-generation", "web-navigation"], "updated_at": "2025-11-25", "url": "https://arxiv.org/abs/2505.15277"}
{"aliases": [], "authors": ["Qiushi Sun", "Zhoumianze Liu", "Chang Ma", "Zichen Ding", "Fangzhi Xu", "Zhangyue Yin", "Haiteng Zhao", "Zhenyu Wu", "Kanzhi Cheng", "Zhaoyang Liu", "Jianing Wang", "Qintong Li", "Xiangru Tang", "Tianbao Xie", "Xiachong Feng", "Xiang Li", "Ben Kao", "Wenhai Wang", "Biqing Qi", "Lingpeng Kong", "Zhiyong Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4743f6bfd0ca48717184cf55545e33a229dd5ab2d454385d33ea2f1b25c15b0c", "submitted_at": "2025-05-26T12:27:27Z", "updated_at": "2026-04-20T16:29:04Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2505.19897v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2505.19897", "identifiers": {"arxiv": "2505.19897"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.19897"}], "name": "ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "ScienceBoard combines an interactive environment with professional scientific software and 169 human-curated tasks in areas including biochemistry, astronomy, and geoinformatics. It evaluates computer-use agents on visually rich research workflows and analyzes their execution limitations.", "topics": ["evaluation", "science", "workflow"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2505.19897"}
{"aliases": [], "authors": ["Shaoqing Zhang", "Kehai Chen", "Zhuosheng Zhang", "Rumei Li", "Rongxiang Weng", "Yang Xiang", "Min Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4fe40099905712f9d7ba89bc7616d09b260fe6113beb8122cade54a66d46e3d7", "submitted_at": "2025-05-27T14:49:30Z", "updated_at": "2025-09-26T09:13:44Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.21279v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2505.21279", "identifiers": {"arxiv": "2505.21279"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.21279"}], "name": "XBOUND: Exploring Capability Boundaries of Device-Control Agents at the State Level", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-05-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "XBOUND evaluates GUI agents at the level of a screen state, considering multiple instructions and possible target actions within it. The study separates action matching from instruction unification and compares effects of grounding and trajectory data.", "topics": ["evaluation", "grounding"], "updated_at": "2025-09-26", "url": "https://arxiv.org/abs/2505.21279"}
{"aliases": [], "authors": ["Haowei Wang", "Junjie Wang", "Xiaojun Jia", "Rupeng Zhang", "Mingyang Li", "Zhe Liu", "Yang Liu", "Qing Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f8aade17e67471c87f259d49aef43b3caf764b6ac14f99e1a4114f4a25dda3d6", "submitted_at": "2025-05-27T17:59:05Z", "updated_at": "2025-05-27T17:59:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.21499v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.21499", "identifiers": {"arxiv": "2505.21499"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.21499"}], "name": "AdInject: Real-World Black-Box Attacks on Web Agents via Advertising Delivery", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies advertising delivery as an indirect-injection channel for web agents under a black-box threat model. AdInject evaluates how static ad content shaped by website context can redirect browser actions.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2025-05-27", "url": "https://arxiv.org/abs/2505.21499"}
{"aliases": [], "authors": ["Zeyi Liao", "Jaylen Jones", "Linxi Jiang", "Yuting Ning", "Eric Fosler-Lussier", "Yu Su", "Zhiqiang Lin", "Huan Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7fe3bdb1357a375de0b1700a6eda7d5987f358bd47fddf3405c27272a3453f2f", "submitted_at": "2025-05-28T03:42:09Z", "updated_at": "2026-03-01T07:50:09Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2505.21936v5"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.21936", "identifiers": {"arxiv": "2505.21936"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.21936"}], "name": "RedTeamCUA: Realistic Adversarial Testing of Computer-Use Agents in Hybrid Web-OS Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "RedTeamCUA combines a virtual-machine OS environment with containerized websites for controlled prompt-injection tests spanning both interfaces. Its RTC-Bench contains 864 examples and can initialize agents at injection points to separate attack susceptibility from navigation failures.", "topics": ["safety-security", "evaluation", "sandbox"], "updated_at": "2026-03-01", "url": "https://arxiv.org/abs/2505.21936"}
{"aliases": [], "authors": ["Yaxin Luo", "Zhaoyi Li", "Jiacheng Liu", "Jiacheng Cui", "Xiaohan Zhao", "Zhiqiang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3f2206333ca25b3f21df7ce49adfd8bc56521b9b8ace7a9b686eccf5b67276fa", "submitted_at": "2025-05-30T17:59:55Z", "updated_at": "2025-05-30T17:59:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.24878v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2505.24878", "identifiers": {"arxiv": "2505.24878"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.24878"}], "name": "Open CaptchaWorld: A Comprehensive Web-based Platform for Testing and Benchmarking Multimodal LLM Agents", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-05-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Open CaptchaWorld evaluates multimodal agents on interactive browser-based CAPTCHA puzzles. Its benchmark distinguishes task success from the depth of visual reasoning and interaction needed to complete a puzzle.", "topics": ["evaluation", "web-navigation", "screen-understanding"], "updated_at": "2025-05-30", "url": "https://arxiv.org/abs/2505.24878"}
{"aliases": [], "authors": ["Jingyi Yang", "Shuai Shao", "Dongrui Liu", "Jing Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bb94fdba775cb0fb54b312fdff9fa71afd3aefa56b812ff31ceccad33eb5f800", "submitted_at": "2025-05-31T16:04:59Z", "updated_at": "2025-06-20T01:24:06Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2506.00618v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.00618", "identifiers": {"arxiv": "2506.00618"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.00618"}], "name": "RiOSWorld: Benchmarking the Risk of Multimodal Computer-Use Agents", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces RiOSWorld, a computer-use risk benchmark separating user-originated and environmental threats. It evaluates both risky-goal intent and goal completion across web and other computer applications.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-06-20", "url": "https://arxiv.org/abs/2506.00618"}
{"aliases": [], "authors": ["Xiao Yang", "Jiawei Chen", "Jun Luo", "Zhengwei Fang", "Yinpeng Dong", "Hang Su", "Jun Zhu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1ec2467703bc5d0f97ef8d207289ed56f6d12b811688d20b7833eac06038757b", "submitted_at": "2025-06-02T12:56:27Z", "updated_at": "2025-06-02T12:56:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.01616v1"}}, "environments": ["web", "android", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:08Z", "role": "primary", "sha256": "00d275cd5407cec847fa9d43fe73cfc4384c5570b68a5e058c97921799be0bec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2302.01560%2C2303.16563%2C2305.15486%2C2305.16291%2C2305.17144%2C2306.00003%2C2306.00349%2C2306.17882%2C2310.01557%2C2310.08235%2C2310.08367%2C2311.05997%2C2311.17600%2C2312.02519%2C2312.11865%2C2402.01118%2C2402.10196%2C2402.11208%2C2403.10249%2C2403.12037%2C2404.06647%2C2404.12241%2C2404.13246%2C2407.15325%2C2409.00138%2C2410.14573%2C2410.18927%2C2410.19012%2C2411.09523%2C2411.10349%2C2411.13543%2C2412.01547%2C2501.05366%2C2501.09004%2C2502.07463%2C2502.14538%2C2503.02732%2C2503.08175%2C2503.11387%2C2504.10147%2C2504.11703%2C2504.14064%2C2504.15810%2C2504.16208%2C2504.20635%2C2504.21776%2C2505.12981%2C2505.13076%2C2506.01616%2C2506.13205&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:55Z", "role": "primary", "sha256": "d69da19edaed185ad3071a97db3bb3fcfbdb14c0b514b7d27794b39fbe8e2541", "supports": ["environments", "notes"], "url": "https://raw.githubusercontent.com/thu-ml/MLA-Trust/HEAD/README.md"}], "id": "arxiv:2506.01616", "identifiers": {"arxiv": "2506.01616"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:08Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.01616"}], "name": "MLA-Trust: Benchmarking Trustworthiness of Multimodal LLM Agents in GUI Environments", "notes": "The contribution was reviewed from the retained paper abstract; Android setup is supported separately by the official project README. No benchmark results were independently reproduced.", "published_at": "2025-06-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Benchmarks truthfulness, controllability, safety, and privacy of multimodal LLM agents across interactive web and mobile GUI tasks.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2025-06-02", "url": "https://arxiv.org/abs/2506.01616"}
{"aliases": [], "authors": ["Tri Cao", "Bennett Lim", "Yue Liu", "Yuan Sui", "Yuexin Li", "Shumin Deng", "Lin Lu", "Nay Oo", "Shuicheng Yan", "Bryan Hooi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3cd988217d1e3371bec827b472abf46c205ac407b5b60b7b581d7105f86e8bc5", "submitted_at": "2025-06-03T05:21:50Z", "updated_at": "2026-03-01T16:12:40Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.02456v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.02456", "identifiers": {"arxiv": "2506.02456"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.02456"}], "name": "VPI-Bench: Visual Prompt Injection Attacks for Computer-Use Agents", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-06-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces VPI-Bench, interactive web-platform variants containing visually embedded malicious instructions. It evaluates computer-use and browser-use agents under these threats and examines system-prompt defenses.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-03-01", "url": "https://arxiv.org/abs/2506.02456"}
{"aliases": [], "authors": ["Yougang Lyu", "Xiaoyu Zhang", "Lingyong Yan", "Maarten de Rijke", "Zhaochun Ren", "Xiuying Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bebdf1c8f34582408ef0050245f6b1e454d1bd2e75903a4e0817a4f3c00fb4db", "submitted_at": "2025-06-03T13:08:17Z", "updated_at": "2025-06-03T13:08:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.02839v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2506.02839", "identifiers": {"arxiv": "2506.02839"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.02839"}], "name": "DeepShop: A Benchmark for Deep Research Shopping Agents", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-06-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "DeepShop evaluates shopping agents with evolving query complexity across product attributes, filters and sorting preferences. Its evaluation combines checks of these individual requirements with overall task success.", "topics": ["evaluation", "web-navigation"], "updated_at": "2025-06-03", "url": "https://arxiv.org/abs/2506.02839"}
{"aliases": [], "authors": ["Mathieu Andreux", "Breno Baldas Skuk", "Hamza Benchekroun", "Emilien Biré", "Antoine Bonnet", "Riaz Bordie", "Nathan Bout", "Matthias Brunel", "Pierre-Louis Cedoz", "Antoine Chassang", "Mickaël Chen", "Alexandra D. Constantinou", "Antoine d'Andigné", "Hubert de La Jonquière", "Aurélien Delfosse", "Ludovic Denoyer", "Alexis Deprez", "Augustin Derupti", "Michael Eickenberg", "Mathïs Federico", "Charles Kantor", "Xavier Koegler", "Yann Labbé", "Matthew C. H. Lee", "Erwan Le Jumeau de Kergaradec", "Amir Mahla", "Avshalom Manevich", "Adrien Maret", "Charles Masson", "Rafaël Maurin", "Arturo Mena", "Philippe Modard", "Axel Moyal", "Axel Nguyen Kerbel", "Julien Revelle", "Mats L. Richter", "María Santos", "Laurent Sifre", "Maxime Theillard", "Marc Thibault", "Louis Thiry", "Léo Tronchon", "Nicolas Usunier", "Tony Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "89c2d658e03204b56f6d273da9572820f9a23ca359a441ea87baa451a93c5b58", "submitted_at": "2025-06-03T13:29:03Z", "updated_at": "2025-06-11T09:33:38Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.02865v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.02865", "identifiers": {"arxiv": "2506.02865"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.02865"}], "name": "Surfer-H Meets Holo1: Cost-Efficient Web Agent Powered by Open Weights", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes Surfer-H, a web agent paired with the Holo1 vision-language model family trained on web content, synthetic examples, and agent-generated data. The work introduces WebClick for web-interface localization and evaluates browser task execution on WebVoyager.", "topics": ["web-navigation", "grounding", "data-generation", "evaluation"], "updated_at": "2025-06-11", "url": "https://arxiv.org/abs/2506.02865"}
{"aliases": [], "authors": ["Man Luo", "David Cobbley", "Xin Su", "Shachar Rosenman", "Vasudev Lal", "Shao-Yen Tseng", "Phillip Howard"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "068eae2cab9899967850f21616a701f3eeb258b9654c83d34f68be266e48854e", "submitted_at": "2025-06-03T17:27:04Z", "updated_at": "2025-06-03T17:27:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.03095v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.03095", "identifiers": {"arxiv": "2506.03095"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.03095"}], "name": "DPO Learning with LLMs-Judge Signal for Computer Use Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-06-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses an LLM judge to evaluate and filter synthetic GUI interaction trajectories for training a compact vision-language agent. OSWorld experiments assess the policy designed for local execution.", "topics": ["data-generation", "evaluation", "efficiency"], "updated_at": "2025-06-03", "url": "https://arxiv.org/abs/2506.03095"}
{"aliases": [], "authors": ["Pei Yang", "Hai Ci", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8a4c980694819dc5eb4ff322e6a703cccfe0b1e45efb60f22b120592d03edf36", "submitted_at": "2025-06-04T16:26:56Z", "updated_at": "2025-10-18T12:11:31Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2506.04135v4"}}, "environments": ["macos", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.04135", "identifiers": {"arxiv": "2506.04135"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.04135"}], "name": "macOSWorld: A Multilingual Interactive Benchmark for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "macOSWorld provides 202 interactive tasks across 30 applications, with instructions and operating-system interfaces in five languages. It evaluates GUI agents on macOS workflows and includes a separate subset for deception-attack testing.", "topics": ["evaluation", "workflow", "safety-security"], "updated_at": "2025-10-18", "url": "https://arxiv.org/abs/2506.04135"}
{"aliases": [], "authors": ["Zihan Deng", "Changyu Du", "Stavros Nousias", "André Borrmann"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2bf4f91de32fdf3bd9b81ba64ace02e4a7d02ff094794f30dfd560389f334cea", "submitted_at": "2025-06-08T16:45:31Z", "updated_at": "2025-06-30T08:31:07Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.07217v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.07217", "identifiers": {"arxiv": "2506.07217"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.07217"}], "name": "BIMgent: Towards Autonomous Building Modeling via Computer-use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "BIMgent plans software-specific workflows and executes GUI actions for building-information modeling. It evaluates multimodal building design and reconstruction tasks in professional authoring software.", "topics": ["industry", "planning", "workflow", "evaluation"], "updated_at": "2025-06-30", "url": "https://arxiv.org/abs/2506.07217"}
{"aliases": [], "authors": ["Ruofan Liu", "Xiwen Teoh", "Yun Lin", "Guanjie Chen", "Ruofei Ren", "Denys Poshyvanyk", "Jin Song Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bea3e76b3124ebbcd8f7817b3e32d667d8aaae6e31f1b75ac3fc8c39e7637c6f", "submitted_at": "2025-06-09T03:09:48Z", "updated_at": "2025-06-09T03:09:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.07385v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:51Z", "role": "primary", "sha256": "b91987360225499418fcd3645d7d5effac91e8202bf7c7e17c9f9e31d25b5be8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.01423%2C2506.03231%2C2506.04301%2C2506.05336%2C2506.07385%2C2506.07398%2C2506.08136%2C2506.08837%2C2506.10953%2C2506.11425%2C2506.11763%2C2506.13538%2C2506.14683%2C2506.15461%2C2506.18959%2C2506.19502%2C2506.20668%2C2506.21252%2C2506.21495%2C2506.21614%2C2507.01599%2C2507.02279%2C2507.02825%2C2507.04036%2C2507.07400%2C2507.09481%2C2507.09884%2C2507.13348%2C2507.14447%2C2507.16704%2C2507.19902%2C2507.21678%2C2507.22925%2C2508.01548%2C2508.02866%2C2508.04002%2C2508.04389%2C2508.05193%2C2508.05294%2C2508.05748%2C2508.07871%2C2508.07999%2C2508.08228%2C2508.09124%2C2508.09981%2C2508.15804%2C2508.16153%2C2508.17398%2C2508.18489%2C2508.19611&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:30Z", "role": "primary", "sha256": "23030fd266519162c51c7f4a87bc38e75bb171ef7fa95e30c2d7afd820a0bc3f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2506.07385v1"}], "id": "arxiv:2506.07385", "identifiers": {"arxiv": "2506.07385"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:30Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.07385"}], "name": "GUIPilot: A Consistency-based Mobile GUI Testing Approach for Detecting Application-specific Bugs", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUIPilot, which compares mobile design mock-ups with implemented screens through widget alignment and tests expected transitions by executing inferred UI actions. It evaluates both visual inconsistencies and workflow-specific bugs.", "topics": ["evaluation", "mobile-navigation", "robustness"], "updated_at": "2025-06-09", "url": "https://arxiv.org/abs/2506.07385"}
{"aliases": [], "authors": ["Yunhe Yan", "Shihe Wang", "Jiajun Du", "Yexuan Yang", "Yuxuan Shan", "Qichen Qiu", "Xianqing Jia", "Xinge Wang", "Xin Yuan", "Xu Han", "Mao Qin", "Yinxiao Chen", "Chen Peng", "Shangguang Wang", "Mengwei Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2334e017fe49b39dc5d460963988f05d11051f2101bebe23be76005809fafd45", "submitted_at": "2025-06-09T11:50:33Z", "updated_at": "2025-06-09T11:50:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.07672v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:12:09Z", "role": "primary", "sha256": "f64464b1516369a87f0fbe28828e32ff61687dae4ce063aedddd8bd45a7d40eb", "supports": ["scope", "summary", "environments"], "url": "https://arxiv.org/html/2506.07672"}], "id": "arxiv:2506.07672", "identifiers": {"arxiv": "2506.07672"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.07672"}], "name": "MCPWorld: A Unified Benchmarking Testbed for API, GUI, and Hybrid Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced. The platform or GUI evaluation setting was additionally checked in the primary full text.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "MCPWorld evaluates API, GUI, and hybrid computer-use agents with 201 tasks in applications whose source can be instrumented and extended with MCP support. It verifies completion from application behavior and packages the testbed in containers with GPU support.", "topics": ["evaluation", "tool-integration", "sandbox"], "updated_at": "2025-06-09", "url": "https://arxiv.org/abs/2506.07672"}
{"aliases": [], "authors": ["Penghao Wu", "Shengnan Ma", "Bo Wang", "Jiaheng Yu", "Lewei Lu", "Ziwei Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ce3b7d2c25b5a99f110f00be7389d4230d625ac1f7f33d5ab81f3d6f441f3610", "submitted_at": "2025-06-09T17:59:57Z", "updated_at": "2025-06-09T17:59:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.08012v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2506.08012", "identifiers": {"arxiv": "2506.08012"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.08012"}], "name": "GUI-Reflection: Empowering Multimodal GUI Models with Self-Reflection Behavior", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-Reflection synthesizes reflection and error-correction examples from successful trajectories and adds dedicated learning stages for these behaviors. It introduces a reflection task suite, a mobile environment for online data collection, and iterative reflection tuning.", "topics": ["robustness", "data-generation", "evaluation", "mobile-navigation"], "updated_at": "2025-06-09", "url": "https://arxiv.org/abs/2506.08012"}
{"aliases": [], "authors": ["Zefang Liu", "Yinzhu Quan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "43aca3b0b8a1a0d5729a4e90dc1516bbb68e3611de3220876cdb29afc81d847f", "submitted_at": "2025-06-09T18:39:48Z", "updated_at": "2026-05-11T08:25:57Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2506.08136v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:51Z", "role": "primary", "sha256": "b91987360225499418fcd3645d7d5effac91e8202bf7c7e17c9f9e31d25b5be8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.01423%2C2506.03231%2C2506.04301%2C2506.05336%2C2506.07385%2C2506.07398%2C2506.08136%2C2506.08837%2C2506.10953%2C2506.11425%2C2506.11763%2C2506.13538%2C2506.14683%2C2506.15461%2C2506.18959%2C2506.19502%2C2506.20668%2C2506.21252%2C2506.21495%2C2506.21614%2C2507.01599%2C2507.02279%2C2507.02825%2C2507.04036%2C2507.07400%2C2507.09481%2C2507.09884%2C2507.13348%2C2507.14447%2C2507.16704%2C2507.19902%2C2507.21678%2C2507.22925%2C2508.01548%2C2508.02866%2C2508.04002%2C2508.04389%2C2508.05193%2C2508.05294%2C2508.05748%2C2508.07871%2C2508.07999%2C2508.08228%2C2508.09124%2C2508.09981%2C2508.15804%2C2508.16153%2C2508.17398%2C2508.18489%2C2508.19611&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:34Z", "role": "primary", "sha256": "2f1d53911fe997bbba01c9d8a2fa7002f18837492448e15056fd7a7a6c78f97d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2506.08136v3"}], "id": "arxiv:2506.08136", "identifiers": {"arxiv": "2506.08136"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:34Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.08136"}], "name": "EconWebArena: Benchmarking Autonomous Agents on Economic Tasks in Realistic Web Environments", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces EconWebArena, a curated benchmark of economic-data tasks requiring live website navigation, multimodal interpretation, and precise extraction. It evaluates visual grounding, planning, and interaction design within multistep workflows.", "topics": ["evaluation", "web-navigation", "planning"], "updated_at": "2026-05-11", "url": "https://arxiv.org/abs/2506.08136"}
{"aliases": [], "authors": ["Yuan Guo", "Tingjia Miao", "Zheng Wu", "Pengzhou Cheng", "Ming Zhou", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8dbc84aa582b94775d490cf84e17eff05be84af7dbcc2de7de61b886fa3c9178", "submitted_at": "2025-06-10T16:45:29Z", "updated_at": "2025-06-10T16:45:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.08972v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2506.08972", "identifiers": {"arxiv": "2506.08972"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.08972"}], "name": "Atomic-to-Compositional Generalization for Mobile Agents with A New Benchmark and Scheduling System", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-NEXUS evaluates compositional mobile tasks in local utility apps and online Chinese and English service apps. AGENT-NEXUS decomposes those long tasks into self-contained atomic subtasks to study generalization beyond isolated screens or short action chains.", "topics": ["evaluation", "long-horizon", "planning", "mobile-navigation"], "updated_at": "2025-06-10", "url": "https://arxiv.org/abs/2506.08972"}
{"aliases": [], "authors": ["Jingxu Xie", "Dylan Xu", "Xuandong Zhao", "Dawn Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ae36b271f9b4799093831cabee772845edfaeca97107414e424e170ccf89904d", "submitted_at": "2025-06-17T05:46:52Z", "updated_at": "2026-03-02T00:51:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.14205v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:13+00:00", "role": "primary", "sha256": "42e8b8dead7caea3f8b21e93a075e8c895dfc30a25e1470bf97a90940bbf0106", "supports": ["scope", "environments"], "url": "https://raw.githubusercontent.com/sunblaze-ucb/AgentSynth/HEAD/README.md"}], "id": "arxiv:2506.14205", "identifiers": {"arxiv": "2506.14205"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.14205"}], "name": "AgentSynth: Scalable Task Generation for Generalist Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-06-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "AgentSynth composes easily generated subtasks into longer computer-use tasks and trajectories with adjustable difficulty. Its evaluation uses the number of subtasks to probe how agent success changes as task complexity grows.", "topics": ["data-generation", "evaluation", "long-horizon"], "updated_at": "2026-03-02", "url": "https://arxiv.org/abs/2506.14205"}
{"aliases": [], "authors": ["Jingqi Yang", "Zhilong Song", "Jiawei Chen", "Mingli Song", "Sheng Zhou", "linjun sun", "Xiaogang Ouyang", "Chun Chen", "Can Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f9e64118f03c2a9f3f55e35f79c38835f1eb823265c83c933fdceca2610b9904", "submitted_at": "2025-06-17T12:50:35Z", "updated_at": "2025-06-17T12:50:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.14477v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.14477", "identifiers": {"arxiv": "2506.14477"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.14477"}], "name": "GUI-Robust: A Comprehensive Dataset for Testing GUI Agent Robustness in Real-World Anomalies", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds GUI task data with real-world interaction anomalies and evaluates how agent behavior changes under those conditions.", "topics": ["robustness", "evaluation", "data-generation"], "updated_at": "2025-06-17", "url": "https://arxiv.org/abs/2506.14477"}
{"aliases": [], "authors": ["Thomas Kuntz", "Agatha Duzan", "Hao Zhao", "Francesco Croce", "Zico Kolter", "Nicolas Flammarion", "Maksym Andriushchenko"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f551b861012fc4eddbf010a619bc65e77b524b05958761b738a88bcae377e9a9", "submitted_at": "2025-06-17T17:59:31Z", "updated_at": "2025-10-29T10:34:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.14866v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.14866", "identifiers": {"arxiv": "2506.14866"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.14866"}], "name": "OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2025-06-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Extends OSWorld with tasks covering user misuse, prompt injection, and model misbehavior, with assessment of both task success and safety.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-10-29", "url": "https://arxiv.org/abs/2506.14866"}
{"aliases": [], "authors": ["Xingjian Tao", "Yiwei Wang", "Yujun Cai", "Zhicheng Yang", "Jing Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e3884b1807cbee971364947aa2f3756937c4c774d9942bb4da3d0fabb69bda07", "submitted_at": "2025-06-18T12:55:35Z", "updated_at": "2025-06-18T12:55:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.15425v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.15425", "identifiers": {"arxiv": "2506.15425"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.15425"}], "name": "Understanding GUI Agent Localization Biases through Logit Sharpness", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a finer-grained taxonomy of GUI localization errors and a logit-based Peak Sharpness Score for analyzing coordinate-prediction uncertainty. A context-aware cropping method refines the visual input without retraining the grounding model.", "topics": ["grounding", "evaluation", "robustness", "training-free"], "updated_at": "2025-06-18", "url": "https://arxiv.org/abs/2506.15425"}
{"aliases": [], "authors": ["Yining Hong", "Rui Sun", "Bingxuan Li", "Xingcheng Yao", "Maxine Wu", "Alexander Chien", "Da Yin", "Ying Nian Wu", "Zhecan James Wang", "Kai-Wei Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3e8f714b234a1aad60d0306df23c90410aebfd257fb5906644d0382479e8f3d0", "submitted_at": "2025-06-18T17:58:17Z", "updated_at": "2025-07-29T22:40:49Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2506.15677v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.15677", "identifiers": {"arxiv": "2506.15677"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.15677"}], "name": "Embodied Web Agents: Bridging Physical-Digital Realms for Integrated Agent Intelligence", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The physical component is a simulation; the supported catalog connection is its functional web interaction.", "published_at": "2025-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces environments that couple 3D physical-world simulations with functional web interfaces. The accompanying benchmark evaluates tasks requiring agents to coordinate online information and physical actions across cooking, navigation and shopping scenarios.", "topics": ["evaluation", "web-navigation", "planning"], "updated_at": "2025-07-29", "url": "https://arxiv.org/abs/2506.15677"}
{"aliases": [], "authors": ["Reyna Abhyankar", "Qi Qi", "Yiying Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "df3f654ef1364aeb6cdaeaea1c6334da949a798b2f8256c9ddc4e6a88a588e19", "submitted_at": "2025-06-19T05:26:40Z", "updated_at": "2026-05-18T17:55:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.16042v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.16042", "identifiers": {"arxiv": "2506.16042"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.16042"}], "name": "OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "OSWorld-Human adds human-determined trajectories to OSWorld to evaluate the efficiency of desktop computer-use agents. The study compares 16 agents and analyzes model-call latency, changes in step duration during task execution, and excess action steps.", "topics": ["evaluation", "efficiency", "observability"], "updated_at": "2026-05-18", "url": "https://arxiv.org/abs/2506.16042"}
{"aliases": [], "authors": ["Atharv Singh Patlan", "Ashwin Hebbar", "Pramod Viswanath", "Prateek Mittal"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2d346389013d465f578417a7fca80ee527f5e7b582e93eb73dae148f11479971", "submitted_at": "2025-06-18T14:29:02Z", "updated_at": "2025-06-18T14:29:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.17318v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.17318", "identifiers": {"arxiv": "2506.17318"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.17318"}], "name": "Context manipulation attacks : Web agents are susceptible to corrupted memory", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies context manipulation in web agents by altering stored task representations rather than webpage prompts alone. Evaluation on Browser-use and Agent-E examines memory-directed attacks and the implications for secure context handling.", "topics": ["safety-security", "memory", "evaluation"], "updated_at": "2025-06-18", "url": "https://arxiv.org/abs/2506.17318"}
{"aliases": [], "authors": ["Jinjie Wei", "Jiyao Liu", "Lihao Liu", "Ming Hu", "Junzhi Ning", "Mingcheng Li", "Weijie Yin", "Junjun He", "Xiao Liang", "Chao Feng", "Dingkang Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f12d2ae58243b5f89a6a510e8f3ef4073a605c4bafdf63c37fd38db57ea84602", "submitted_at": "2025-06-22T06:30:52Z", "updated_at": "2026-07-21T05:30:48Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.17913v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.17913", "identifiers": {"arxiv": "2506.17913"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.17913"}], "name": "Learning, Reasoning, Refinement: A Framework for Kahneman's Dual-System Intelligence in GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "CogniGUI combines hierarchical visual parsing of interface elements with a GRPO-based grounding agent that compares alternative interaction paths. Its ScreenSeek benchmark evaluates navigation across applications, changing interface states, and consistency between interfaces.", "topics": ["grounding", "planning", "reinforcement-learning", "evaluation"], "updated_at": "2026-07-21", "url": "https://arxiv.org/abs/2506.17913"}
{"aliases": [], "authors": ["Siyuan Liang", "Tianmeng Fang", "Zhe Liu", "Aishan Liu", "Yan Xiao", "Jinyuan He", "Ee-Chien Chang", "Xiaochun Cao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "78dd523ad6123fe1ee5164685a3853719c9a4442adac8661050f147ff6a9f619", "submitted_at": "2025-07-01T15:10:00Z", "updated_at": "2025-07-01T15:10:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.00841v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:01:16.030926+00:00", "role": "primary", "sha256": "4dcd1ba66ba21c0f2cb2aeb00c8bcab7c0cee4da37ff2fa6c16b45066aaa9db0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:01:16.030926+00:00", "role": "primary", "sha256": "4dcd1ba66ba21c0f2cb2aeb00c8bcab7c0cee4da37ff2fa6c16b45066aaa9db0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:10:56.912207+00:00", "role": "primary", "sha256": "a25da6700a238167d5e9915cf6fb3a4bcd910cefc300f4be6eed56271be26f00", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/pdf/2507.00841v1"}], "id": "arxiv:2507.00841", "identifiers": {"arxiv": "2507.00841"}, "kind": "paper", "last_verified_at": "2026-10-18T10:01:16.030926+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.00841"}], "name": "SafeMobile: Chain-level Jailbreak Detection and Automated Evaluation for Multimodal Mobile Agents", "notes": "The reviewed evaluation uses Android emulation and a specific MobileSafetyBench-derived setup. Its judge and score threshold are evaluator choices, not universal safety measurements, and its task population is not the whole later MobileSafetyBench corpus. No results were reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (II.B Defense objective; III Chain-level guard and GPTJudge; IV.B Evaluation setup; V.D Judge validity); not an entire-paper reading or reproduction claim.", "published_at": "2025-07-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies chain-level risk detection for mobile multimodal agents and a language-model-assisted evaluator, assessing safety interventions together with functional preservation.", "topics": ["safety-security", "evaluation", "mobile-navigation"], "updated_at": "2025-07-01", "url": "https://arxiv.org/abs/2507.00841"}
{"aliases": [], "authors": ["Zihao Sun", "Zijing Shi", "Ling Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1745ec9cf10c465c0f389f885df6db502f1884cca271bccea1b45acd2c787af9", "submitted_at": "2025-07-01T16:43:57Z", "updated_at": "2026-09-24T12:58:22Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2507.00938v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.00938", "identifiers": {"arxiv": "2507.00938"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.00938"}], "name": "WebArxiv: A Reproducible Benchmark for Evaluating Multimodal Web Agents on arXiv Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebArxiv provides 510 static-snapshot tasks with deterministic answers for scholarly browsing, including constrained paper retrieval, content extraction, and cross-paper comparison. The study also tests dynamic retrieval of interaction history to address repetitive or incomplete agent reasoning.", "topics": ["evaluation", "web-navigation", "memory"], "updated_at": "2026-09-24", "url": "https://arxiv.org/abs/2507.00938"}
{"aliases": [], "authors": ["Dheeraj Vattikonda", "Santhoshi Ravichandran", "Emiliano Penaloza", "Hadi Nekoei", "Megh Thakkar", "Thibault Le Sellier de Chezelles", "Nicolas Gontier", "Miguel Muñoz-Mármol", "Sahar Omidi Shayegan", "Stefania Raimondo", "Xue Liu", "Alexandre Drouin", "Laurent Charlin", "Alexandre Piché", "Alexandre Lacoste", "Massimo Caccia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "15fa527a869ab66f4f39900ab07ea8b8db930703a8e950ed1e195e1e66164099", "submitted_at": "2025-07-05T17:12:33Z", "updated_at": "2026-02-13T17:24:17Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2507.04103v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.04103", "identifiers": {"arxiv": "2507.04103"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.04103"}], "name": "How to Train Your LLM Web Agent: A Statistical Diagnosis", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies compute allocation for web-agent post-training with teacher-generated supervised fine-tuning followed by on-policy reinforcement learning. It samples 1,370 training configurations and uses bootstrap analysis to examine hyperparameter sensitivity on WorkArena and MiniWoB++.", "topics": ["imitation-learning", "reinforcement-learning", "efficiency", "evaluation"], "updated_at": "2026-02-13", "url": "https://arxiv.org/abs/2507.04103"}
{"aliases": [], "authors": ["Guohong Liu", "Jialei Ye", "Jiacheng Liu", "Yuanchun Li", "Wei Liu", "Pengzhi Gao", "Jian Luan", "Yunxin Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "85ebf8b3b5916e66e4b4221c73bb29519156c41e8d1695bcc46d275dd50cf58d", "submitted_at": "2025-07-06T03:31:36Z", "updated_at": "2026-04-14T14:47:25Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2507.04227v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.04227", "identifiers": {"arxiv": "2507.04227"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.04227"}], "name": "Mobile GUI Agents under Real-world Threats: Are We There Yet?", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-07-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces an app-content instrumentation framework and complementary interactive and static test suites for mobile GUI agents. It studies how untrusted third-party content changes task execution and state understanding.", "topics": ["mobile-navigation", "safety-security", "evaluation"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2507.04227"}
{"aliases": [], "authors": ["Daniel Jones", "Giorgio Severi", "Martin Pouliot", "Gary Lopez", "Joris de Gruyter", "Santiago Zanella-Beguelin", "Justin Song", "Blake Bullwinkel", "Pamela Cortez", "Amanda Minnich"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3ad2f299303076e3474cf247cf25d20ba63b1f1aa8601ab74243c00896bb56e4", "submitted_at": "2025-07-07T19:50:21Z", "updated_at": "2025-07-07T19:50:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.05445v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.05445", "identifiers": {"arxiv": "2507.05445"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.05445"}], "name": "A Systematization of Security Vulnerabilities in Computer Use Agents", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. This is a threat analysis with case studies; the annotation does not imply a standardized public benchmark or independently validated defenses.", "published_at": "2025-07-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Analyzes security boundaries in computer-use agents through a risk taxonomy and concrete adversarial case studies. It links observed failures to input provenance, interface-action binding, memory, and delegation, and proposes CUA-specific evaluation principles.", "topics": ["safety-security", "evaluation"], "updated_at": "2025-07-07", "url": "https://arxiv.org/abs/2507.05445"}
{"aliases": [], "authors": ["Sanidhya Vijayvargiya", "Aditya Bharat Soni", "Xuhui Zhou", "Zora Zhiruo Wang", "Nouha Dziri", "Graham Neubig", "Maarten Sap"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db56d120a7f04f9181a107bcd75c61a3eb3ac36939e2a0517d622332c70f3eac", "submitted_at": "2025-07-08T16:18:54Z", "updated_at": "2026-02-16T22:34:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2507.06134v2"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T07:54:03Z", "role": "primary", "sha256": "13d5969845070cd2877dc51ae5300c2d623e5daef00c8a1552ffa46af3f721a5", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2401.05507%2C2406.07275%2C2409.16427%2C2410.00079%2C2503.07701%2C2503.22738%2C2505.15859%2C2505.17735%2C2505.23596%2C2507.05257%2C2507.06134%2C2507.21953%2C2507.23336%2C2508.06960%2C2508.08997%2C2509.02473%2C2509.23694%2C2510.09116%2C2510.14438%2C2511.18423%2C2512.12818%2C2512.13564%2C2601.10758%2C2601.12449%2C2601.21841%2C2603.01152%2C2603.19993%2C2603.20576%2C2604.04503%2C2604.17562%2C2604.23941%2C2605.16265%2C2606.06448%2C2606.06462%2C2606.29193%2C2607.01647%2C2607.13157&max_results=37"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T07:54:03Z", "role": "primary", "sha256": "13d5969845070cd2877dc51ae5300c2d623e5daef00c8a1552ffa46af3f721a5", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2401.05507%2C2406.07275%2C2409.16427%2C2410.00079%2C2503.07701%2C2503.22738%2C2505.15859%2C2505.17735%2C2505.23596%2C2507.05257%2C2507.06134%2C2507.21953%2C2507.23336%2C2508.06960%2C2508.08997%2C2509.02473%2C2509.23694%2C2510.09116%2C2510.14438%2C2511.18423%2C2512.12818%2C2512.13564%2C2601.10758%2C2601.12449%2C2601.21841%2C2603.01152%2C2603.19993%2C2603.20576%2C2604.04503%2C2604.17562%2C2604.23941%2C2605.16265%2C2606.06448%2C2606.06462%2C2606.29193%2C2607.01647%2C2607.13157&max_results=37"}], "id": "arxiv:2507.06134", "identifiers": {"arxiv": "2507.06134"}, "kind": "paper", "last_verified_at": "2026-10-16T07:54:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.06134"}], "name": "OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2025-07-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates agent safety across real browser, code, filesystem, shell, and messaging tools using rule-based and model-based assessments.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-02-16", "url": "https://arxiv.org/abs/2507.06134"}
{"aliases": [], "authors": ["Ziang Ye", "Yang Zhang", "Wentao Shi", "Xiaoyu You", "Fuli Feng", "Tat-Seng Chua"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "010a3a54ba37417c97e52434211375447418dd2e7ce1da6c339532534f0a920e", "submitted_at": "2025-07-09T14:36:00Z", "updated_at": "2025-09-24T14:33:17Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2507.06899v2"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.06899", "identifiers": {"arxiv": "2507.06899"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.06899"}], "name": "VisualTrap: A Stealthy Backdoor Attack on GUI Agents via Visual Grounding Manipulation", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-07-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies backdoors that redirect GUI visual grounding even when the task plan is correct. VisualTrap evaluates whether poisoned grounding training transfers to downstream tasks and across web, mobile, and desktop interface settings.", "topics": ["safety-security", "grounding", "evaluation"], "updated_at": "2025-09-24", "url": "https://arxiv.org/abs/2507.06899"}
{"aliases": [], "authors": ["Joel Nitu", "Heidrun Mühle", "Andreas Stöckl"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9fe2d9d877a0666d46a065bd8b673c282c879ef0a2620e3b267793a91858b519", "submitted_at": "2025-07-17T07:10:27Z", "updated_at": "2025-11-12T13:34:24Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2507.12844v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.12844", "identifiers": {"arxiv": "2507.12844"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.12844"}], "name": "Machine-Readable Ads: Accessibility and Trust Patterns for AI Web Agents interacting with Online Advertisements", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-07-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares DOM-based and pixel-based web agents on a cloned news site containing advertisements, consent dialogs and paywalls. The study examines how interface presentation affects ad interaction and consequential subscription choices.", "topics": ["evaluation", "ui-design", "safety-security"], "updated_at": "2025-11-12", "url": "https://arxiv.org/abs/2507.12844"}
{"aliases": [], "authors": ["Boyuan Zheng", "Zeyi Liao", "Scott Salisbury", "Zeyuan Liu", "Michael Lin", "Qinyuan Zheng", "Zifan Wang", "Xiang Deng", "Dawn Song", "Huan Sun", "Yu Su"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a8f082fd40174e01f40e97d64f7aa3d90e5f76e831df8f93afdc21aa2e35bfd5", "submitted_at": "2025-07-18T18:06:27Z", "updated_at": "2025-07-18T18:06:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.14293v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.14293", "identifiers": {"arxiv": "2507.14293"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.14293"}], "name": "WebGuard: Building a Generalizable Guardrail for Web Agents", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-07-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebGuard, a human-annotated dataset for predicting outcomes and risk levels of state-changing website actions. It provides generalization splits and evaluates specialized guardrail models.", "topics": ["safety-security", "reward-models", "evaluation"], "updated_at": "2025-07-18", "url": "https://arxiv.org/abs/2507.14293"}
{"aliases": [], "authors": ["Sam Johnson", "Viet Pham", "Thai Le"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "79963ac401384f0c23ecc65d58e616daacc93103456ceb39241dcd6d71c6591c", "submitted_at": "2025-07-20T03:10:13Z", "updated_at": "2025-07-20T03:10:13Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.14799v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.14799", "identifiers": {"arxiv": "2507.14799"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.14799"}], "name": "Manipulating LLM Web Agents with Indirect Prompt Injection Attack via HTML Accessibility Tree", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-07-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies indirect prompt injection through HTML content consumed as accessibility trees. The evaluation tests how manipulated page observations redirect browser-agent actions on real websites.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2025-07-20", "url": "https://arxiv.org/abs/2507.14799"}
{"aliases": [], "authors": ["Xuetian Chen", "Yinghao Chen", "Xinfeng Yuan", "Zhuo Peng", "Lu Chen", "Yuekeng Li", "Zhoujia Zhang", "Yingqian Huang", "Leyan Huang", "Jiaqing Liang", "Tianbao Xie", "Zhiyong Wu", "Qiushi Sun", "Biqing Qi", "Bowen Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fcd5a34bf6187f48da8e08842678e1e21cd558b2bcc7aab2611ceee7bc0d0179", "submitted_at": "2025-07-25T10:14:53Z", "updated_at": "2025-07-25T10:14:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.19132v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.19132", "identifiers": {"arxiv": "2507.19132"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.19132"}], "name": "OS-MAP: How Far Can Computer-Using Agents Go in Breadth and Depth?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "OS-MAP organizes 416 computer-use tasks across 15 applications by a five-level automation taxonomy and a hierarchy of user demands. The benchmark separates required autonomy from generalization scope to diagnose where agents struggle as task demands increase.", "topics": ["evaluation", "automation", "planning"], "updated_at": "2025-07-25", "url": "https://arxiv.org/abs/2507.19132"}
{"aliases": [], "authors": ["Xuehui Wang", "Zhenyu Wu", "JingJing Xie", "Zichen Ding", "Bowen Yang", "Zehao Li", "Zhaoyang Liu", "Qingyun Li", "Xuan Dong", "Zhe Chen", "Weiyun Wang", "Xiangyu Zhao", "Jixuan Chen", "Haodong Duan", "Tianbao Xie", "Chenyu Yang", "Shiqian Su", "Yue Yu", "Yuan Huang", "Yiqian Liu", "Xiao Zhang", "Yanting Zhang", "Xiangyu Yue", "Weijie Su", "Xizhou Zhu", "Wei Shen", "Jifeng Dai", "Wenhai Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8b91f7bacbac917e374aa6f656dcfbcb2e220ab3227fa2cff295c63a1c005f9f", "submitted_at": "2025-07-25T17:59:26Z", "updated_at": "2025-07-25T17:59:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.19478v1"}}, "environments": ["web", "desktop", "windows", "macos", "linux", "ios", "android", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.19478", "identifiers": {"arxiv": "2507.19478"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.19478"}], "name": "MMBench-GUI: Hierarchical Multi-Platform Evaluation Framework for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "MMBench-GUI evaluates interface understanding, element grounding, task automation, and collaboration across six named desktop, mobile, and web platforms. It introduces the Efficiency-Quality Area metric for examining execution efficiency alongside task quality in online GUI automation.", "topics": ["evaluation", "grounding", "efficiency", "planning"], "updated_at": "2025-07-25", "url": "https://arxiv.org/abs/2507.19478"}
{"aliases": [], "authors": ["Qinglong Yang", "Haoming Li", "Haotian Zhao", "Xiaokai Yan", "Jingtao Ding", "Fengli Xu", "Yong Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "947c85efc189635a118b823b8ca0787ff6b1c922d1af8b21a8686b7df9a27cf8", "submitted_at": "2025-06-09T06:38:41Z", "updated_at": "2026-03-16T09:14:01Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2507.21071v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T17:11:36Z", "role": "primary", "sha256": "90a03448678a584f29ef11ae2afe1ef7040b4f3658c492f6e48eafe8fb6d2fa2", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2507.21071", "identifiers": {"arxiv": "2507.21071"}, "kind": "paper", "last_verified_at": "2026-10-17T17:11:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.21071"}], "name": "FingerTip 20K: A Benchmark for Proactive and Personalized Mobile LLM Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "FingerTip 20K gathers 20K human Android demonstrations from users' long-term everyday app use, with user context, to evaluate proactive task suggestion and preference-aware task execution.", "topics": ["evaluation", "data-generation", "mobile-navigation", "interaction-history"], "updated_at": "2026-03-16", "url": "https://arxiv.org/abs/2507.21071"}
{"aliases": [], "authors": ["Zihan Zheng", "Tianle Cui", "Taoran Wang", "Fengtao Wang", "Jiahui Pan", "Lewei He", "Qianglong Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "57d89909534af90cbc10b97a992238bde3ae976ac6cc97778b4345635cef552c", "submitted_at": "2025-08-02T11:53:41Z", "updated_at": "2026-04-20T02:44:32Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2508.01330v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2508.01330", "identifiers": {"arxiv": "2508.01330"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.01330"}], "name": "NaturalGAIA: A Verifiable Benchmark and Hierarchical Framework for Long-Horizon GUI Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "NaturalGAIA builds verifiable long-horizon GUI tasks from human interaction intents while separating causal task structure from its verbal description. Its LightManus-Jarvis framework divides dynamic planning and context management from execution based on visual and structural interface information.", "topics": ["evaluation", "long-horizon", "planning"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2508.01330"}
{"aliases": [], "authors": ["Yuhan Guo", "Cong Guo", "Aiwen Sun", "Hongliang He", "Xinyu Yang", "Yue Lu", "Yingji Zhang", "Xuntao Guo", "Dong Zhang", "Jianzhuang Liu", "Jiang Duan", "Yijia Xiao", "Liangjian Wen", "Hai-Ming Xu", "Yong Dai"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7406b48ed9826d51215d955c31c3c67cd2ce9e08bf9a93a55ac4e85a546315f1", "submitted_at": "2025-08-03T17:17:52Z", "updated_at": "2026-07-05T06:17:25Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2508.01858v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:13+00:00", "role": "primary", "sha256": "1c8ad21928ad5e15bba474a060d225b7da000b37878a8e98dfe5ce94008e3086", "supports": ["scope", "environments"], "url": "https://raw.githubusercontent.com/Gnonymous/Web-CogReasoner/HEAD/README.md"}], "id": "arxiv:2508.01858", "identifiers": {"arxiv": "2508.01858"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.01858"}], "name": "Web-CogReasoner: Towards Multimodal Knowledge-Induced Cognitive Reasoning for Web Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-08-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Web-CogReasoner separates factual and conceptual web knowledge from procedural reasoning, using a structured curriculum to train a multimodal agent. Web-CogDataset and Web-CogBench support study of these capabilities alongside web-navigation evaluation.", "topics": ["data-generation", "planning", "evaluation", "web-navigation"], "updated_at": "2026-07-05", "url": "https://arxiv.org/abs/2508.01858"}
{"aliases": [], "authors": ["Chao Hao", "Shuai Wang", "Kaiwen Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "499a462f9f14b8afdd367bb7cf7a7ffb1b2e166c212fb5c36686d3145f08ab5f", "submitted_at": "2025-08-06T02:38:02Z", "updated_at": "2025-08-06T02:38:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.04025v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.04025", "identifiers": {"arxiv": "2508.04025"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.04025"}], "name": "Uncertainty-Aware GUI Agent: Adaptive Perception through Component Recommendation and Human-in-the-Loop Refinement", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "RecAgent narrows attention to recommended UI components and requests user feedback when task intent or decisions remain uncertain. It introduces ComplexAction to evaluate specified single-step GUI actions in difficult interface scenarios.", "topics": ["mobile-navigation", "human-in-the-loop", "screen-understanding", "evaluation"], "updated_at": "2025-08-06", "url": "https://arxiv.org/abs/2508.04025"}
{"aliases": [], "authors": ["Zeyu Huang", "Juyuan Wang", "Longfeng Chen", "Boyi Xiao", "Leng Cai", "Yawen Zeng", "Jin Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d4efa329c3b106219c9b25f675be0c561743085ac999d2964458a5f22ff6e62f", "submitted_at": "2025-08-12T16:18:30Z", "updated_at": "2025-08-14T16:36:45Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.09057v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2508.09057", "identifiers": {"arxiv": "2508.09057"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.09057"}], "name": "MVISU-Bench: Benchmarking Mobile Agents for Real-World Tasks by Multi-App, Vague, Interactive, Single-App and Unethical Instructions", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-08-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "MVISU-Bench evaluates bilingual mobile tasks involving cross-app work, vague requests, user interaction and unethical instructions alongside ordinary single-app tasks. Its Aider prompting module is designed to clarify intent and mitigate risky execution within agent workflows.", "topics": ["evaluation", "mobile-navigation", "human-in-the-loop", "safety-security"], "updated_at": "2025-08-14", "url": "https://arxiv.org/abs/2508.09057"}
{"aliases": [], "authors": ["Fengxian Ji", "Jingpu Yang", "Zirui Song", "Yuanxi Wang", "Zhexuan Cui", "Yuke Li", "Qian Jiang", "Miao Fang", "Xiuying Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f8c91a1f85602760c706e56426c0d2d870bbc101ca500aa3c7c391938c3f1676", "submitted_at": "2025-08-12T15:12:42Z", "updated_at": "2025-08-12T15:12:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.09241v1"}}, "environments": ["desktop", "web", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.09241", "identifiers": {"arxiv": "2508.09241"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.09241"}], "name": "FineState-Bench: A Comprehensive Benchmark for Fine-Grained State Control in GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "FineState-Bench evaluates fine-grained GUI perception and control across desktop, web, and mobile interfaces. Its Visual Diagnostic Assistant separates perception and localization contributions in controlled interaction experiments.", "topics": ["evaluation", "grounding", "observability", "screen-understanding"], "updated_at": "2025-08-12", "url": "https://arxiv.org/abs/2508.09241"}
{"aliases": [], "authors": ["Ralph Peeters", "Aaron Steiner", "Luca Schwarz", "Julian Yuya Caspary", "Christian Bizer"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0307d242a90f0650b5d5fdfbaf9653842d3ce1461d2653b0b1ba6b21c31ee2dc", "submitted_at": "2025-08-18T15:41:22Z", "updated_at": "2026-04-30T09:17:11Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2508.13024v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.13024", "identifiers": {"arxiv": "2508.13024"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.13024"}], "name": "WebMall -- A Multi-Shop Benchmark for Evaluating Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebMall provides four simulated online shops with heterogeneous product data for comparison shopping, related-product searches, and checkout tasks. It evaluates web agents with different observation formats and memory configurations in a fixed offline environment.", "topics": ["evaluation", "web-navigation", "sandbox"], "updated_at": "2026-04-30", "url": "https://arxiv.org/abs/2508.13024"}
{"aliases": [], "authors": ["Yutong Bian", "Xianhao Lin", "Yupeng Xie", "Tianyang Liu", "Mingchen Zhuge", "Siyuan Lu", "Haoming Tang", "Jinlin Wang", "Jiayi Zhang", "Jiaqi Chen", "Xiangru Tang", "Yongxin Ni", "Sirui Hong", "Chenglin Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c5034c06e63607a1279668964e91c14ed762a95b6e3953337ef8b8b63c4fd57d", "submitted_at": "2025-08-17T07:31:11Z", "updated_at": "2025-08-17T07:31:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.14104v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2508.14104", "identifiers": {"arxiv": "2508.14104"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.14104"}], "name": "You Don't Know Until You Click:Automated GUI Testing for Production-Ready Software Evaluation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "RealDevWorld evaluates generated software repositories with 194 development tasks and AppEvalPilot, an agent judge that operates the resulting GUIs. The evaluation examines functional behavior, visual fidelity, and runtime responses to produce task-specific diagnostic feedback.", "topics": ["evaluation", "automation", "observability"], "updated_at": "2025-08-17", "url": "https://arxiv.org/abs/2508.14104"}
{"aliases": [], "authors": ["Xianren Zhang", "Shreyas Prasad", "Di Wang", "Qiuhai Zeng", "Suhang Wang", "Wenbo Yan", "Mat Hans"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6c8336a707723227a9d651a3e9347e45907f28ed7421b577761ed57f7d70593d", "submitted_at": "2025-08-18T21:58:43Z", "updated_at": "2026-04-21T00:38:54Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.15832v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.15832", "identifiers": {"arxiv": "2508.15832"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.15832"}], "name": "A Functionality-Grounded Benchmark for Evaluating Web Agents in E-commerce Domains", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Amazon-Bench generates functionality-specific web tasks from page content and interactive controls, covering operations beyond product search. Its evaluator checks task completion and unintended changes to account state, such as address or purchase settings.", "topics": ["evaluation", "safety-security", "data-generation", "web-navigation"], "updated_at": "2026-04-21", "url": "https://arxiv.org/abs/2508.15832"}
{"aliases": [], "authors": ["Zhixin Lin", "Jungang Li", "Shidong Pan", "Yibo Shi", "Yue Yao", "Dongliang Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "815c8c1fa5993d2ef58798abd43a2b8945c476c27c777d30c1cbfacadd87af84", "submitted_at": "2025-08-27T00:41:28Z", "updated_at": "2025-09-03T07:48:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.19493v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2508.19493", "identifiers": {"arxiv": "2508.19493"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.19493"}], "name": "Mind the Third Eye! Benchmarking Privacy Awareness in MLLM-powered Smartphone Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SAPA-Bench with 7,138 smartphone scenarios to evaluate whether MLLM-powered mobile agents detect and protect sensitive personal information on screen.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2025-09-03", "url": "https://arxiv.org/abs/2508.19493"}
{"aliases": [], "authors": ["Qihang Ai", "Pi Bu", "Yue Cao", "Yingyao Wang", "Jihao Gu", "Jingxuan Xing", "Zekun Zhu", "Wei Jiang", "Zhicheng Zheng", "Jun Song", "Yuning Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "385ed1b6e60c27660f0cbf281b0f2bbde052691801f7faae99cc053dfbc0d127", "submitted_at": "2025-08-27T08:40:05Z", "updated_at": "2026-04-25T13:29:08Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.19679v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2508.19679", "identifiers": {"arxiv": "2508.19679"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.19679"}], "name": "InquireMobile: Teaching VLM-based Mobile Agent to Request Human Assistance via Reinforcement Fine-Tuning", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "InquireMobile trains mobile agents to seek human confirmation at critical decisions using staged reinforcement fine-tuning and pre-action reasoning. InquireBench evaluates safe interaction and proactive inquiry, distinguishing asking for help from completing the underlying task.", "topics": ["human-in-the-loop", "reinforcement-learning", "evaluation", "mobile-navigation"], "updated_at": "2026-04-25", "url": "https://arxiv.org/abs/2508.19679"}
{"aliases": [], "authors": ["Yi-Hao Peng", "Dingzeyu Li", "Jeffrey P. Bigham", "Amy Pavel"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5dc6411807fd58daed4824e86c39377b24d9fd022ac19daceb836bfb382353cf", "submitted_at": "2025-08-29T09:39:00Z", "updated_at": "2025-08-29T09:39:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.21456v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:56Z", "role": "primary", "sha256": "3429cc2fc7737050cac964c976293fc4793e52c5913636c1f67d04243fbefb6a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.20410%2C2508.20890%2C2508.21456%2C2509.00088%2C2509.00531%2C2509.00969%2C2509.01684%2C2509.03536%2C2509.04018%2C2509.04908%2C2509.05368%2C2509.05831%2C2509.06917%2C2509.07098%2C2509.07553%2C2509.08820%2C2509.10345%2C2509.10540%2C2509.11253%2C2509.12543%2C2509.12987%2C2509.13615%2C2509.14547%2C2509.14744%2C2509.15061%2C2509.15237%2C2509.17336%2C2509.19185%2C2509.19199%2C2509.21072%2C2509.21291%2C2509.21766%2C2509.21799%2C2509.22715%2C2509.22830%2C2509.23738%2C2509.24704%2C2509.25455%2C2509.25693%2C2510.00172%2C2510.00515%2C2510.02190%2C2510.04491%2C2510.05096%2C2510.05156%2C2510.05788%2C2510.07284%2C2510.07915%2C2510.08787%2C2510.09517&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:40Z", "role": "primary", "sha256": "713ecfafe72365a57604306ccaba68f94158fc9b1aa169b408945922cdd1418c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2508.21456v1"}], "id": "arxiv:2508.21456", "identifiers": {"arxiv": "2508.21456"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:40Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.21456"}], "name": "Morae: Proactively Pausing UI Agents for User Choices", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-08-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Morae, a UI agent that detects decision points and pauses for user choices using interface code and screenshots. A study with blind and low-vision participants evaluates this mixed-initiative approach on real web tasks.", "topics": ["human-in-the-loop", "web-navigation", "evaluation"], "updated_at": "2025-08-29", "url": "https://arxiv.org/abs/2508.21456"}
{"aliases": [], "authors": ["Ziyi Guan", "Jason Chun Lok Li", "Zhijian Hou", "Pingping Zhang", "Donglai Xu", "Yuzhi Zhao", "Mengyang Wu", "Jinpeng Chen", "Thanh-Toan Nguyen", "Pengfei Xian", "Wenao Ma", "Shengchao Qin", "Graziano Chesi", "Ngai Wong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c22f63bb5de22dd922ba37ea45df9404def017fee492e8ade03c83ab8b4f9b58", "submitted_at": "2025-08-30T05:32:32Z", "updated_at": "2025-08-30T05:32:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.00366v1"}}, "environments": ["android", "web", "desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.00366", "identifiers": {"arxiv": "2509.00366"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.00366"}], "name": "KG-RAG: Enhancing GUI Agent Decision-Making via Knowledge Graph-Driven Retrieval-Augmented Generation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "KG-RAG converts UI transition graphs into a searchable vector store and uses intent-guided retrieval to construct actionable navigation paths. It introduces KG-Android-Bench and KG-Harmony-Bench and evaluates transfer from mobile apps to web and desktop tasks.", "topics": ["memory", "planning", "mobile-navigation", "evaluation"], "updated_at": "2025-08-30", "url": "https://arxiv.org/abs/2509.00366"}
{"aliases": [], "authors": ["Jaewoo Ahn", "Junseo Kim", "Heeseung Yun", "Jaehyeon Son", "Dongmin Park", "Jaewoong Cho", "Gunhee Kim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3bf86652aaeb252972f3010be38b75b78dfaa9a8063c480c80d0cea47397c2e7", "submitted_at": "2025-09-01T01:33:16Z", "updated_at": "2025-10-15T10:33:27Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.01052v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.01052", "identifiers": {"arxiv": "2509.01052"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.01052"}], "name": "FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-01", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "FlashAdventure evaluates GUI agents on completing story arcs in 34 adventure games. It pairs an automated gameplay judge with COAST, a framework that keeps long-term clue memory to plan and complete sequential objectives.", "topics": ["evaluation", "long-horizon", "memory", "planning"], "updated_at": "2025-10-15", "url": "https://arxiv.org/abs/2509.01052"}
{"aliases": [], "authors": ["Abhinav Kumar", "Jaechul Roh", "Ali Naseh", "Amir Houmansadr", "Eugene Bagdasarian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0f6ff68c59cede9cf59c039416a87ea39b926e3a3179b9f81af9354767befb93", "submitted_at": "2025-09-01T16:56:16Z", "updated_at": "2025-09-01T16:56:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.01619v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.01619", "identifiers": {"arxiv": "2509.01619"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.01619"}], "name": "Throttling Web Agents Using Reasoning Gates", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The contribution is an access-cost mechanism spanning websites and MCP servers, not a GUI navigation policy.", "published_at": "2025-09-01", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces reasoning gates that require agents to solve generated, verifiable puzzles before receiving access to resources. The throttling approach is deployed on a custom website and MCP servers and evaluated with web agents.", "topics": ["safety-security", "evaluation", "tool-integration"], "updated_at": "2025-09-01", "url": "https://arxiv.org/abs/2509.01619"}
{"aliases": [], "authors": ["Gowen Loo", "Chang Liu", "Qinghong Yin", "Xiang Chen", "Jiawei Chen", "Jingyuan Zhang", "Yu Tian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a8b2db00f62afe350e55e2266b3627de9b0cf089178b0cf80eb13b06275eee15", "submitted_at": "2025-09-04T05:22:42Z", "updated_at": "2025-09-04T05:22:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.03891v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2509.03891", "identifiers": {"arxiv": "2509.03891"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.03891"}], "name": "MobileRAG: Enhancing Mobile Agent with Retrieval-Augmented Generation", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileRAG combines retrieval and memory components to support mobile tasks that need external knowledge or long action sequences. MobileRAG-Eval is introduced to study these requirements on complex real-world app tasks.", "topics": ["memory", "mobile-navigation", "evaluation", "long-horizon"], "updated_at": "2025-09-04", "url": "https://arxiv.org/abs/2509.03891"}
{"aliases": [], "authors": ["Hongyi Jing", "Jiafu Chen", "Chen Rao", "Ziqiang Dang", "Jiajie Teng", "Tianyi Chu", "Juncheng Mo", "Shuo Fang", "Huaizhong Lin", "Rui Lv", "Chenguang Ma", "Lei Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f74081d31c98d18e10a5febc5de436a3788fc4156215e4be8fc54b1c53260f58", "submitted_at": "2025-09-05T08:24:12Z", "updated_at": "2025-09-05T08:24:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.04908v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:56Z", "role": "primary", "sha256": "3429cc2fc7737050cac964c976293fc4793e52c5913636c1f67d04243fbefb6a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.20410%2C2508.20890%2C2508.21456%2C2509.00088%2C2509.00531%2C2509.00969%2C2509.01684%2C2509.03536%2C2509.04018%2C2509.04908%2C2509.05368%2C2509.05831%2C2509.06917%2C2509.07098%2C2509.07553%2C2509.08820%2C2509.10345%2C2509.10540%2C2509.11253%2C2509.12543%2C2509.12987%2C2509.13615%2C2509.14547%2C2509.14744%2C2509.15061%2C2509.15237%2C2509.17336%2C2509.19185%2C2509.19199%2C2509.21072%2C2509.21291%2C2509.21766%2C2509.21799%2C2509.22715%2C2509.22830%2C2509.23738%2C2509.24704%2C2509.25455%2C2509.25693%2C2510.00172%2C2510.00515%2C2510.02190%2C2510.04491%2C2510.05096%2C2510.05156%2C2510.05788%2C2510.07284%2C2510.07915%2C2510.08787%2C2510.09517&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:46Z", "role": "primary", "sha256": "e4304967a9a63ead71f00e1933965321540df913f534c0711cc18b9190560cde", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2509.04908v1"}], "id": "arxiv:2509.04908", "identifiers": {"arxiv": "2509.04908"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:46Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.04908"}], "name": "SparkUI-Parser: Enhancing GUI Perception with Robust Grounding and Parsing", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-09-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SparkUI-Parser, which adds continuous coordinate decoding and a rejection mechanism for nonexistent targets to a multimodal model. It also presents ScreenParse for evaluating structural interface perception.", "topics": ["grounding", "screen-understanding", "evaluation"], "updated_at": "2025-09-05", "url": "https://arxiv.org/abs/2509.04908"}
{"aliases": [], "authors": ["Pengxiang Zhao", "Guangyi Liu", "YaoZhen Liang", "Weiqing He", "Zhengxi Lu", "WenHao Wang", "Yuehao Huang", "Yuxiang Chai", "Zhaolu Kang", "Yaxuan Guo", "Hao Wang", "Kexin Zhang", "Liang Liu", "Yong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b1e0e807f5f396bfd9108ab6cfacd48ed4780abfb792589da0b41d7d246c717", "submitted_at": "2025-09-08T09:43:48Z", "updated_at": "2026-04-15T11:26:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.06477v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.06477", "identifiers": {"arxiv": "2509.06477"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.06477"}], "name": "MAS-Bench: A Unified Benchmark for Shortcut-Augmented Hybrid Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAS-Bench evaluates mobile agents that combine GUI actions with APIs, deep links, and reusable scripts. Its 139 tasks across 11 apps test use of 88 predefined shortcuts as well as the ability to generate new reusable workflows.", "topics": ["evaluation", "mobile-navigation", "tool-integration", "workflow"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2509.06477"}
{"aliases": [], "authors": ["Haitao Hu", "Peng Chen", "Yanpeng Zhao", "Yuqi Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ae582aceb31075c460a9e4a6dce9978574c3e0ada731b26c13738a4a4fafcc32", "submitted_at": "2025-09-09T13:59:00Z", "updated_at": "2025-09-09T13:59:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.07764v1"}}, "environments": ["desktop", "linux", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "topics", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:25:42Z", "role": "primary", "sha256": "7f303629ef6d9c7df128c6d12d472abd9bf238e0204515f18839249adef317b9", "supports": ["summary", "environments", "scope", "notes"], "url": "https://arxiv.org/html/2509.07764v1"}], "id": "arxiv:2509.07764", "identifiers": {"arxiv": "2509.07764"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.07764"}], "name": "AgentSentinel: An End-to-End and Real-Time Security Defense Framework for Computer-Use Agents", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The official full paper confirms GUI/desktop interaction alongside Bash and file tools. Its abstract and body enumerate attack categories differently; no category count is promoted into this annotation.", "published_at": "2025-09-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AgentSentinel, which intercepts sensitive computer-agent operations and audits them using task context together with system traces. BadComputerUse evaluates this enforcement approach across attack scenarios involving tool and computer control.", "topics": ["safety-security", "observability", "evaluation"], "updated_at": "2025-09-09", "url": "https://arxiv.org/abs/2509.07764"}
{"aliases": [], "authors": ["Jingyu Tang", "Chaoran Chen", "Jiawen Li", "Zhiping Zhang", "Bingcan Guo", "Ibrahim Khalilov", "Simret Araya Gebreegziabher", "Bingsheng Yao", "Dakuo Wang", "Yanfang Ye", "Tianshi Li", "Ziang Xiao", "Yaxing Yao", "Toby Jia-Jun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "247ed053a4c8f62106d9f8e9af0801d94265ed712f4ae9f095fd8aaccbe33c59", "submitted_at": "2025-09-12T22:26:31Z", "updated_at": "2025-09-12T22:26:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.10723v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.10723", "identifiers": {"arxiv": "2509.10723"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.10723"}], "name": "Dark Patterns Meet GUI Agents: LLM Agent Susceptibility to Manipulative Interfaces and the Role of Human Oversight", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-09-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies how GUI agents, people, and human-agent teams respond to manipulative interface designs. The two-phase evaluation examines failure modes and the benefits and costs of human oversight.", "topics": ["human-in-the-loop", "safety-security", "evaluation"], "updated_at": "2025-09-12", "url": "https://arxiv.org/abs/2509.10723"}
{"aliases": [], "authors": ["Yitong Zhang", "Ximo Li", "Liyi Cai", "Jia Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9ab96edd8e52b3a490455b46878c92c23d15c0f9e105e71dbd6848e058da42ec", "submitted_at": "2025-09-14T12:47:54Z", "updated_at": "2026-08-24T12:04:58Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2509.11250v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.11250", "identifiers": {"arxiv": "2509.11250"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.11250"}], "name": "Environmental Injection Attacks against GUI Agents in Realistic Dynamic Environments", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-09-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Chameleon, an environmental-injection evaluation framework that varies webpage content and trigger context. It combines simulated page variation with attention-based optimization and tests GUI agents across realistic websites.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": "2026-08-24", "url": "https://arxiv.org/abs/2509.11250"}
{"aliases": [], "authors": ["Zhengxi Lu", "Jiabo Ye", "Fei Tang", "Yongliang Shen", "Haiyang Xu", "Ziwei Zheng", "Weiming Lu", "Ming Yan", "Fei Huang", "Jun Xiao", "Yueting Zhuang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1e31638e531aff49fe59bef6e80b35069b1f851ea0cbf1901130557148df4783", "submitted_at": "2025-09-15T03:24:08Z", "updated_at": "2025-09-24T15:05:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.11543v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2509.11543", "identifiers": {"arxiv": "2509.11543"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.11543"}], "name": "UI-S1: Advancing GUI Automation via Semi-online Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-S1 simulates multi-turn online learning over offline trajectories, using a patch module to recover divergences while retaining the model’s own outputs in context. It combines discounted future returns with step- and episode-level advantages and introduces a semi-online performance proxy.", "topics": ["reinforcement-learning", "evaluation", "mobile-navigation"], "updated_at": "2025-09-24", "url": "https://arxiv.org/abs/2509.11543"}
{"aliases": [], "authors": ["Zongru Wu", "Rui Mao", "Zhiyuan Tian", "Pengzhou Cheng", "Tianjie Ju", "Zheng Wu", "Lingzhong Dong", "Haiyue Sheng", "Zhuosheng Zhang", "Gongshen Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "be8620b021bd9218c1143282f1c7a7ea06be2abe6fc0928a310cfc8ac3172a18", "submitted_at": "2025-09-17T01:14:14Z", "updated_at": "2026-03-18T08:12:54Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2509.13615v5"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:56Z", "role": "primary", "sha256": "3429cc2fc7737050cac964c976293fc4793e52c5913636c1f67d04243fbefb6a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.20410%2C2508.20890%2C2508.21456%2C2509.00088%2C2509.00531%2C2509.00969%2C2509.01684%2C2509.03536%2C2509.04018%2C2509.04908%2C2509.05368%2C2509.05831%2C2509.06917%2C2509.07098%2C2509.07553%2C2509.08820%2C2509.10345%2C2509.10540%2C2509.11253%2C2509.12543%2C2509.12987%2C2509.13615%2C2509.14547%2C2509.14744%2C2509.15061%2C2509.15237%2C2509.17336%2C2509.19185%2C2509.19199%2C2509.21072%2C2509.21291%2C2509.21766%2C2509.21799%2C2509.22715%2C2509.22830%2C2509.23738%2C2509.24704%2C2509.25455%2C2509.25693%2C2510.00172%2C2510.00515%2C2510.02190%2C2510.04491%2C2510.05096%2C2510.05156%2C2510.05788%2C2510.07284%2C2510.07915%2C2510.08787%2C2510.09517&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:53Z", "role": "primary", "sha256": "28426488ebca15b74a693ea9fcb2b0ce3165d3baf97b76a4884b34a0d1997af4", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2509.13615v5"}], "id": "arxiv:2509.13615", "identifiers": {"arxiv": "2509.13615"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:53Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.13615"}], "name": "See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-09-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a benchmark for binary GUI-toggle instructions and State-aware Reasoning, which compares the observed control state with the requested state before choosing an action. It evaluates agents on toggle tasks and broader interactive benchmarks.", "topics": ["state-tracking", "grounding", "evaluation"], "updated_at": "2026-03-18", "url": "https://arxiv.org/abs/2509.13615"}
{"aliases": [], "authors": ["Daniel Röder", "Akhil Juneja", "Roland Roller", "Sven Schmeier"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ac93e8cfca06ff86fad06408051abaa6b6510a44d7ad9f2d34c478b11391a32f", "submitted_at": "2025-09-17T19:34:49Z", "updated_at": "2025-09-17T19:34:49Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.14382v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.14382", "identifiers": {"arxiv": "2509.14382"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.14382"}], "name": "Detecting Pipeline Failures through Fine-Grained Analysis of Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a modular framework for identifying errors at intermediate stages of a web-agent pipeline. A SeeAct and Mind2Web case study examines weaknesses that are obscured by final task-success metrics.", "topics": ["evaluation", "observability", "web-navigation"], "updated_at": "2025-09-17", "url": "https://arxiv.org/abs/2509.14382"}
{"aliases": [], "authors": ["Junliang Liu", "Jingyu Xiao", "Wenxin Tang", "Zhixian Wang", "Zipeng Xie", "Wenxuan Wang", "Minrui Zhang", "Shuanghe Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8fed6f081e0330c54a304c79e8a4183865aae48f0e944545029d3de9852f8028", "submitted_at": "2025-09-26T02:38:14Z", "updated_at": "2026-03-04T16:31:39Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.21782v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2509.21782", "identifiers": {"arxiv": "2509.21782"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.21782"}], "name": "Benchmarking MLLM-based Web Understanding: Reasoning, Robustness and Safety", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates multimodal LLMs on web interface reasoning, visual perturbation robustness, and safety-critical web understanding tasks.", "topics": ["evaluation", "screen-understanding", "safety-security", "robustness"], "updated_at": "2026-03-04", "url": "https://arxiv.org/abs/2509.21782"}
{"aliases": [], "authors": ["Gaole Dai", "Shiqi Jiang", "Ting Cao", "Yuqing Yang", "Yuanchun Li", "Rui Tan", "Mo Li", "Lili Qiu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c96f310fb0f28888c59dce71d08697b603e7a7562fdf99b51e3504b93e9e81b3", "submitted_at": "2025-09-26T03:29:36Z", "updated_at": "2026-04-15T09:19:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.21823v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.21823", "identifiers": {"arxiv": "2509.21823"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.21823"}], "name": "ProRe: A Proactive Reward System for GUI Agents via Reasoner-Actor Collaboration", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "ProRe evaluates GUI task outcomes by having a reasoner assign targeted environment-probing tasks to specialist actors. The additional observations support reward judgments, with separate evaluation of judge accuracy and downstream agent task success.", "topics": ["reward-models", "evaluation", "planning"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2509.21823"}
{"aliases": [], "authors": ["Renqi Chen", "Zeyin Tao", "Jianming Guo", "Jingzhe Zhu", "Yiheng Peng", "Qingqing Sun", "Tianyi Zhang", "Shuai Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d23b9f33282f77b9d4965078ca993bcf21163dc1d3210aaa4e554d4fc12b37c5", "submitted_at": "2025-09-26T07:05:01Z", "updated_at": "2026-04-13T03:14:30Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.21982v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.21982", "identifiers": {"arxiv": "2509.21982"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.21982"}], "name": "RISK: A Framework for GUI Agents in E-commerce Risk Management", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "RISK combines browser-interaction training data, a benchmark, and reinforcement fine-tuning for e-commerce risk-management workflows. Its trajectories and rewards distinguish individual actions from multi-step stateful tasks involving deeply embedded web content.", "topics": ["industry", "data-generation", "evaluation", "reinforcement-learning"], "updated_at": "2026-04-13", "url": "https://arxiv.org/abs/2509.21982"}
{"aliases": [], "authors": ["Tao Xiong", "Xavier Hu", "Yurun Chen", "Yuhang Liu", "Changqiao Wu", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5b5375e77f087367383bb5de96cec46d8de6550d9a44bff6965b9201b6718a05", "submitted_at": "2025-09-27T11:42:36Z", "updated_at": "2026-08-30T18:12:14Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2509.23263v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2509.23263", "identifiers": {"arxiv": "2509.23263"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.23263"}], "name": "GUI-PRA: Process Reward Agent for GUI Tasks", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-PRA evaluates GUI progress by generating task-specific criteria and actively inspecting visual evidence with tools. It tests offline judgment on OS-Critic Bench and the effects of this feedback on online mobile policies in AndroidWorld.", "topics": ["reward-models", "evaluation", "screen-understanding"], "updated_at": "2026-08-30", "url": "https://arxiv.org/abs/2509.23263"}
{"aliases": [], "authors": ["Zhecheng Li", "Guoxian Song", "Yiwei Wang", "Zhen Xiong", "Junsong Yuan", "Yujun Cai"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b42e784de8dd007c60d234b7331bd98b5a04c58c13696bf436c81c7eccee37f", "submitted_at": "2025-09-29T00:06:31Z", "updated_at": "2025-09-29T00:06:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.24133v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.24133", "identifiers": {"arxiv": "2509.24133"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.24133"}], "name": "Generalist Scanner Meets Specialist Locator: A Synergistic Coarse-to-Fine Framework for Robust GUI Grounding", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "GMS pairs a general vision-language scanner that proposes interface regions with a specialist locator that predicts coordinates within them. Hierarchical coarse-to-fine search coordinates the two components for GUI grounding on ScreenSpot-Pro.", "topics": ["grounding", "planning", "evaluation"], "updated_at": "2025-09-29", "url": "https://arxiv.org/abs/2509.24133"}
{"aliases": [], "authors": ["Yutong Dai", "Krithika Ramakrishnan", "Jing Gu", "Matthew Fernandez", "Yanqi Luo", "Viraj Prabhu", "Zhenyu Hu", "Silvio Savarese", "Caiming Xiong", "Zeyuan Chen", "Ran Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0e22d26dfbda8530f32d4686f26d91ea1a8e8c5470e857f2e13803258a2f7aaf", "submitted_at": "2025-09-30T16:48:49Z", "updated_at": "2025-09-30T16:48:49Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.26506v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.26506", "identifiers": {"arxiv": "2509.26506"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.26506"}], "name": "SCUBA: Salesforce Computer Use Benchmark", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "SCUBA evaluates enterprise UI navigation and CRM workflows with 300 task instances derived from Salesforce user interviews. Salesforce sandbox execution and milestone-based assessment support comparisons between zero-shot agents and agents given demonstrations.", "topics": ["evaluation", "industry", "workflow", "sandbox"], "updated_at": "2025-09-30", "url": "https://arxiv.org/abs/2509.26506"}
{"aliases": [], "authors": ["Yinuo Liu", "Xilong Wang", "Ruohan Xu", "Yuqi Jia", "Neil Zhenqiang Gong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e997e87268cfe66869bf19fdc8445e5487c980736a63cbe76f779bd1f142c396", "submitted_at": "2025-10-01T18:34:06Z", "updated_at": "2026-09-23T16:44:27Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.01354v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.01354", "identifiers": {"arxiv": "2510.01354"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.01354"}], "name": "WAInjectBench: Benchmarking Prompt Injection Detections for Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The reported contribution evaluates attack detection; end-to-end prevention of browser-agent compromise is not inferred.", "published_at": "2025-10-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WAInjectBench with a threat-model taxonomy and malicious and benign text-image datasets for web-agent prompt-injection detection. It compares detector families across explicit and less visible forms of injected content.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-09-23", "url": "https://arxiv.org/abs/2510.01354"}
{"aliases": [], "authors": ["Erfan Shayegani", "Keegan Hines", "Yue Dong", "Nael Abu-Ghazaleh", "Roman Lutz", "Spencer Whitehead", "Vidhisha Balachandran", "Besmira Nushi", "Vibhav Vineet"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a6426fa5eeb3dfafef30e47dd46a6a43dff3e24bf0ff485b7d10ab9efe133f94", "submitted_at": "2025-10-02T04:52:15Z", "updated_at": "2025-10-02T04:52:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.01670v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.01670", "identifiers": {"arxiv": "2510.01670"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.01670"}], "name": "Just Do It!? Computer-Use Agents Exhibit Blind Goal-Directedness", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "BLIND-ACT tests whether computer-use agents persist with ambiguous, contradictory, or infeasible goals without adequate contextual reasoning. Its OSWorld-based tasks and judged trajectories examine these behaviors and the effect of mitigation prompts.", "topics": ["evaluation", "safety-security", "robustness"], "updated_at": "2025-10-02", "url": "https://arxiv.org/abs/2510.01670"}
{"aliases": [], "authors": ["Gonzalo Gonzalez-Pumariega", "Vincent Tu", "Chih-Lun Lee", "Jiachen Yang", "Ang Li", "Xin Eric Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "29c00ba4713da26f912fec8046d498b4fd10cf71a3109f66e92fa23f7cebcf44", "submitted_at": "2025-10-02T17:37:08Z", "updated_at": "2026-02-03T23:48:45Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.02250v2"}}, "environments": ["desktop", "windows", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.02250", "identifiers": {"arxiv": "2510.02250"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.02250"}], "name": "Scaling Agents for Computer Use", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Behavior Judge, which converts execution trajectories into behavior narratives and compares multiple candidate rollouts. Experiments examine this selection approach on OSWorld, WindowsAgentArena and AndroidWorld.", "topics": ["evaluation", "reward-models", "long-horizon"], "updated_at": "2026-02-03", "url": "https://arxiv.org/abs/2510.02250"}
{"aliases": [], "authors": ["Sagnik Anupam", "Davis Brown", "Shuo Li", "Eric Wong", "Hamed Hassani", "Osbert Bastani"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7906e06c2efb8778a43ec774821f90947e70cec0b78d6d43ac8901f7f0562332", "submitted_at": "2025-10-02T15:22:21Z", "updated_at": "2025-10-07T15:12:39Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.02418v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2510.02418", "identifiers": {"arxiv": "2510.02418"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.02418"}], "name": "BrowserArena: Evaluating LLM Agents on Real-World Web Navigation Tasks", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces BrowserArena, a live-web platform for head-to-head agent comparisons on user-submitted tasks. Step-level human annotations support targeted analysis of navigation, pop-up handling and CAPTCHA-related failures.", "topics": ["evaluation", "web-navigation", "human-in-the-loop"], "updated_at": "2025-10-07", "url": "https://arxiv.org/abs/2510.02418"}
{"aliases": [], "authors": ["Su Kara", "Fazle Faisal", "Suman Nath"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "88e25f6d876dc6ad11219e3faa0e7d95f4abb39a9aba32213896225aff64ec42", "submitted_at": "2025-09-28T20:51:05Z", "updated_at": "2025-09-28T20:51:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.03285v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.03285", "identifiers": {"arxiv": "2510.03285"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.03285"}], "name": "WAREX: Web Agent Reliability Evaluation on Existing Benchmarks", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "WAREX augments existing web-agent benchmarks with reliability challenges arising from unstable systems and changing or disruptive website behavior. Evaluation on WebArena, WebVoyager and REAL examines how these disturbances affect task completion.", "topics": ["evaluation", "robustness", "web-navigation"], "updated_at": "2025-09-28", "url": "https://arxiv.org/abs/2510.03285"}
{"aliases": [], "authors": ["Sanjari Srivastava", "Gang Li", "Cheng Chang", "Rishu Garg", "Manpreet Kaur", "Charlene Y. Lee", "Yuezhang Li", "Yining Mao", "Ignacio Cases", "Yanan Xie", "Peng Qi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3af3fb75d012ebcf4ad082ab7b83b0553596bbbb8598751d0a70c86d929a22ee", "submitted_at": "2025-10-10T21:20:51Z", "updated_at": "2026-05-18T22:06:26Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.09872v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2510.09872", "identifiers": {"arxiv": "2510.09872"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.09872"}], "name": "WARC-Bench: Web Archive Based Benchmark for GUI Subtask Executions", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "WARC-Bench packages dynamic web pages as replayable archives and evaluates short navigation subtasks involving components such as date pickers and scrollable containers. It also studies supervised training followed by reinforcement learning with verifiable rewards on these browser interactions.", "topics": ["evaluation", "web-navigation", "sandbox", "reinforcement-learning"], "updated_at": "2026-05-18", "url": "https://arxiv.org/abs/2510.09872"}
{"aliases": [], "authors": ["Zonghao Ying", "Yangguang Shao", "Jianle Gan", "Gan Xu", "Wenxin Zhang", "Quanchen Zou", "Junzheng Shi", "Zhenfei Yin", "Mingchuan Zhang", "Aishan Liu", "Xianglong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "43c9f7c91539df587a4f08888820731aef5bb3fca2cfb9e742443330aaf1c35f", "submitted_at": "2025-10-11T07:18:12Z", "updated_at": "2026-04-14T02:53:37Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.10073v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.10073", "identifiers": {"arxiv": "2510.10073"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.10073"}], "name": "SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SecureWebArena with simulated interactive websites, an attack-vector taxonomy and labeled agent trajectories. Its evaluation separates failures in reasoning, behavioral traces and task outcomes across vision-language agent families.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2510.10073"}
{"aliases": [], "authors": ["Longjie Guo", "Chenjie Yuan", "Mingyuan Zhong", "Robert Wolfe", "Ruican Zhong", "Yue Xu", "Bingbing Wen", "Hua Shen", "Lucy Lu Wang", "Alexis Hiniker"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e6152a61f81c435e76a432bc10086827c0cf0d2f44026bac03d6b80ebae8ca77", "submitted_at": "2025-10-13T06:12:03Z", "updated_at": "2026-02-23T08:31:13Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.11035v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.11035", "identifiers": {"arxiv": "2510.11035"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.11035"}], "name": "SusBench: An Online Benchmark for Evaluating Dark Pattern Susceptibility of Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SusBench, an online benchmark that inserts dark-pattern designs into consumer websites for agent evaluation. A human study checks the realism of the interventions, and agent trials compare susceptibility across pattern types.", "topics": ["ui-design", "safety-security", "evaluation"], "updated_at": "2026-02-23", "url": "https://arxiv.org/abs/2510.11035"}
{"aliases": [], "authors": ["Kai Mei", "Jiang Guo", "Shuaichen Chang", "Mingwen Dong", "Dongkyu Lee", "Xing Niu", "Jiarong Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7c37dae372655e8ef556163744e47ed42db3533175ca7d1db7560033774ed2a0", "submitted_at": "2025-10-13T19:52:04Z", "updated_at": "2026-03-09T15:25:16Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.11892v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.11892", "identifiers": {"arxiv": "2510.11892"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.11892"}], "name": "R-WoM: Retrieval-augmented World Model For Computer-use Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "R-WoM evaluates how well world models predict nearby GUI states, align complete procedures, and recognize milestone transitions. It retrieves external tutorials to guide simulated future interactions, with experiments on subsets of OSWorld and WebArena.", "topics": ["world-models", "memory", "planning", "evaluation"], "updated_at": "2026-03-09", "url": "https://arxiv.org/abs/2510.11892"}
{"aliases": [], "authors": ["Xiaoxue Ren", "Penghao Jiang", "Kaixin Li", "Zhiyong Huang", "Xiaoning Du", "Jiaojiao Jiang", "Zhenchang Xing", "Jiamou Sun", "Terry Yue Zhuo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e92db1e568c3c512aa00e2363bbdc9869cf834d638c331b45d4753d29a282004", "submitted_at": "2025-10-14T06:52:15Z", "updated_at": "2025-10-14T06:52:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.12200v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.12200", "identifiers": {"arxiv": "2510.12200"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.12200"}], "name": "HackWorld: Evaluating Computer-Use Agents on Exploiting Web Application Vulnerabilities", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces HackWorld for evaluating visual computer-use agents in a capture-the-flag collection of web applications. The benchmark examines interface navigation and multistep planning during vulnerability discovery and exploitation tasks.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2025-10-14", "url": "https://arxiv.org/abs/2510.12200"}
{"aliases": [], "authors": ["Yuanyi Song", "Heyuan Huang", "Qiqiang Lin", "Yin Zhao", "Xiangmou Qu", "Jun Wang", "Xingyu Lou", "Weiwen Liu", "Zhuosheng Zhang", "Jun Wang", "Yong Yu", "Weinan Zhang", "Zhaoxiang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fc45dc62068c971cd831fc3db64bd3d8995d4e5fd60e9c83953476d4fb24f7f3", "submitted_at": "2025-10-16T12:30:05Z", "updated_at": "2025-10-16T12:30:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.14621v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2510.14621", "identifiers": {"arxiv": "2510.14621"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.14621"}], "name": "ColorBench: Benchmarking Mobile Agents with Graph-Structured Framework for Complex Long-Horizon Tasks", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "ColorBench represents mobile tasks as graphs of states collected from real devices, providing a quasi-dynamic environment with multiple valid paths and error paths. It evaluates single-app and cross-app tasks using completion, subtask progress, and atomic diagnostics.", "topics": ["evaluation", "mobile-navigation", "long-horizon", "state-tracking"], "updated_at": "2025-10-16", "url": "https://arxiv.org/abs/2510.14621"}
{"aliases": [], "authors": ["Devin Ersoy", "Brandon Lee", "Ananth Shreekumar", "Arjun Arunasalam", "Muhammad Ibrahim", "Antonio Bianchi", "Z. Berkay Celik"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db3c25ba55221ca3d7c45ce4243a4a37b56a698909c0ea3eaaabdcd37388a1ec", "submitted_at": "2025-10-20T21:26:26Z", "updated_at": "2025-10-20T21:26:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.18113v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.18113", "identifiers": {"arxiv": "2510.18113"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.18113"}], "name": "Investigating the Impact of Dark Patterns on LLM-Based Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents LiteAgent for recording web-agent interactions and TrickyArena for enabling or disabling dark patterns in controlled websites. Experiments compare how individual patterns, combined patterns and visual or HTML changes affect agent decisions.", "topics": ["safety-security", "ui-design", "evaluation"], "updated_at": "2025-10-20", "url": "https://arxiv.org/abs/2510.18113"}
{"aliases": [], "authors": ["Ho Fai Leung", "Xiaoyan Xi", "Fei Zuo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "33b21a63d51148f2b5c30478c302e3df9b92e03e2e106a1c68224f2c835a8608", "submitted_at": "2025-10-21T10:11:33Z", "updated_at": "2025-10-21T10:11:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.18488v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.18488", "identifiers": {"arxiv": "2510.18488"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.18488"}], "name": "AndroidControl-Curated: Revealing the True Potential of GUI Agents through Benchmark Purification", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "AndroidControl-Curated audits ambiguities and factual errors in AndroidControl and produces a refined evaluation dataset. The paper also studies post-training a small Magma-R1 model on a curated subset to examine the effect of data quality on GUI action assessment.", "topics": ["evaluation", "data-generation", "grounding"], "updated_at": "2025-10-21", "url": "https://arxiv.org/abs/2510.18488"}
{"aliases": [], "authors": ["Yaoyao Qian", "Yuanli Wang", "Jinda Zhang", "Yun Zong", "Meixu Chen", "Hanhan Zhou", "Jindan Huang", "Yifan Zeng", "Xinyu Hu", "Chan Hee Song", "Danqing Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d404ff922ef38e57481c01a5e3108bd38457d88f83d7adf6d3c3c7b5ddf124e2", "submitted_at": "2025-10-22T03:29:25Z", "updated_at": "2025-10-22T03:29:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.19205v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.19205", "identifiers": {"arxiv": "2510.19205"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.19205"}], "name": "WebGraphEval: Multi-Turn Trajectory Evaluation for Web Agents using Graph Representation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebGraphEval, which merges multiple agents' trajectories into a weighted graph of canonical actions. Structural analyses examine recurring behavior, inefficient paths and important decision points beyond a final success score.", "topics": ["evaluation", "observability", "interaction-history"], "updated_at": "2025-10-22", "url": "https://arxiv.org/abs/2510.19205"}
{"aliases": [], "authors": ["Qiushi Sun", "Mukai Li", "Zhoumianze Liu", "Zhihui Xie", "Fangzhi Xu", "Zhangyue Yin", "Kanzhi Cheng", "Zehao Li", "Zichen Ding", "Qi Liu", "Zhiyong Wu", "Zhuosheng Zhang", "Ben Kao", "Lingpeng Kong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7e3fde732c9e760437f2c60d585f7ba0d2ed13bbe9e051838844a21174159099", "submitted_at": "2025-10-28T13:22:39Z", "updated_at": "2026-07-26T01:16:06Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2510.24411v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.24411", "identifiers": {"arxiv": "2510.24411"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.24411"}], "name": "OS-Sentinel: Towards Safety-Enhanced Mobile GUI Agents via Hybrid Validation in Realistic Workflows", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-10-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MobileRisk-Live and OS-Sentinel, combining a formal verifier for system-level violations with a multimodal judge for contextual risks in mobile GUI workflows.", "topics": ["safety-security", "evaluation", "mobile-navigation"], "updated_at": "2026-07-26", "url": "https://arxiv.org/abs/2510.24411"}
{"aliases": [], "authors": ["Hongrui Jia", "Jitong Liao", "Xi Zhang", "Haiyang Xu", "Tianbao Xie", "Chaoya Jiang", "Ming Yan", "Si Liu", "Wei Ye", "Fei Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d115356f723746ce95c3e40fed933786a0747b3b560c8f73a6b7af05ca22d24f", "submitted_at": "2025-10-28T15:56:36Z", "updated_at": "2025-11-11T04:10:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.24563v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.24563", "identifiers": {"arxiv": "2510.24563"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.24563"}], "name": "OSWorld-MCP: Benchmarking MCP Tool Invocation In Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "OSWorld-MCP evaluates how agents choose between GUI operations and MCP tool calls in computer tasks. It combines generated and curated tools, with 158 manually validated tools across seven applications, to examine tool use alongside task completion.", "topics": ["evaluation", "tool-integration", "planning"], "updated_at": "2025-11-11", "url": "https://arxiv.org/abs/2510.24563"}
{"aliases": [], "authors": ["Chenghao Du", "Quanfeng Huang", "Tingxuan Tang", "Zihao Wang", "Adwait Nadkarni", "Yue Xiao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "61b611712fff54af25aae660f2353fa6c3aaa764b0767831ebf2278c120d9dea", "submitted_at": "2025-10-31T03:35:59Z", "updated_at": "2026-07-07T02:18:26Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2510.27140v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2510.27140", "identifiers": {"arxiv": "2510.27140"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.27140"}], "name": "Measuring the Security of Mobile LLM Agents under Adversarial Prompts from Untrusted Third-Party Channels", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Measures how untrusted third-party notifications, messages, and app content can hijack mobile LLM agents into executing unauthorized cross-app actions.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2026-07-07", "url": "https://arxiv.org/abs/2510.27140"}
{"aliases": [], "authors": ["Shaojie Zhang", "Pei Fu", "Ruoceng Zhang", "Jiahui Yang", "Anan Du", "Xiuwen Xi", "Shaokang Wang", "Ying Huang", "Bin Qin", "Zhenbo Luo", "Jian Luan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6869e86d7cd4242ba8dcf942bb1cef3c157dff4f0ef37f7138ff5025d082c7b8", "submitted_at": "2025-10-31T08:07:02Z", "updated_at": "2026-05-27T02:32:58Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.27266v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.27266", "identifiers": {"arxiv": "2510.27266"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.27266"}], "name": "Enhancing Trustworthy GUI Grounding via Self-Critiqued Reinforcement Learning", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "HyperClick trains GUI grounders to produce a click and an accompanying confidence estimate using rewards for correctness and confidence alignment. Its evaluation examines localization, calibration, and abstention behavior.", "topics": ["grounding", "reinforcement-learning", "robustness", "evaluation"], "updated_at": "2026-05-27", "url": "https://arxiv.org/abs/2510.27266"}
{"aliases": [], "authors": ["Jian Mu", "Chaoyun Zhang", "Chiming Ni", "Lu Wang", "Bo Qiao", "Kartik Mathur", "Qianhui Wu", "Yuhang Xie", "Xiaojun Ma", "Mengyu Zhou", "Si Qin", "Liqun Li", "Yu Kang", "Minghua Ma", "Qingwei Lin", "Saravan Rajmohan", "Dongmei Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6b063a4d2b3a552aa05a153745978473a7d165f2c6b4b97e4c91eff11b7b0f80", "submitted_at": "2025-11-06T12:19:02Z", "updated_at": "2025-11-10T12:27:15Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2511.04307v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.04307", "identifiers": {"arxiv": "2511.04307"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.04307"}], "name": "GUI-360$^\\circ$: A Comprehensive Dataset and Benchmark for Computer-Using Agents", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2025-11-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Collects Windows application trajectories with screenshots and action traces for grounding, screen parsing, and action prediction.", "topics": ["data-generation", "grounding", "evaluation"], "updated_at": "2025-11-10", "url": "https://arxiv.org/abs/2511.04307"}
{"aliases": [], "authors": ["Xiangwu Guo", "Difei Gao", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8a5f43873a47e5b47250cedd3d860abe2dc5ec9b85fe657444bf2b7378f3e8f7", "submitted_at": "2025-11-09T15:13:45Z", "updated_at": "2025-11-09T15:13:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.06417v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.06417", "identifiers": {"arxiv": "2511.06417"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.06417"}], "name": "AUTO-Explorer: Automated Data Collection for GUI Agent", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-11-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Auto-Explorer, an automated GUI exploration method that records software states for training-data collection. UIXplore evaluates exploration coverage, and collected data is used to fine-tune and test GUI element grounding.", "topics": ["data-generation", "grounding", "evaluation"], "updated_at": "2025-11-09", "url": "https://arxiv.org/abs/2511.06417"}
{"aliases": [], "authors": ["Aarash Feizi", "Shravan Nayak", "Xiangru Jian", "Kevin Qinghong Lin", "Kaixin Li", "Rabiul Awal", "Xing Han Lù", "Johan Obando-Ceron", "Juan A. Rodriguez", "Nicolas Chapados", "David Vazquez", "Adriana Romero-Soriano", "Reihaneh Rabbany", "Perouz Taslakian", "Christopher Pal", "Spandana Gella", "Sai Rajeswar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0f2b80f294ddf5e3331dcb47a836864032f88f2fdbe35588bfef000cd9ceddc8", "submitted_at": "2025-11-10T17:35:21Z", "updated_at": "2026-06-09T21:30:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2511.07332v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.07332", "identifiers": {"arxiv": "2511.07332"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.07332"}], "name": "Grounding Computer Use Agents on Human Demonstrations", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "GroundCUA turns expert desktop demonstrations into densely annotated screenshots and grounding instructions. The GroundNext model family uses this supervision and subsequent reinforcement learning, with both grounding benchmarks and an OSWorld agent configuration used for evaluation.", "topics": ["grounding", "data-generation", "evaluation", "reinforcement-learning"], "updated_at": "2026-06-09", "url": "https://arxiv.org/abs/2511.07332"}
{"aliases": [], "authors": ["Leyang Yang", "Ziwei Wang", "Xiaoxuan Tang", "Sheng Zhou", "Dajun Chen", "Wei Jiang", "Yong Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dc7f5f0fe1ca373272cda4bbeca3a465e1a7eaec829e8a59bbecbe6049dc7b4a", "submitted_at": "2025-11-12T09:49:31Z", "updated_at": "2025-11-12T09:49:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.09157v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.09157", "identifiers": {"arxiv": "2511.09157"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.09157"}], "name": "ProBench: Benchmarking GUI Agents with Accurate Process Information", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "ProBench extends mobile GUI evaluation with tasks whose correctness depends on the interaction process as well as the final state. Its ProcessProvider supplies process information for assessing agent behavior and analyzing failure modes.", "topics": ["evaluation", "mobile-navigation", "observability"], "updated_at": "2025-11-12", "url": "https://arxiv.org/abs/2511.09157"}
{"aliases": [], "authors": ["Kevin Qinghong Lin", "Siyuan Hu", "Linjie Li", "Zhengyuan Yang", "Lijuan Wang", "Philip Torr", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7d63831bc0ff00979615415ce94abc140b646a4375cb33b950c4a7ccf00bf166", "submitted_at": "2025-11-19T16:00:02Z", "updated_at": "2025-11-19T16:00:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.15567v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.15567", "identifiers": {"arxiv": "2511.15567"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.15567"}], "name": "Computer-Use Agents as Judges for Generative User Interface", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "AUI-Gym evaluates generated websites by having computer-use agents attempt tasks whose executability is checked programmatically. A feedback dashboard condenses navigation histories into visual summaries that a coding agent can use to revise an interface.", "topics": ["evaluation", "ui-design", "observability"], "updated_at": "2025-11-19", "url": "https://arxiv.org/abs/2511.15567"}
{"aliases": [], "authors": ["Sen Chen", "Tong Zhao", "Yi Bin", "Fei Ma", "Wenqi Shao", "Zheng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1679f7b56c7b60f92195e488c887114bcdcce7975a80fafbf2572c2b1f01ce0b", "submitted_at": "2025-11-20T17:43:46Z", "updated_at": "2025-11-20T17:43:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.16590v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.16590", "identifiers": {"arxiv": "2511.16590"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.16590"}], "name": "D-GARA: A Dynamic Benchmarking Framework for GUI Agent Robustness in Real-World Anomalies", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "D-GARA evaluates Android GUI agents under runtime anomalies such as permission dialogs, battery warnings, and update prompts. Its modular environment embeds these interruptions in Android applications to study recovery and robustness.", "topics": ["evaluation", "robustness", "mobile-navigation"], "updated_at": "2025-11-20", "url": "https://arxiv.org/abs/2511.16590"}
{"aliases": [], "authors": ["Horia Cristescu", "Charles Park", "Trong Canh Nguyen", "Sergiu Talmacel", "Alexandru-Gabriel Ilie", "Stefan Adam"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4c746b329229d90dbe3d04a138f54c8b205b16ec6173cf681d47a319c8f24ddf", "submitted_at": "2025-11-21T10:47:22Z", "updated_at": "2025-11-21T10:47:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.17131v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.17131", "identifiers": {"arxiv": "2511.17131"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.17131"}], "name": "UI-CUBE: Enterprise-Grade Computer Use Agent Benchmarking Beyond Task Accuracy to Operational Reliability", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-11-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-CUBE evaluates 226 tasks ranging from individual controls to copy-paste and enterprise workflows. Interface variations, multiple resolutions, and application-state validation support diagnosis of operational reliability beyond task accuracy alone.", "topics": ["evaluation", "industry", "workflow", "robustness"], "updated_at": "2025-11-21", "url": "https://arxiv.org/abs/2511.17131"}
{"aliases": [], "authors": ["Marta Sumyk", "Oleksandr Kosovan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ff9343bb706687caf2ea813220ee7221819cd5892ba950f3168b91aebe0a74e0", "submitted_at": "2025-11-25T08:40:33Z", "updated_at": "2025-11-25T08:40:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.20067v1"}}, "environments": ["desktop", "macos"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.20067", "identifiers": {"arxiv": "2511.20067"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.20067"}], "name": "\"Are We Done Yet?\": A Vision-Based Judge for Autonomous Task Completion of Computer Use Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study tests vision-language models as task-completion judges using screenshots and instructions from 42 built-in macOS applications. It evaluates judgment accuracy on human-labeled tasks and separately studies how judge feedback affects agent task success.", "topics": ["evaluation", "self-reflection"], "updated_at": "2025-11-25", "url": "https://arxiv.org/abs/2511.20067"}
{"aliases": [], "authors": ["Karen Ullrich", "Jingtong Su", "Claudia Shi", "Arjun Subramonian", "Amir Bar", "Ivan Evtimov", "Nikolaos Tsilivis", "Randall Balestriero", "Julia Kempe", "Mark Ibrahim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2c288e7c42b5eedeea3f27a5622e87981f101f472db05ead6d6bede1371500f0", "submitted_at": "2025-11-25T19:00:22Z", "updated_at": "2025-11-25T19:00:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.20766v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2511.20766", "identifiers": {"arxiv": "2511.20766"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.20766"}], "name": "OpenApps: Simulating Environment Variations to Measure UI-Agent Reliability", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-11-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OpenApps, a configurable app ecosystem for measuring UI-agent reliability under changes in appearance and content. Its evaluation compares task success and failure behavior across many versions of the same apps.", "topics": ["evaluation", "robustness", "data-generation"], "updated_at": "2025-11-25", "url": "https://arxiv.org/abs/2511.20766"}
{"aliases": [], "authors": ["Ruihan Chen", "Qiming Li", "Xiaocheng Feng", "Weihong Zhong", "Xiaoliang Yang", "Yuxuan Gu", "Zekun Zhou", "Yunfei Lu", "Haoyu Ren", "Kun Chen", "Dandan Tu", "Bing Qin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf16987178fc24fb483bfa82efecb1924c7528fd16ac94f1144fcc345748850f", "submitted_at": "2025-11-30T06:47:33Z", "updated_at": "2026-04-28T03:32:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2512.00756v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.00756", "identifiers": {"arxiv": "2512.00756"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.00756"}], "name": "MPR-GUI: Benchmarking and Enhancing Multilingual Perception and Reasoning in GUI Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "MPR-GUI evaluates GUI perception and reasoning through strictly aligned tasks in six languages. GUI-XLI adjusts non-English hidden representations toward English representations at language-sensitive layers during inference to address the measured cross-language gaps.", "topics": ["evaluation", "screen-understanding", "robustness"], "updated_at": "2026-04-28", "url": "https://arxiv.org/abs/2512.00756"}
{"aliases": [], "authors": ["Hyunjun Kim", "Sooyoung Ryu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7e445d0ca79f727a6c1d51e83381c3fd1d8e1fddeff7de99fe70ca06bc1bff49", "submitted_at": "2025-12-01T01:18:21Z", "updated_at": "2025-12-01T01:18:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.01174v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}], "id": "arxiv:2512.01174", "identifiers": {"arxiv": "2512.01174"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:55Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.01174"}], "name": "DrawingBench: Evaluating Spatial Reasoning and UI Interaction Capabilities of Large Language Models through Mouse-Based Drawing Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "DrawingBench evaluates sequences of low-level mouse-based GUI actions through 250 drawing prompts and rule-based criteria. Action inspection and multi-turn external feedback support analysis of spatial reasoning, tool-state management, and refinement.", "topics": ["evaluation", "grounding", "human-in-the-loop", "state-tracking"], "updated_at": "2025-12-01", "url": "https://arxiv.org/abs/2512.01174"}
{"aliases": [], "authors": ["Aiden Yiliu Li", "Bizhi Yu", "Daoan Lei", "Tianhe Ren", "Shilong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4868c3c85545532e9c208623596d3810c85e32e3cf11429a68847ac11c2ba1cb", "submitted_at": "2025-12-01T18:37:19Z", "updated_at": "2025-12-01T18:37:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.01979v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.01979", "identifiers": {"arxiv": "2512.01979"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.01979"}], "name": "Chain-of-Ground: Improving GUI Grounding via Iterative Reasoning and Reference Feedback", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Chain-of-Ground uses iterative visual reasoning and reference feedback to refine GUI localization without additional training. It evaluates on ScreenSpot-Pro and introduces TPanel-UI to probe grounding on industrial control panels with visual distortions.", "topics": ["grounding", "training-free", "robustness", "evaluation"], "updated_at": "2025-12-01", "url": "https://arxiv.org/abs/2512.01979"}
{"aliases": [], "authors": ["Haolong Yan", "Yeqing Shen", "Xin Huang", "Jia Wang", "Kaijun Tan", "Zhixuan Liang", "Hongxin Li", "Zheng Ge", "Osamu Yoshie", "Si Li", "Xiangyu Zhang", "Daxin Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fda844ac6767e6d9dfd16919042469e2070ec7f092e0b01fb5a243ed6eb4c282", "submitted_at": "2025-12-02T05:11:23Z", "updated_at": "2025-12-02T05:11:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.02423v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}], "id": "arxiv:2512.02423", "identifiers": {"arxiv": "2512.02423"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:55Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.02423"}], "name": "GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-12-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUI Exploration Lab, a simulation engine for composing screens, interface elements, and navigation graphs. It studies how supervised learning, single-turn RL, and multiturn RL affect GUI navigation and exploration.", "topics": ["reinforcement-learning", "evaluation", "data-generation"], "updated_at": "2025-12-02", "url": "https://arxiv.org/abs/2512.02423"}
{"aliases": [], "authors": ["Jinzhe Tan", "Karim Benyekhlef"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d362a839fd1ee917ebdde6fcd794640523e930cd711769154532200903fe7de5", "submitted_at": "2025-11-28T07:26:29Z", "updated_at": "2025-11-28T07:26:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.04105v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.04105", "identifiers": {"arxiv": "2512.04105"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.04105"}], "name": "LegalWebAgent: Empowering Access to Justice via LLM-Based Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-11-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "LegalWebAgent separates user-needs clarification, webpage browsing, and actions such as form completion or scheduling. A 15-task evaluation studies legal-service workflows relevant to Québec civil-law users, using HTML and screenshot observations.", "topics": ["industry", "web-navigation", "human-in-the-loop", "evaluation"], "updated_at": "2025-11-28", "url": "https://arxiv.org/abs/2512.04105"}
{"aliases": [], "authors": ["Zhiyuan Jiang", "Shenghao Xie", "Wenyi Li", "Wenqiang Zu", "Peihang Li", "Jiahao Qiu", "Siqi Pei", "Lei Ma", "Tiejun Huang", "Mengdi Wang", "Shilong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bac971bd0e7cd946f81e28d5976eb933cbb95b97ca846ee624a0c36b48760de7", "submitted_at": "2025-12-05T18:39:12Z", "updated_at": "2025-12-05T18:39:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.05941v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.05941", "identifiers": {"arxiv": "2512.05941"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.05941"}], "name": "Zoom in, Click out: Unlocking and Evaluating the Potential of Zooming for GUI Grounding", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "ZoomClick uses repeated zooming and context adjustment to improve GUI grounding without training. The paper studies zoom depth, crop sizes, and related choices and introduces GUIZoom-Bench to evaluate how grounding models respond to zoomed views.", "topics": ["grounding", "training-free", "evaluation", "screen-understanding"], "updated_at": "2025-12-05", "url": "https://arxiv.org/abs/2512.05941"}
{"aliases": [], "authors": ["Alisha Ukani", "Hamed Haddadi", "Ali Shahin Shamsabadi", "Peter Snyder"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2576701bbdc4fd8f1f571ed691b06185d1e16bf4955a5c99b10200d796ba6525", "submitted_at": "2025-12-08T17:16:12Z", "updated_at": "2025-12-08T17:16:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.07725v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.07725", "identifiers": {"arxiv": "2512.07725"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.07725"}], "name": "Privacy Practices of Browser Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The abstract describes artifact release as planned; the annotation does not independently verify availability.", "published_at": "2025-12-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents a privacy evaluation framework for browser agents covering component weaknesses, website defenses, tracking, privacy-related prompts and personal-data leakage. The framework is applied to deployed browser-agent tools.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2025-12-08", "url": "https://arxiv.org/abs/2512.07725"}
{"aliases": [], "authors": ["Chunyi Li", "Longfei Li", "Zicheng Zhang", "Xiaohong Liu", "Min Tang", "Weisi Lin", "Guangtao Zhai"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "484a51656db4466d5e8630d64fa78d3cf10043568488f5a2e9bae679fa800cbe", "submitted_at": "2025-12-12T14:49:32Z", "updated_at": "2025-12-12T14:49:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.11611v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.11611", "identifiers": {"arxiv": "2512.11611"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.11611"}], "name": "Using GUI Agent for Electronic Design Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-EDA collects more than 2,000 screenshot-answer-action pairs from engineers working across five CAD tools and five physical domains. The accompanying benchmark evaluates more than 30 GUI agents on specialized electronic-design workflows.", "topics": ["industry", "data-generation", "evaluation", "workflow"], "updated_at": "2025-12-12", "url": "https://arxiv.org/abs/2512.11611"}
{"aliases": [], "authors": ["Luoxi Meng", "Henry Feng", "Ilia Shumailov", "Earlence Fernandes"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fa6b4d9fa34610ec6fa8c0ac8b075127365aad57a20d814a2e838a223cf8bd56", "submitted_at": "2025-12-14T08:25:31Z", "updated_at": "2026-03-04T01:49:57Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2512.12594v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:09Z", "role": "primary", "sha256": "f8cee8b281f54ecb3112bb26b59d166617d33f08f098f44818968a9c9b777ce6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2512.06866%2C2512.07287%2C2512.08769%2C2512.12594%2C2512.12730%2C2512.13159%2C2512.14012%2C2512.16925%2C2512.16962%2C2512.16975%2C2512.21373%2C2512.23647%2C2512.23738%2C2601.01743%2C2601.02046%2C2601.02163%2C2601.02454%2C2601.02702%2C2601.03468%2C2601.03928%2C2601.04795%2C2601.05107%2C2601.05432%2C2601.06411%2C2601.06487%2C2601.06490%2C2601.06565%2C2601.06663%2C2601.07143%2C2601.07375%2C2601.08012%2C2601.08079%2C2601.08225%2C2601.08237%2C2601.08731%2C2601.09113%2C2601.09749%2C2601.09883%2C2601.10378%2C2601.10440%2C2601.11653%2C2601.11895%2C2601.11903%2C2601.11974%2C2601.12042%2C2601.12307%2C2601.12560%2C2601.12641%2C2601.13186%2C2601.13383&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:09Z", "role": "primary", "sha256": "f8cee8b281f54ecb3112bb26b59d166617d33f08f098f44818968a9c9b777ce6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2512.06866%2C2512.07287%2C2512.08769%2C2512.12594%2C2512.12730%2C2512.13159%2C2512.14012%2C2512.16925%2C2512.16962%2C2512.16975%2C2512.21373%2C2512.23647%2C2512.23738%2C2601.01743%2C2601.02046%2C2601.02163%2C2601.02454%2C2601.02702%2C2601.03468%2C2601.03928%2C2601.04795%2C2601.05107%2C2601.05432%2C2601.06411%2C2601.06487%2C2601.06490%2C2601.06565%2C2601.06663%2C2601.07143%2C2601.07375%2C2601.08012%2C2601.08079%2C2601.08225%2C2601.08237%2C2601.08731%2C2601.09113%2C2601.09749%2C2601.09883%2C2601.10378%2C2601.10440%2C2601.11653%2C2601.11895%2C2601.11903%2C2601.11974%2C2601.12042%2C2601.12307%2C2601.12560%2C2601.12641%2C2601.13186%2C2601.13383&max_results=50"}], "id": "arxiv:2512.12594", "identifiers": {"arxiv": "2512.12594"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:09Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.12594"}], "name": "ceLLMate: Sandboxing Browser AI Agents", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2025-12-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "ceLLMate constrains browser-agent authority through an extension that applies policies at the HTTP layer. It evaluates this sandbox against prompt-injection attacks in WASP, avoiding dependence on brittle low-level click and keystroke policies.", "topics": ["safety-security", "sandbox", "web-navigation", "robustness", "evaluation"], "updated_at": "2026-03-04", "url": "https://arxiv.org/abs/2512.12594"}
{"aliases": [], "authors": ["Youngmin Im", "Byeongung Jo", "Jaeyoung Wi", "Seungwoo Baek", "Tae Hoon Min", "Joo Hyung Lee", "Sangeun Oh", "Insik Shin", "Sunjae Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5982901598e43dcadf63b4a950b2fff68e85be71d0f483648b535d82d7745bd7", "submitted_at": "2025-12-14T10:41:39Z", "updated_at": "2026-05-31T23:48:10Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2512.12634v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.12634", "identifiers": {"arxiv": "2512.12634"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.12634"}], "name": "MobiBench: Multi-Branch, Modular Benchmark for Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobiBench evaluates mobile GUI agents offline while allowing multiple valid action paths. It decomposes agent pipelines into modules to study component contributions, configurations, and performance bottlenecks.", "topics": ["evaluation", "mobile-navigation", "observability"], "updated_at": "2026-05-31", "url": "https://arxiv.org/abs/2512.12634"}
{"aliases": [], "authors": ["Shufan Li", "Konstantinos Kallidromitis", "Akash Gokul", "Yusuke Kato", "Kazuki Kozuka", "Aditya Grover"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e6228822cf64c6aeb1836bee52a6515123bd0d44d77cceb23e785bb4ff9aef47", "submitted_at": "2025-12-16T02:16:42Z", "updated_at": "2026-10-07T23:06:49Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2512.14014v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.14014", "identifiers": {"arxiv": "2512.14014"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.14014"}], "name": "MobileWorldBench: Towards Semantic World Modeling For Mobile Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates semantic world models for mobile interfaces and studies natural-language state transitions as a tool for agent planning.", "topics": ["world-models", "planning", "evaluation"], "updated_at": "2026-10-07", "url": "https://arxiv.org/abs/2512.14014"}
{"aliases": [], "authors": ["Haolong Yan", "Jia Wang", "Xin Huang", "Yeqing Shen", "Ziyang Meng", "Zhimin Fan", "Kaijun Tan", "Jin Gao", "Lieyu Shi", "Mi Yang", "Shiliang Yang", "Zhirui Wang", "Brian Li", "Kang An", "Chenyang Li", "Lei Lei", "Mengmeng Duan", "Danxun Liang", "Guodong Liu", "Hang Cheng", "Hao Wu", "Jie Dong", "Junhao Huang", "Mei Chen", "Renjie Yu", "Shunshan Li", "Xu Zhou", "Yiting Dai", "Yineng Deng", "Yingdan Liang", "Zelin Chen", "Wen Sun", "Chengxu Yan", "Chunqin Xu", "Dong Li", "Fengqiong Xiao", "Guanghao Fan", "Guopeng Li", "Guozhen Peng", "Hongbing Li", "Hang Li", "Hongming Chen", "Jingjing Xie", "Jianyong Li", "Jingyang Zhang", "Jiaju Ren", "Jiayu Yuan", "Jianpeng Yin", "Kai Cao", "Liang Zhao", "Liguo Tan", "Liying Shi", "Mengqiang Ren", "Min Xu", "Manjiao Liu", "Mao Luo", "Mingxin Wan", "Na Wang", "Nan Wu", "Ning Wang", "Peiyao Ma", "Qingzhou Zhang", "Qiao Wang", "Qinlin Zeng", "Qiong Gao", "Qiongyao Li", "Shangwu Zhong", "Shuli Gao", "Shaofan Liu", "Shisi Gao", "Shuang Luo", "Xingbin Liu", "Xiaojia Liu", "Xiaojie Hou", "Xin Liu", "Xuanti Feng", "Xuedan Cai", "Xuan Wen", "Xianwei Zhu", "Xin Liang", "Xin Liu", "Xin Zhou", "Yifan Sui", "Yingxiu Zhao", "Yukang Shi", "Yunfang Xu", "Yuqing Zeng", "Yixun Zhang", "Zejia Weng", "Zhonghao Yan", "Zhiguo Huang", "Zhuoyu Wang", "Zihan Yan", "Zheng Ge", "Jing Li", "Yibo Zhu", "Binxing Jiao", "Xiangyu Zhang", "Daxin Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2fc824bdeabca3451fd895447a0321c507da1ac7bc7cc24184c7768992549daf", "submitted_at": "2025-12-17T13:26:30Z", "updated_at": "2025-12-19T17:36:21Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2512.15431v2"}}, "environments": ["android", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.15431", "identifiers": {"arxiv": "2512.15431"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.15431"}], "name": "Step-GUI Technical Report", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Step-GUI uses calibrated trajectory rewards to turn generated GUI interaction data into training signals for compact agent models. The report also introduces GUI-MCP for hierarchical GUI operations and local task delegation, and AndroidDaily for static-action and end-to-end mobile evaluation.", "topics": ["data-generation", "reward-models", "tool-integration", "evaluation", "mobile-navigation"], "updated_at": "2025-12-19", "url": "https://arxiv.org/abs/2512.15431"}
{"aliases": [], "authors": ["Zhenyu Wu", "Jingjing Xie", "Zehao Li", "Bowen Yang", "Qiushi Sun", "Zhaoyang Liu", "Zhoumianze Liu", "Yu Qiao", "Xiangyu Yue", "Zun Wang", "Zichen Ding"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "67a1bd40b117b34e5fb994b1c2c93cd285714097d1bc9730baeeaba6c31a7b8a", "submitted_at": "2025-12-18T08:29:50Z", "updated_at": "2025-12-18T08:29:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.16295v1"}}, "environments": ["web", "desktop", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.16295", "identifiers": {"arxiv": "2512.16295"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.16295"}], "name": "OS-Oracle: A Comprehensive Framework for Cross-Platform GUI Critic Models", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "OS-Oracle synthesizes GUI critic examples and trains a pre-action critic with supervised learning followed by consistency-preserving GRPO. It introduces OS-Critic Bench across mobile, web, and desktop settings and tests critic-assisted agents in OSWorld and AndroidWorld.", "topics": ["reward-models", "reinforcement-learning", "data-generation", "evaluation"], "updated_at": "2025-12-18", "url": "https://arxiv.org/abs/2512.16295"}
{"aliases": [], "authors": ["Beitong Zhou", "Zhexiao Huang", "Yuan Guo", "Zhangxuan Gu", "Tianyu Xia", "Zichen Luo", "Fei Tang", "Dehan Kong", "Yanyi Shang", "Suling Ou", "Zhenlin Guo", "Changhua Meng", "Shuheng Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4bc8a1da3993ead9b36af47a9332fe9a1a2427101ad0d91da654b4a4b25141df", "submitted_at": "2025-12-18T13:09:09Z", "updated_at": "2025-12-18T13:09:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.16501v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.16501", "identifiers": {"arxiv": "2512.16501"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.16501"}], "name": "VenusBench-GD: A Comprehensive Multi-Platform GUI Benchmark for Diverse Grounding Tasks", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "VenusBench-GD is a bilingual, cross-platform GUI grounding benchmark organized into basic and advanced tasks. Its annotation pipeline and six-subtask taxonomy support comparisons of general multimodal models and GUI-specialized models across different grounding demands.", "topics": ["grounding", "evaluation", "data-generation", "robustness"], "updated_at": "2025-12-18", "url": "https://arxiv.org/abs/2512.16501"}
{"aliases": [], "authors": ["Brandon Huang", "Hang Hua", "Zhuoran Yu", "Trevor Darrell", "Rogerio Feris", "Roei Herzig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "37cd0aa99ed8bad167e25f6e2b48e7cd1cf01249508555f657215213ee94a301", "submitted_at": "2025-12-19T04:09:24Z", "updated_at": "2025-12-31T17:30:11Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2512.17221v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.17221", "identifiers": {"arxiv": "2512.17221"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.17221"}], "name": "DAVE: A VLM Vision Encoder for Document Understanding and Web Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "DAVE trains a vision encoder for document images and web agents using self-supervised pretraining followed by supervised parsing and localization. It combines decoder-specific encoder merging with feature fusion and evaluates web localization and agent benchmarks alongside document tasks.", "topics": ["screen-understanding", "grounding", "evaluation"], "updated_at": "2025-12-31", "url": "https://arxiv.org/abs/2512.17221"}
{"aliases": [], "authors": ["Quyu Kong", "Xu Zhang", "Zhenyu Yang", "Nolan Gao", "Chen Liu", "Panrong Tong", "Chenglin Cai", "Hanzhang Zhou", "Jianan Zhang", "Liangyu Chen", "Zhidan Liu", "Steven Hoi", "Yue Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cf3258fd2286b0ee6babb69b9bfc459ecf40536eec498a3af3403d2971791f3d", "submitted_at": "2025-12-22T14:31:28Z", "updated_at": "2025-12-30T07:15:15Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2512.19432v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.19432", "identifiers": {"arxiv": "2512.19432"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.19432"}], "name": "MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive and MCP-Augmented Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces long mobile workflows that combine agent-user interaction, cross-application tasks, and MCP-based tool access.", "topics": ["evaluation", "tool-integration", "long-horizon"], "updated_at": "2025-12-30", "url": "https://arxiv.org/abs/2512.19432"}
{"aliases": [], "authors": ["Yue Cao", "Yingyao Wang", "Pi Bu", "Jingxuan Xing", "Wei Jiang", "Zekun Zhu", "Junpeng Ma", "Sashuai Zhou", "Tong Lu", "Jun Song", "Yu Cheng", "Yuning Jiang", "Bo Zheng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "597935923ce9e59720771f3d92e921f2842f908ad81b3ab31a87ed9d9f5c5774", "submitted_at": "2025-12-24T17:40:42Z", "updated_at": "2025-12-24T17:40:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.21302v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.21302", "identifiers": {"arxiv": "2512.21302"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.21302"}], "name": "AndroidLens: Long-latency Evaluation with Nested Sub-targets for Android GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "AndroidLens provides 571 extended mobile tasks in Chinese and English, with multiple valid paths and preserved environmental anomalies. Its dynamic evaluation uses nested milestones and Average Task Progress to measure partial completion of long workflows.", "topics": ["evaluation", "long-horizon", "mobile-navigation", "robustness"], "updated_at": "2025-12-24", "url": "https://arxiv.org/abs/2512.21302"}
{"aliases": [], "authors": ["Karolina Korgul", "Yushi Yang", "Arkadiusz Drohomirecki", "Piotr Błaszczyk", "Will Howard", "Lukas Aichberger", "Chris Russell", "Philip H. S. Torr", "Adam Mahdi", "Adel Bibi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a255cccbe3f16f498918126724c391911e1d745c093189790cbab715fd28a2d2", "submitted_at": "2025-12-29T01:09:10Z", "updated_at": "2026-07-02T16:12:04Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2512.23128v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.23128", "identifiers": {"arxiv": "2512.23128"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.23128"}], "name": "It's a TRAP! Task-Redirecting Agent Persuasion Benchmark for Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-12-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces TRAP, which evaluates task-redirection attacks embedded in realistic web interfaces. Controlled website clones and a modular injection framework support comparisons of persuasion techniques and contextual changes across agents.", "topics": ["safety-security", "evaluation", "ui-design"], "updated_at": "2026-07-02", "url": "https://arxiv.org/abs/2512.23128"}
{"aliases": ["ShowUI-pi", "ShowUI Pi"], "authors": ["Siyuan Hu", "Kevin Qinghong Lin", "Mike Zheng Shou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b7d1f358b450f2a6c3deb00f1b16c54d443f48fda3a7dc70c77456bebf35ad7c", "submitted_at": "2025-12-31T16:51:14Z", "updated_at": "2025-12-31T16:51:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.24965v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.24965", "identifiers": {"arxiv": "2512.24965"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.24965"}], "name": "ShowUI-$π$: Flow-based Generative Models as GUI Dexterous Hands", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Models discrete clicks and continuous drags with flow-based action generation, and introduces drag trajectories and the ScreenDrag benchmark.", "topics": ["grounding", "evaluation", "data-generation"], "updated_at": "2025-12-31", "url": "https://arxiv.org/abs/2512.24965"}
{"aliases": [], "authors": ["Jiaxin Ai", "Yukang Feng", "Fanrui Zhang", "Jianwen Sun", "Zizhen Li", "Chuanhao Li", "Yifan Chang", "Wenxiao Wu", "Ruoxi Wang", "Mingliang Zhai", "Kaipeng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "afab3dcd2444bf8333f2a2fea887819f6fca55c41209d8c449f6211d1c62b819", "submitted_at": "2025-12-30T01:49:46Z", "updated_at": "2025-12-30T01:49:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.02399v1"}}, "environments": ["windows", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "topics"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:06:36Z", "role": "primary", "sha256": "f9a44f8090369a296a584a9b94c4491662ee25bb8f7239ef75dfd8f775be8887", "supports": ["summary", "environments", "scope"], "url": "https://arxiv.org/html/2601.02399v1"}], "id": "arxiv:2601.02399", "identifiers": {"arxiv": "2601.02399"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.02399"}], "name": "ProSoftArena: Benchmarking Hierarchical Capabilities of Multimodal Agents in Professional Software Environments", "notes": "The paper abstract and official full-text sections were reviewed. The full text supports Windows 11 virtual-machine execution, screenshot/state observations, and mouse/keyboard actions. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "ProSoftArena evaluates 436 professional-software tasks across six disciplines and 13 applications in a Windows 11 virtual-machine environment. Its capability hierarchy, execution-based checks, and human-in-the-loop evaluation cover precise GUI operations and multi-software workflows.", "topics": ["evaluation", "industry", "science", "cross-app"], "updated_at": "2025-12-30", "url": "https://arxiv.org/abs/2601.02399"}
{"aliases": [], "authors": ["Hao Bai", "Alexey Taymanov", "Tong Zhang", "Aviral Kumar", "Spencer Whitehead"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ae8751c4fae3b4f838eba59ffa8be37a1a5f4fa497dfd69fa3075f8581547f51", "submitted_at": "2026-01-05T09:35:11Z", "updated_at": "2026-05-02T18:45:09Z", "version": 6, "versioned_url": "https://arxiv.org/abs/2601.02439v6"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.02439", "identifiers": {"arxiv": "2601.02439"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.02439"}], "name": "WebGym: Scaling Training Environments for Visual Web Agents with Realistic Tasks", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebGym supplies realistic web tasks with rubric-based evaluation and an asynchronous rollout system for reinforcement learning. It studies how training on an agent’s own interaction traces scales with task coverage and transfers to unseen websites.", "topics": ["web-navigation", "reinforcement-learning", "data-generation", "evaluation", "efficiency"], "updated_at": "2026-05-02", "url": "https://arxiv.org/abs/2601.02439"}
{"aliases": [], "authors": ["Ziyun Zhang", "Zezhou Wang", "Xiaoyi Zhang", "Zongyu Guo", "Jiahao Li", "Bin Li", "Yan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d4d0a59f1eb8197b4d389f5f40957fb8aeb1e22d0739f9877b26d57ed70fe59b", "submitted_at": "2026-01-07T17:40:08Z", "updated_at": "2026-06-30T13:54:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.04126v3"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.04126", "identifiers": {"arxiv": "2601.04126"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.04126"}], "name": "InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "InfiniteWeb generates functional, interconnected websites and task evaluators using shared specifications and task-driven development. It evaluates website construction and the transfer of GUI policies trained in these synthetic web environments to OSWorld and Online-Mind2Web.", "topics": ["data-generation", "reinforcement-learning", "evaluation", "sandbox"], "updated_at": "2026-06-30", "url": "https://arxiv.org/abs/2601.04126"}
{"aliases": [], "authors": ["Yifei Gao", "Jiang Wu", "Xiaoyi Chen", "Yifan Yang", "Zhe Cui", "Tianyi Ma", "Jiaming Zhang", "Jitao Sang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "47a340b19cf399f2ae7851482030d85e572b0f6ad8415c67657beec737a80e06", "submitted_at": "2026-01-08T02:07:53Z", "updated_at": "2026-01-08T02:07:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.04500v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.04500", "identifiers": {"arxiv": "2601.04500"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.04500"}], "name": "GUITester: Enabling GUI Agents for Exploratory Defect Discovery", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUITestBench for interactive defect discovery and GUITester, which separates task navigation from anomaly verification. Testing-oriented plans and history-based reflection help distinguish software defects from agent execution mistakes.", "topics": ["automation", "evaluation", "self-reflection"], "updated_at": "2026-01-08", "url": "https://arxiv.org/abs/2601.04500"}
{"aliases": [], "authors": ["Zezhou Wang", "Ziyun Zhang", "Xiaoyi Zhang", "Zhuzhong Qian", "Yan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e342723078b50e612df749df80eadfd16e0fe29f5875777c1b3189d8b5981512", "submitted_at": "2026-01-09T13:26:38Z", "updated_at": "2026-02-10T15:12:17Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.05787v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.05787", "identifiers": {"arxiv": "2601.05787"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.05787"}], "name": "From Off-Policy to On-Policy: Enhancing GUI Agents via Bi-level Expert-to-Policy Assimilation", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "BEPA adapts a small set of expert GUI traces into policy-aligned guidance through trajectories reachable by the learner and a dynamically updated task cache. It evaluates this two-level approach to reinforcement learning on desktop and web benchmarks.", "topics": ["reinforcement-learning", "data-generation", "evaluation"], "updated_at": "2026-02-10", "url": "https://arxiv.org/abs/2601.05787"}
{"aliases": [], "authors": ["Xinyi Wu", "Jiagui Chen", "Geng Hong", "Jiayi Dong", "Xudong Pan", "Jiarun Dai", "Min Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "626702d22786336ff65b6e89d97e2034ff7aa144fab4ecc286d169fca6993410", "submitted_at": "2026-01-13T10:21:28Z", "updated_at": "2026-01-13T10:21:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.08406v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.08406", "identifiers": {"arxiv": "2601.08406"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.08406"}], "name": "WebTrap Park: An Automated Platform for Systematic Security Evaluation of Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebTrap Park, a platform that turns web-agent security risks into executable tasks and assesses concrete interactions with live pages. It compares agent frameworks without requiring modifications to their implementations.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2026-01-13", "url": "https://arxiv.org/abs/2601.08406"}
{"aliases": [], "authors": ["Wenyuan Zhang", "Xinghua Zhang", "Haiyang Yu", "Shuaiyi Nie", "Bingli Wu", "Juwei Yue", "Tingwen Liu", "Yongbin Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73df8c5b417c3387530b7730be3aed8b69374c06fdd3573f56147b19359c6912", "submitted_at": "2026-01-13T14:48:34Z", "updated_at": "2026-04-15T14:59:17Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.08605v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.08605", "identifiers": {"arxiv": "2601.08605"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.08605"}], "name": "ExpSeek: Self-Triggered Experience Seeking for Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "ExpSeek uses step-level entropy signals to decide when a web agent should request experience-based guidance. It supplies guidance tailored to the current step and evaluates the intervention across web-agent benchmarks.", "topics": ["memory", "planning", "web-navigation", "evaluation"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2601.08605"}
{"aliases": [], "authors": ["Yibo Lyu", "Gongwei Chen", "Rui Shao", "Weili Guan", "Liqiang Nie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d8b697b91942291403491581a61c135c6bfc8b0f3c2a9b9bab0f926af7208fce", "submitted_at": "2026-01-14T17:12:48Z", "updated_at": "2026-05-13T02:54:19Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.09636v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.09636", "identifiers": {"arxiv": "2601.09636"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.09636"}], "name": "PersonalAlign: Hierarchical Implicit Intent Alignment for Personalized GUI Agent with Long-Term User-Centric Records", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AndroidIntent to evaluate how GUI agents infer omitted preferences and recurring routines from long-term user records. Its HIM-Agent maintains hierarchical memory of preferences and routines for personalized execution and proactive suggestions.", "topics": ["memory", "evaluation", "mobile-navigation"], "updated_at": "2026-05-13", "url": "https://arxiv.org/abs/2601.09636"}
{"aliases": [], "authors": ["Yi Qian", "Kunwei Qian", "Xingbang He", "Ligeng Chen", "Jikang Zhang", "Tiantai Zhang", "Haiyang Wei", "Linzhang Wang", "Hao Wu", "Bing Mao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "21848dea9e000125550fff7fe3844a57bf100215bf9eb800cbdbb612675f7074", "submitted_at": "2026-01-18T10:54:54Z", "updated_at": "2026-07-15T15:32:11Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.12349v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.12349", "identifiers": {"arxiv": "2601.12349"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.12349"}], "name": "Mind the Gap: Action Rebinding Attacks against Android GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies action-rebinding attacks in which an Android GUI agent’s planned input is redirected to a different application during the observation–action delay. The evaluation examines cross-application execution and the consequences of treating an earlier screen as the current action target.", "topics": ["safety-security", "cross-app", "evaluation"], "updated_at": "2026-07-15", "url": "https://arxiv.org/abs/2601.12349"}
{"aliases": [], "authors": ["Zecheng Li", "Zhihui Cao", "Wenke Huang", "Yudong Zhang", "Keying Qi", "Rui Wang", "Zeyu Zheng", "Jian Zhao", "Hao Zhu", "Hengxin Wu", "Yuran Wang", "Guitao Fan", "Guokun Wu", "Yicong Liu", "Zhilin Gao", "Haikun Xu", "He Yang", "Minqi Xiang", "Xingyu Liu", "Zuojian Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aa8cb4137e862bf8031a55bc8f81e1d6c903733f2cfac9cf758e5f2f92884481", "submitted_at": "2026-01-19T13:50:43Z", "updated_at": "2026-01-19T13:50:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.13060v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.13060", "identifiers": {"arxiv": "2601.13060"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.13060"}], "name": "MagicGUI-RMS: A Multi-Agent Reward Model System for Self-Evolving GUI Agents via Automated Feedback Reflux", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "MagicGUI-RMS combines domain-specific and general reward models to assess GUI actions and propose corrections. A structured data pipeline and feedback-recycling process support repeated reward-model and agent improvement.", "topics": ["reward-models", "data-generation", "evaluation", "robustness"], "updated_at": "2026-01-19", "url": "https://arxiv.org/abs/2601.13060"}
{"aliases": [], "authors": ["Simret Araya Gebreegziabher", "Yukun Yang", "Charles Chiang", "Hojun Yoo", "Chaoran Chen", "Hyo Jin Do", "Zahra Ashktorab", "Werner Geyer", "Diego Gómez-Zará", "Toby Jia-Jun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1bdc29648446cd79e0c0b9f72814457b55258d501283ce2fa2dedcd8262f885b", "submitted_at": "2026-01-22T22:42:40Z", "updated_at": "2026-01-22T22:42:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.16356v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.16356", "identifiers": {"arxiv": "2601.16356"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.16356"}], "name": "The Behavioral Fabric of LLM-Powered GUI Agents: Human Values and Interaction Outcomes", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a controlled web-task testbed for studying how stated preferences and human values influence GUI-agent decisions. It analyzes how personas alter navigation and how prominent interface cues can override the intended preference guidance.", "topics": ["evaluation", "ui-design", "human-in-the-loop"], "updated_at": "2026-01-22", "url": "https://arxiv.org/abs/2601.16356"}
{"aliases": [], "authors": ["Ying Mo", "Yu Bai", "Dapeng Sun", "Yuqian Shi", "Yukai Miao", "Li Chen", "Dan Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fa7c8ce1b0e584c5279691ff06cd522d7c8ea1f4801d7074324c6845ce482efb", "submitted_at": "2026-01-25T06:58:15Z", "updated_at": "2026-01-25T06:58:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.17722v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.17722", "identifiers": {"arxiv": "2601.17722"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.17722"}], "name": "EntWorld: A Holistic Environment and Benchmark for Verifiable Enterprise GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces EntWorld, an enterprise GUI benchmark built by deriving task logic from database schemas. Long workflows across enterprise domains are assessed with deterministic SQL-based state checks rather than screenshot similarity alone.", "topics": ["evaluation", "industry", "workflow"], "updated_at": "2026-01-25", "url": "https://arxiv.org/abs/2601.17722"}
{"aliases": [], "authors": ["Nan Chen", "Jing Lu", "Zilong Wang", "Luna K. Qiu", "Siming Chen", "Yuqing Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ae1716f673e47cd0e61b5f2225ca4871dfd8522e35037fc8d226f9c7cd0e3e75", "submitted_at": "2026-01-26T03:02:26Z", "updated_at": "2026-01-26T03:02:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.18092v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.18092", "identifiers": {"arxiv": "2601.18092"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.18092"}], "name": "From Struggle to Success: Context-Aware Guidance for Screen Reader Users in Computer Use", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-26", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces AskEase, an on-demand assistant that combines contextual information to provide screen-reader-friendly computer-use guidance. A study with screen-reader users evaluates task completion and perceived workload while preserving human control of the interaction.", "topics": ["human-in-the-loop", "evaluation", "workflow"], "updated_at": "2026-01-26", "url": "https://arxiv.org/abs/2601.18092"}
{"aliases": [], "authors": ["Shaokang Wang", "Pei Fu", "Ruoceng Zhang", "Shaojie Zhang", "Xiuwen Xi", "Jiahui Yang", "Bin Qin", "Ying Huang", "Zhenbo Luo", "Jian Luan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ccb46b86b0c0b854ed2f1e17f9e12bf031c5ed1349f3d3c44d3276b2a3a195ee", "submitted_at": "2026-01-26T06:29:41Z", "updated_at": "2026-06-26T07:19:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.18197v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}], "id": "arxiv:2601.18197", "identifiers": {"arxiv": "2601.18197"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:55Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.18197"}], "name": "GAIA: A Data Flywheel System for Training GUI Test-Time Scaling Critic Models", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "GAIA trains a GUI action critic on positive and negative examples, then uses that critic to collect improved examples for another training round. It studies how recycling these data affects the test-time action choices of different GUI agents.", "topics": ["reward-models", "data-generation", "evaluation"], "updated_at": "2026-06-26", "url": "https://arxiv.org/abs/2601.18197"}
{"aliases": [], "authors": ["Xuan Wang", "Siyuan Su", "Quantong Fu", "Yongxiang Hu", "Yangfan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cb90eed33e989d38cd4058d54ba7d073f1b8e5c1453a9106dfe5e533da7b08a1", "submitted_at": "2026-01-26T09:35:10Z", "updated_at": "2026-08-20T07:50:44Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.18305v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.18305", "identifiers": {"arxiv": "2601.18305"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.18305"}], "name": "SwipeGen: Bridging the Execution Gap in GUI Agents via Human-like Swipe Synthesis", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SwipeGen for synthesizing varied swipe trajectories and SwipeBench for assessing their execution quality. The study examines how richer control of trajectory, speed, and timing changes GUI agents’ interaction performance.", "topics": ["action-prediction", "data-generation", "evaluation"], "updated_at": "2026-08-20", "url": "https://arxiv.org/abs/2601.18305"}
{"aliases": [], "authors": ["Yanxi Wang", "Zhiling Zhang", "Wenbo Zhou", "Weiming Zhang", "Jie Zhang", "Qiannan Zhu", "Yu Shi", "Shuxin Zheng", "Jiyan He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1ca912ad67a681e8e908a5a403e982f39b5080bee4df376eb0b598468f3bb504", "submitted_at": "2026-01-26T11:33:40Z", "updated_at": "2026-05-13T07:11:36Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.18842v3"}}, "environments": ["android", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.18842", "identifiers": {"arxiv": "2601.18842"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.18842"}], "name": "GUIGuard-Bench: Toward a General Evaluation for Privacy-Preserving GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUIGuard-Bench, which annotates private regions in GUI trajectories with categories, risk levels, and relevance to the user’s task. It evaluates privacy recognition and the effect of screenshot protection on offline planning and task utility.", "topics": ["safety-security", "evaluation", "grounding"], "updated_at": "2026-05-13", "url": "https://arxiv.org/abs/2601.18842"}
{"aliases": [], "authors": ["Libo Sun", "Jiwen Zhang", "Siyuan Wang", "Zhongyu Wei"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1930b4efca7ee38fc6cd1ceb283c14ed9b1388e7b758b951ba9469ed3067c3e0", "submitted_at": "2026-01-27T04:58:48Z", "updated_at": "2026-01-31T15:17:38Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.19199v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.19199", "identifiers": {"arxiv": "2601.19199"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.19199"}], "name": "MAGNET: Towards Adaptive GUI Agents with Memory-Driven Knowledge Evolution", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAGNET maintains separate memories for stable interface functions and task procedures across changing layouts and workflows. It updates frequently used knowledge and evaluates adaptation in AndroidWorld and offline distribution-shift settings.", "topics": ["memory", "grounding", "robustness", "evaluation"], "updated_at": "2026-01-31", "url": "https://arxiv.org/abs/2601.19199"}
{"aliases": [], "authors": ["Qinzhuo Wu", "Zhizhuo Yang", "Hanhao Li", "Pengzhi Gao", "Wei Liu", "Jian Luan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1cbf16dd3c790996758e00c085a9c2bccabf5b24f945bbd8440cc1f953e38568", "submitted_at": "2026-01-28T07:49:48Z", "updated_at": "2026-01-29T02:32:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.20335v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.20335", "identifiers": {"arxiv": "2601.20335"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.20335"}], "name": "MobileBench-OL: A Comprehensive Chinese Benchmark for Evaluating Mobile GUI Agents in Real-World Environment", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MobileBench-OL, an online benchmark of Chinese-app tasks spanning execution, reasoning, exploration, and noisy conditions. Its automatic evaluation and reset framework supports repeated testing of mobile GUI agents in real environments.", "topics": ["evaluation", "mobile-navigation", "robustness"], "updated_at": "2026-01-29", "url": "https://arxiv.org/abs/2601.20335"}
{"aliases": [], "authors": ["Le Zhang", "Yixiong Xiao", "Xinjiang Lu", "Jingjia Cao", "Yusai Zhao", "Jingbo Zhou", "Lang An", "Zikan Feng", "Wanxiang Sha", "Yu Shi", "Congxi Xiao", "Jian Xiong", "Yankai Zhang", "Hua Wu", "Haifeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ad764cf86a2ecab81f3f730d6d26ea582544a355a90a8be55be220c98aabd7bf", "submitted_at": "2026-01-28T08:45:17Z", "updated_at": "2026-01-28T08:45:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.20380v1"}}, "environments": ["android", "desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.20380", "identifiers": {"arxiv": "2601.20380"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.20380"}], "name": "OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Execution", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "OmegaUse combines exploratory and taxonomy-guided data synthesis with supervised fine-tuning followed by GRPO for GUI control. The report introduces OS-Nav for offline Android and Ubuntu evaluation alongside established grounding and action-prediction benchmarks.", "topics": ["data-generation", "reinforcement-learning", "grounding", "evaluation"], "updated_at": "2026-01-28", "url": "https://arxiv.org/abs/2601.20380"}
{"aliases": [], "authors": ["Jing Wu", "Wenjie Ai", "Daphne Barretto", "Yiye Chen", "Qingyu Chen", "Yuhang He", "Pranit Chawla", "Nicholas Gydé", "Yanan Jian", "Vibhav Vineet"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0139f7a1402d8dbfdecf65cc53c032495b2084f24cc38ff43841642c9102f29c", "submitted_at": "2026-01-28T14:35:23Z", "updated_at": "2026-08-26T22:31:44Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.20650v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.20650", "identifiers": {"arxiv": "2601.20650"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.20650"}], "name": "OS-Marathon: Benchmarking Computer-Use Agents on Vast-Horizon, Repetitive Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OS-Marathon to evaluate long repetitive computer-use workflows whose execution grows with the amount of input data. It studies failure accumulation under task decomposition and GraphDemo, which adapts agents using a demonstration of the recurring workflow.", "topics": ["evaluation", "long-horizon", "human-in-the-loop"], "updated_at": "2026-08-26", "url": "https://arxiv.org/abs/2601.20650"}
{"aliases": [], "authors": ["Ziwei Liu", "Borui Kang", "Hangjie Yuan", "Zixiang Zhao", "Wei Li", "Yifan Zhu", "Tao Feng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1083cf063ccf1781f2cdfe5e266b75b52079fd12ad16d05755577b880c010aec", "submitted_at": "2026-01-28T16:06:31Z", "updated_at": "2026-03-25T10:25:50Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2601.20732v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.20732", "identifiers": {"arxiv": "2601.20732"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.20732"}], "name": "Continual GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Continual GUI Agents defines grounding under changing application domains and screen resolutions. Its GUI-AiF training framework uses separate point- and region-alignment rewards to adapt grounding as these distributions shift.", "topics": ["grounding", "reinforcement-learning", "robustness", "evaluation"], "updated_at": "2026-03-25", "url": "https://arxiv.org/abs/2601.20732"}
{"aliases": [], "authors": ["Wei Minn", "Biniam Fisseha Demissie", "Yan Naing Tun", "Jiakun Liu", "Mariano Ceccato", "Lwin Khin Shar", "David Lo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ac3f148047f7e63a33780fed3fd1ca461bead2128914c46f353efab4675df444", "submitted_at": "2026-01-29T04:21:11Z", "updated_at": "2026-01-29T04:21:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.21253v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}], "id": "arxiv:2601.21253", "identifiers": {"arxiv": "2601.21253"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.21253"}], "name": "CovAgent: Overcoming the 30% Curse of Mobile Application Coverage with Agentic AI and Dynamic Instrumentation", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-01-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "CovAgent augments Android GUI fuzzers by analyzing decompiled app code and component transitions to identify unsatisfied activation conditions. A second agent generates instrumentation that enables otherwise difficult-to-reach activities, with coverage-based evaluation.", "topics": ["automation", "mobile-navigation", "evaluation"], "updated_at": "2026-01-29", "url": "https://arxiv.org/abs/2601.21253"}
{"aliases": [], "authors": ["Yao Zhang", "Shijie Tang", "Zeyu Li", "Zhen Han", "Volker Tresp"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "68ae67856eb871d9a8f259dfb180bde02d29ee97fd7abf0e60719fc2c7abd7ab", "submitted_at": "2026-01-29T15:39:50Z", "updated_at": "2026-04-09T17:48:11Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.21872v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.21872", "identifiers": {"arxiv": "2601.21872"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.21872"}], "name": "WebArbiter: A Principle-Guided Reasoning Process Reward Model for Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebArbiter generates reasoned action-preference verdicts for web navigation, using reasoning distillation followed by reinforcement learning. It introduces WebPRMBench and evaluates both preference judgments and reward-guided trajectory search.", "topics": ["reward-models", "reinforcement-learning", "evaluation", "web-navigation"], "updated_at": "2026-04-09", "url": "https://arxiv.org/abs/2601.21872"}
{"aliases": [], "authors": ["Kuai Yu", "Naicheng Yu", "Han Wang", "Rui Yang", "Huan Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1380697158d3a91334427cf067f3f91a751bd89d31103edecfa68171b28b8d66", "submitted_at": "2026-01-29T16:40:15Z", "updated_at": "2026-02-27T16:47:40Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.21961v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.21961", "identifiers": {"arxiv": "2601.21961"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.21961"}], "name": "How do Visual Attributes Influence Web Agents? A Comprehensive Evaluation of User Interface Design Factors", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces VAF, a controlled pipeline that varies webpage appearance while retaining item semantics. It measures both agent clicks and reasoning references to study which visual design factors change browsing decisions.", "topics": ["ui-design", "evaluation", "web-navigation"], "updated_at": "2026-02-27", "url": "https://arxiv.org/abs/2601.21961"}
{"aliases": [], "authors": ["Linjia Kang", "Zhimin Wang", "Yongkang Zhang", "Duo Wu", "Jinghe Wang", "Ming Ma", "Haopeng Yan", "Zhi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0d66365682d0ea4cee744c6c59d3a585c9272a31da5f8ff7c27983d719f070a7", "submitted_at": "2026-01-30T10:03:20Z", "updated_at": "2026-01-30T10:03:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.22781v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.22781", "identifiers": {"arxiv": "2601.22781"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.22781"}], "name": "Learning with Challenges: Adaptive Difficulty-Aware Data Generation for Mobile GUI Agent Training", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileGen estimates a mobile agent’s strengths across structural and semantic task difficulty, then samples new training targets around that profile. A controllable multi-agent generator creates the corresponding instructions and GUI trajectories.", "topics": ["data-generation", "mobile-navigation", "evaluation"], "updated_at": "2026-01-30", "url": "https://arxiv.org/abs/2601.22781"}
{"aliases": [], "authors": ["Chaoqun Cui", "Jing Huang", "Shijing Wang", "Liming Zheng", "Qingchao Kong", "Zhixiong Zeng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "86225f2e52e1394e2a0a56b1e82327d756bff95739e01ce7cc5898af49944aba", "submitted_at": "2026-01-31T07:36:54Z", "updated_at": "2026-07-25T06:54:35Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.00575v2"}}, "environments": ["desktop", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.00575", "identifiers": {"arxiv": "2602.00575"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.00575"}], "name": "Agentic Reward Modeling: Verifying GUI Agent via Progressive Trajectory-Grounded Interaction", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "VAGEN verifies GUI trajectories by first extracting readily available evidence and escalating to tool-based environment probes when needed. Evaluation on OSWorld-Verified and AndroidWorld studies verification accuracy and its cost.", "topics": ["reward-models", "tool-integration", "evaluation", "efficiency"], "updated_at": "2026-07-25", "url": "https://arxiv.org/abs/2602.00575"}
{"aliases": [], "authors": ["Yurun Chen", "Zeyi Liao", "Ping Yin", "Taotao Xie", "Keting Yin", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d1a6f98c0ab0a221c169796d9680e4b2a7bf7a0775c0493254f47d5e0a9b3c4e", "submitted_at": "2026-02-02T07:04:06Z", "updated_at": "2026-02-02T07:04:06Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.01725v1"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:38+00:00", "role": "primary", "sha256": "6e1266c5a47cd667d681d3143a0331b74b6f64416b7e01f44a93a001c38ef74b", "supports": ["summary", "environments", "scope"], "url": "https://arxiv.org/html/2602.01725v1"}], "id": "arxiv:2602.01725", "identifiers": {"arxiv": "2602.01725"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.01725"}], "name": "SafePred: A Predictive Guardrail for Computer-Using Agents via World Models", "notes": "The annotation was reviewed against the complete retained official abstract and the specified sections of the same recorded paper version. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "SafePred uses world-model risk forecasts to guide action filtering and task replanning. Its desktop and web safety evaluations distinguish policy compliance, task success and success under the specified policy.", "topics": ["safety-security", "world-models", "planning", "evaluation"], "updated_at": "2026-02-02", "url": "https://arxiv.org/abs/2602.01725"}
{"aliases": [], "authors": ["Qingni Wang", "Yue Fan", "Xin Eric Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c9557fc608a7fd7667746e9b3e06bc699b1796cdd77fe5e975ba1bd5a64690c1", "submitted_at": "2026-02-02T18:22:45Z", "updated_at": "2026-02-03T08:31:29Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.02419v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.02419", "identifiers": {"arxiv": "2602.02419"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.02419"}], "name": "SafeGround: Know When to Trust GUI Grounding Models via Uncertainty Calibration", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SafeGround, which estimates GUI-grounding uncertainty from the spatial dispersion of sampled predictions and calibrates a decision threshold before testing. It studies risk-aware selection on ScreenSpot-Pro, with statistical control scoped to the threshold’s false-discovery rate.", "topics": ["grounding", "robustness", "evaluation"], "updated_at": "2026-02-03", "url": "https://arxiv.org/abs/2602.02419"}
{"aliases": [], "authors": ["Sizhe Tang", "Rongqian Chen", "Tian Lan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0a01499ce704bb7916a3157ff2d3fc755a1115d9b27c7e529c153bce360a5404", "submitted_at": "2026-02-03T02:07:12Z", "updated_at": "2026-02-03T02:07:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.02995v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.02995", "identifiers": {"arxiv": "2602.02995"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.02995"}], "name": "Agent Alpha: Tree Search Unifying Generation, Exploration and Evaluation for Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Agent Alpha uses step-level Monte Carlo tree search to reuse successful prefixes and recover from early GUI mistakes. Comparison-based evaluation and diversity-constrained expansion guide the search, with experiments on OSWorld.", "topics": ["planning", "efficiency", "evaluation"], "updated_at": "2026-02-03", "url": "https://arxiv.org/abs/2602.02995"}
{"aliases": [], "authors": ["Xilong Wang", "Yinuo Liu", "Zhun Wang", "Dawn Song", "Neil Gong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "86702018906e64294d7b8daea618c5f152e8b94b604d27d6b164e82aff462564", "submitted_at": "2026-02-03T17:55:04Z", "updated_at": "2026-02-03T17:55:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.03792v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.03792", "identifiers": {"arxiv": "2602.03792"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.03792"}], "name": "WebSentinel: Detecting and Localizing Prompt Injection Attacks for Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebSentinel identifies suspicious webpage segments and checks each against its surrounding page context to detect and locate prompt injections. Its evaluation uses both contaminated and clean webpage datasets.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2026-02-03", "url": "https://arxiv.org/abs/2602.03792"}
{"aliases": [], "authors": ["Syed Nazmus Sakib", "Nafiul Haque", "Tapodhir Karmakar Taton", "Shahrear Bin Amin", "Shifat E. Arman"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "76fbeae086fb9ef21f820b02777a2ceb66ee37f8f1c978c7426589ea4e888e49", "submitted_at": "2026-02-05T06:24:23Z", "updated_at": "2026-09-08T14:40:01Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2602.05354v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.05354", "identifiers": {"arxiv": "2602.05354"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.05354"}], "name": "Do Web Agents Investigate Before They Decide?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MIRAGE, a web decision benchmark where agents must actively uncover evidence that can contradict the visible context. Separate investigation, reasoning, decision, and hallucination measures diagnose failures beyond reaching the right page.", "topics": ["evaluation", "web-navigation", "planning"], "updated_at": "2026-09-08", "url": "https://arxiv.org/abs/2602.05354"}
{"aliases": [], "authors": ["Guangyi Liu", "Pengxiang Zhao", "Yaozhen Liang", "Qinyi Luo", "Shunye Tang", "Yuxiang Chai", "Weifeng Lin", "Han Xiao", "WenHao Wang", "Siheng Chen", "Zhengxi Lu", "Gao Wu", "Hao Wang", "Liang Liu", "Yong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0692d6936a3e3c20a7ad61c69ae770f5669c3783ff837092254f8b3fb170cd0a", "submitted_at": "2026-02-03T17:01:59Z", "updated_at": "2026-08-27T06:48:18Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2602.06075v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.06075", "identifiers": {"arxiv": "2602.06075"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.06075"}], "name": "MemGUI-Bench: Benchmarking Memory of Mobile GUI Agents in Dynamic Environments", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "MemGUI-Bench evaluates short-term retention and cross-session experience reuse in mobile GUI agents using paired tasks and repeated-attempt protocols. Its staged evaluator measures memory fidelity, learning and execution efficiency.", "topics": ["memory", "evaluation", "mobile-navigation", "cross-app"], "updated_at": "2026-08-27", "url": "https://arxiv.org/abs/2602.06075"}
{"aliases": [], "authors": ["Jinbiao Wei", "Yilun Zhao", "Kangqi Ni", "Arman Cohan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ce896d4d846bf07fedc68b7f952532c4fefe24dcb49b51556e0d6c2bc648bb91", "submitted_at": "2026-02-06T19:55:26Z", "updated_at": "2026-04-12T06:09:23Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.07153v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.07153", "identifiers": {"arxiv": "2602.07153"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.07153"}], "name": "ANCHOR: Branch-Point Data Generation for GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "ANCHOR expands verified desktop demonstrations by generating new tasks at meaningful branch points and executing their instructions. State-aware verification and action filtering refine the resulting trajectories, which are evaluated through fine-tuned agents on OSWorld and WindowsAgentArena.", "topics": ["data-generation", "planning", "evaluation"], "updated_at": "2026-04-12", "url": "https://arxiv.org/abs/2602.07153"}
{"aliases": [], "authors": ["Ruijia Cheng", "Jenny T. Liang", "Eldon Schoop", "Jeffrey Nichols"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d556ebbfbf4fc5875395b449a92b8ff27a7ed59d41bc062ac60fcf175f9064f8", "submitted_at": "2026-02-07T00:19:08Z", "updated_at": "2026-02-07T00:19:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.07283v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.07283", "identifiers": {"arxiv": "2602.07283"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.07283"}], "name": "Mapping the Design Space of User Experience for Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Develops a taxonomy of computer-use-agent interaction design through system review and practitioner interviews. A Wizard-of-Oz study then examines how users respond to normal, erroneous, and risky web-based execution, focusing on control, explanation, prompts, and mental models.", "topics": ["human-in-the-loop", "ui-design", "evaluation"], "updated_at": "2026-02-07", "url": "https://arxiv.org/abs/2602.07283"}
{"aliases": [], "authors": ["Yuting Ning", "Jaylen Jones", "Zhehao Zhang", "Chentao Ye", "Weitong Ruan", "Junyi Li", "Rahul Gupta", "Huan Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "853f64c020f47129be01778d5212ade8ef78251959ad634ee04ad0fb9fe2989c", "submitted_at": "2026-02-09T18:41:15Z", "updated_at": "2026-06-25T04:36:12Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.08995v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:26:41+00:00", "role": "primary", "sha256": "487a5d21d2385320315eb5a4a2c76e1017973d6027e219be6d31d9328fd5ff9a", "supports": ["summary", "environments", "scope"], "url": "https://arxiv.org/html/2602.08995v2"}], "id": "arxiv:2602.08995", "identifiers": {"arxiv": "2602.08995"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.08995"}], "name": "When Actions Go Off-Task: Detecting and Correcting Misaligned Actions in Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract and the specified sections of the same recorded paper version. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "MisActBench provides action-level labels for task alignment, while DeAction checks proposed computer-use actions and requests revisions through structured feedback. The paper separately evaluates offline detection and interactive operation in OSWorld and RedTeamCUA.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-06-25", "url": "https://arxiv.org/abs/2602.08995"}
{"aliases": [], "authors": ["Haodong Li", "Jingwei Wu", "Quan Sun", "Guopeng Li", "Juanxi Tian", "Ziyu Guo", "Donghao Zhou", "Tianfei Ren", "Huanyu Zhang", "Yanlin Lai", "Ruichuan An", "Hongbo Peng", "Yuhong Dai", "Jia Wang", "Ziyang Meng", "Zheng Ge", "Xiangyu Zhang", "Daxin Jiang", "Pheng-Ann Heng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "43a01e0eb848a57a84767836d2c3c982bbf6094d22a3e70652dd42af52760bc5", "submitted_at": "2026-02-09T18:52:02Z", "updated_at": "2026-09-26T18:28:57Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2602.09007v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}], "id": "arxiv:2602.09007", "identifiers": {"arxiv": "2602.09007"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.09007"}], "name": "GUI-GenBench: Evaluating Image Generation Models as Interactive GUI Environments", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-02-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-GenBench tests image-generation models as action-conditioned GUI environments across five suites, including transitions, trajectories and grounding. Its VLM-based GUI-Score separates goal completion, interaction logic, consistency and visual properties.", "topics": ["evaluation", "world-models", "grounding", "screen-understanding"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2602.09007"}
{"aliases": [], "authors": ["Jiacheng Liu", "Yaxin Luo", "Jiacheng Cui", "Xinyi Shang", "Xiaohan Zhao", "Zhiqiang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3178684a322fcf07f847473139efdddf0b02213af468f1afbd285549d15bde56", "submitted_at": "2026-02-09T18:55:33Z", "updated_at": "2026-06-21T08:25:39Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.09012v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.09012", "identifiers": {"arxiv": "2602.09012"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.09012"}], "name": "Next-Gen CAPTCHAs: Leveraging the Cognitive Gap for Scalable and Diverse GUI-Agent Defense", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Next-Gen CAPTCHAs introduces generated interactive challenges intended to distinguish human users from GUI agents. The benchmark targets perception, memory, decisions and actions, with backend-supported challenge types designed for scalable instance generation.", "topics": ["evaluation", "safety-security", "data-generation"], "updated_at": "2026-06-21", "url": "https://arxiv.org/abs/2602.09012"}
{"aliases": [], "authors": ["Ananya Gubbi Mohanbabu", "Rosiana Natalie", "Brandon Kim", "Anhong Guo", "Amy Pavel"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c3606d090749737d4223a9f826400c2b2aebde245c623391f2a8c6702b2468f8", "submitted_at": "2026-02-10T01:00:55Z", "updated_at": "2026-02-10T01:00:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.09310v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.09310", "identifiers": {"arxiv": "2602.09310"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.09310"}], "name": "A11y-CUA Dataset: Characterizing the Accessibility Gap in Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces A11y-CUA, a dataset comparing how blind or low-vision users and sighted users perform everyday computer tasks. It contrasts those traces with agent behavior under default, keyboard-only, and magnifier conditions to identify barriers to shared control.", "topics": ["human-in-the-loop", "evaluation", "data-generation"], "updated_at": "2026-02-10", "url": "https://arxiv.org/abs/2602.09310"}
{"aliases": [], "authors": ["Xingyi Zhang", "Yulei Ye", "Kaifeng Huang", "Wenhao Li", "Xiangfeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "86a47f084b23cc70e754808f6404bed0bbd500f78fcac4ccc104f95052a49868", "submitted_at": "2026-02-11T12:54:53Z", "updated_at": "2026-02-11T12:54:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.10814v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.10814", "identifiers": {"arxiv": "2602.10814"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.10814"}], "name": "See, Plan, Snap: Evaluating Multimodal GUI Agents in Scratch", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "ScratchWorld evaluates program construction through the Scratch interface using fine-grained drag-and-drop actions and a complementary semantic-API mode. Runtime tests assess the resulting programs, allowing analysis of the gap between program reasoning and GUI manipulation.", "topics": ["evaluation", "grounding", "automation"], "updated_at": "2026-02-11", "url": "https://arxiv.org/abs/2602.10814"}
{"aliases": [], "authors": ["Xiwen Teoh", "Yun Lin", "Duc-Minh Nguyen", "Ruofei Ren", "Wenjie Zhang", "Jin Song Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2076fc8b0b7b37bdf43fe2e0b0cc5fa4c3a8cb7b44a41fdf80f59a8c05c7e1b3", "submitted_at": "2026-02-12T08:51:07Z", "updated_at": "2026-04-23T15:00:58Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2602.11724v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}], "id": "arxiv:2602.11724", "identifiers": {"arxiv": "2602.11724"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.11724"}], "name": "WebTestPilot: Agentic End-to-End Web Testing against Natural Language Specification by Inferring Oracles with Symbolized GUI Elements", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebTestPilot turns natural-language requirements into web test steps with inferred preconditions and postconditions over symbolized GUI elements. A benchmark of bug-injected web applications tests whether these context-aware oracles distinguish application defects from agent mistakes.", "topics": ["automation", "evaluation", "planning"], "updated_at": "2026-04-23", "url": "https://arxiv.org/abs/2602.11724"}
{"aliases": [], "authors": ["Jiazheng Sun", "Mingxuan Li", "Yingying Zhang", "Jiayang Niu", "Yachen Wu", "Ruihan Jin", "Shuyu Lei", "Pengrongrui Tan", "Zongyu Zhang", "Ruoyi Wang", "Jiachen Yang", "Boyu Yang", "Jiacheng Liu", "Xin Peng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2bbba09ca845f4fb772516d26d59fc3d99f8e87ba3657bb29c3c062b50925ec5", "submitted_at": "2026-02-12T09:25:15Z", "updated_at": "2026-02-12T09:25:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.11750v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.11750", "identifiers": {"arxiv": "2602.11750"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.11750"}], "name": "AmbiBench: Benchmarking Mobile GUI Agents Beyond One-Shot Instructions in the Wild", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "AmbiBench evaluates mobile GUI agents under four levels of instruction clarity, including tasks requiring active clarification. Its MUSE evaluator separately audits outcomes, execution quality and interaction quality to study alignment with user intent.", "topics": ["evaluation", "human-in-the-loop", "mobile-navigation"], "updated_at": "2026-02-12", "url": "https://arxiv.org/abs/2602.11750"}
{"aliases": [], "authors": ["Nicholas Lee", "Lutfi Eren Erdogan", "Chris Joseph John", "Surya Krishnapillai", "Michael W. Mahoney", "Kurt Keutzer", "Amir Gholami"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c64cd65fc7fec601658baf7836bdbe9c0e494bde2b0a4b66ec9d3090ee106b1f", "submitted_at": "2026-02-12T18:58:30Z", "updated_at": "2026-08-14T21:25:10Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.12276v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2602.12276", "identifiers": {"arxiv": "2602.12276"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.12276"}], "name": "Agentic Test-Time Scaling for WebAgents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "CATTS allocates additional inference computation when an agent’s action votes show uncertainty. The paper compares voting and arbiter strategies and evaluates this selective scaling on WebArena-Lite, Online-Mind2Web and GoBrowse.", "topics": ["efficiency", "planning", "evaluation", "web-navigation"], "updated_at": "2026-08-14", "url": "https://arxiv.org/abs/2602.12276"}
{"aliases": [], "authors": ["Lajanugen Logeswaran", "Jaekyeom Kim", "Sungryull Sohn", "Creighton Glasscock", "Honglak Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "555285a1ab51e02661067bdf8bcc3fa7f6ee0fdd63f52f786a03eb2483bd1a31", "submitted_at": "2026-02-13T02:52:18Z", "updated_at": "2026-02-13T02:52:18Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.12544v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.12544", "identifiers": {"arxiv": "2602.12544"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.12544"}], "name": "Scaling Web Agent Training through Automatic Data Generation and Fine-grained Evaluation", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "This work uses constraint-based progress evaluation to retain partially successful web trajectories for training. It introduces BookingArena for structured booking tasks and studies a distilled agent trained with the resulting data.", "topics": ["data-generation", "evaluation", "web-navigation"], "updated_at": "2026-02-13", "url": "https://arxiv.org/abs/2602.12544"}
{"aliases": [], "authors": ["Jaechul Roh", "Eugene Bagdasarian", "Hamed Haddadi", "Ali Shahin Shamsabadi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fb048b4e5d1d244d689957f30cdcb93cf4dbb750b24ae527b62ecfb04b554eb4", "submitted_at": "2026-02-13T23:02:50Z", "updated_at": "2026-02-13T23:02:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.13516v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2602.13516", "identifiers": {"arxiv": "2602.13516"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.13516"}], "name": "SPILLage: Agentic Oversharing on the Web", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Formalizes and measures natural agentic oversharing when browser agents handle private user emails and calendars across live websites.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2026-02-13", "url": "https://arxiv.org/abs/2602.13516"}
{"aliases": [], "authors": ["Yuan Cao", "Dezhi Ran", "Mengzhou Wu", "Yuzhe Guo", "Xin Chen", "Ang Li", "Gang Cao", "Gong Zhi", "Hao Yu", "Linyi Li", "Wei Yang", "Tao Xie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "29b4c8a5bda1df237fad33ce483018b3476684f904841ed0598ec4286234b858", "submitted_at": "2026-02-15T10:58:01Z", "updated_at": "2026-02-15T10:58:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.14093v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.14093", "identifiers": {"arxiv": "2602.14093"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.14093"}], "name": "GUI-GENESIS: Automated Synthesis of Efficient Environments with Verifiable Rewards for GUI Agent Post-Training", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-GENESIS reconstructs applications as lightweight web environments with executable assertions for reward computation. It evaluates the cost of these training environments and transfer from synthetic interaction training to held-out real tasks.", "topics": ["data-generation", "sandbox", "reinforcement-learning", "evaluation"], "updated_at": "2026-02-15", "url": "https://arxiv.org/abs/2602.14093"}
{"aliases": [], "authors": ["A. Said Gurbuz", "Sunghwan Hong", "Ahmed Nassar", "Marc Pollefeys", "Peter Staar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d7c6e23f704562c4ae4db9e0b4f8f2a8dde1318efc08876ee57f5967b178257e", "submitted_at": "2026-02-15T19:00:02Z", "updated_at": "2026-05-01T17:32:06Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.14276v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}], "id": "arxiv:2602.14276", "identifiers": {"arxiv": "2602.14276"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.14276"}], "name": "ScreenParse: Moving Beyond Sparse Grounding with Complete Screen Parsing Supervision", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "ScreenParse supplies dense annotations of UI elements, types and text in web screenshots through an automated rendering and filtering pipeline. ScreenVLM learns to output structured screen markup, and the study evaluates dense parsing and transfer to GUI grounding.", "topics": ["screen-understanding", "data-generation", "grounding", "evaluation"], "updated_at": "2026-05-01", "url": "https://arxiv.org/abs/2602.14276"}
{"aliases": [], "authors": ["Yifan Wu", "Yiran Peng", "Yiyu Chen", "Jianhao Ruan", "Zijie Zhuang", "Cheng Yang", "Jiayi Zhang", "Man Chen", "Yenchi Tseng", "Zhaoyang Yu", "Liang Chen", "Yuyao Zhai", "Bang Liu", "Chenglin Wu", "Yuyu Luo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3773ec51f5316b9006f93aeafd7b654bf0e7418253c0af16ef7b7c00fddea102", "submitted_at": "2026-02-15T20:03:19Z", "updated_at": "2026-02-15T20:03:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.14296v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}], "id": "arxiv:2602.14296", "identifiers": {"arxiv": "2602.14296"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.14296"}], "name": "AutoWebWorld: Synthesizing Infinite Verifiable Web Environments via Finite State Machines", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "AutoWebWorld turns finite-state task models into interactive websites whose actions and goal states can be checked programmatically. It generates verified trajectories for training and evaluates transfer to WebVoyager and Online-Mind2Web.", "topics": ["data-generation", "sandbox", "web-navigation", "evaluation"], "updated_at": "2026-02-15", "url": "https://arxiv.org/abs/2602.14296"}
{"aliases": [], "authors": ["Zikai Xiao", "Jianhong Tu", "Chuhang Zou", "Yuxin Zuo", "Zhi Li", "Peng Wang", "Bowen Yu", "Fei Huang", "Junyang Lin", "Zuozhu Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2dca0e4219a788205bbb060911d0f004d1db4cf2ff8a4962c04fd5777ea194d7", "submitted_at": "2026-02-16T13:06:49Z", "updated_at": "2026-02-16T13:06:49Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.14721v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.14721", "identifiers": {"arxiv": "2602.14721"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.14721"}], "name": "WebWorld: A Large-Scale World Model for Web Agent Training", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-02-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebWorld, an open-web simulator trained on interaction trajectories for generating agent training data and testing candidate actions. WebWorld-Bench evaluates simulated transitions, alongside downstream web-agent training experiments.", "topics": ["world-models", "data-generation", "evaluation"], "updated_at": "2026-02-16", "url": "https://arxiv.org/abs/2602.14721"}
{"aliases": [], "authors": ["Zexue He", "Yu Wang", "Churan Zhi", "Yuanzhe Hu", "Tzu-Ping Chen", "Lang Yin", "Ze Chen", "Tong Arthur Wu", "Siru Ouyang", "Zihan Wang", "Jiaxin Pei", "Julian McAuley", "Yejin Choi", "Alex Pentland"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2317a94d54b4458a1b7c5e92754aeaefcbf270aabfa0bd8ea333c673cd4596b3", "submitted_at": "2026-02-18T09:49:14Z", "updated_at": "2026-09-17T22:55:03Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.16313v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}], "id": "arxiv:2602.16313", "identifiers": {"arxiv": "2602.16313"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.16313"}], "name": "MemoryArena: Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-02-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MemoryArena links tasks across sessions so that agents must distill earlier interaction experience and use it in later decisions. Its evaluation includes web navigation alongside planning, search and formal reasoning, testing memory through subsequent actions rather than recall alone.", "topics": ["memory", "interaction-history", "planning", "evaluation", "long-horizon"], "updated_at": "2026-09-17", "url": "https://arxiv.org/abs/2602.16313"}
{"aliases": [], "authors": ["Serin Kim", "Sangam Lee", "Dongha Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "28a23bf22920f8c33317e9d16f59d7085f188e79157301ff573abe773cf562ce", "submitted_at": "2026-02-19T01:54:26Z", "updated_at": "2026-05-27T12:15:49Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2602.17003v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.17003", "identifiers": {"arxiv": "2602.17003"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.17003"}], "name": "Persona2Web: Benchmarking Personalized Web Agents for Contextual Reasoning with User History", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Persona2Web tests whether web agents can resolve ambiguous requests using preferences implicit in a user’s history. Its reasoning-aware evaluation compares agent architectures and history-access strategies on personalized tasks on the open web.", "topics": ["evaluation", "memory", "web-navigation"], "updated_at": "2026-05-27", "url": "https://arxiv.org/abs/2602.17003"}
{"aliases": [], "authors": ["Zhenquan Yao", "Zitong Huang", "Yihan Zeng", "Jianhua Han", "Hang Xu", "Chun-Mei Feng", "Jianwei Ma", "Wangmeng Zuo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "75f8eb625e969c2b180eeb3e90aa4e513ab5a1dc15657d3849c68b69bc5edf7c", "submitted_at": "2026-03-03T13:02:20Z", "updated_at": "2026-03-07T10:11:08Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.02951v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2603.02951", "identifiers": {"arxiv": "2603.02951"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.02951"}], "name": "CGL: Advancing Continual GUI Learning via Reinforcement Fine-Tuning", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The abstract describes artifact release as future work; this annotation does not assert current availability.", "published_at": "2026-03-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines supervised and reinforcement fine-tuning for continual GUI learning, using policy entropy to adjust their proportions and gradient projection to limit conflicting updates. AndroidControl-CL groups applications into sequential tasks to evaluate adaptation and retention.", "topics": ["reinforcement-learning", "robustness", "evaluation"], "updated_at": "2026-03-07", "url": "https://arxiv.org/abs/2603.02951"}
{"aliases": [], "authors": ["Haoyu Liu", "Dingcheng Li", "Lukas Rutishauser", "Zeyu Zheng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "971770e61bf0af91ad1f17b197f09fc166b26bfb4953d3f2c74d457a87e25f3a", "submitted_at": "2026-03-04T18:29:54Z", "updated_at": "2026-10-05T23:01:49Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.04364v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.04364", "identifiers": {"arxiv": "2603.04364"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.04364"}], "name": "Dual-Modality Multi-Stage Adversarial Safety Training: Robustifying Multimodal Web Agents Against Cross-Modal Attacks", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies webpage injections that affect both screenshots and accessibility trees, then trains web agents through teacher imitation, supervised adversarial examples, and adversarial reinforcement learning. The evaluation tests defenses on MiniWoB++ and held-out tasks.", "topics": ["safety-security", "reinforcement-learning", "evaluation"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2603.04364"}
{"aliases": [], "authors": ["Md Farhan Ishmam", "Kenneth Marino"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1901125b21431e41eacb5d5f0fc7a439d9a637c43762d441ece46a90e135051b", "submitted_at": "2026-03-05T08:43:06Z", "updated_at": "2026-10-14T10:16:21Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2603.04949v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.04949", "identifiers": {"arxiv": "2603.04949"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.04949"}], "name": "TimeWarp: Evaluating Web Agents by Revisiting the Past", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "TimeWarp tests web agents across multiple versions of the same sites, varying visual design, frontend code and workflows. TimeTraj transfers plans across versions to collect training trajectories and study robustness to interface change.", "topics": ["evaluation", "robustness", "web-navigation", "data-generation"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2603.04949"}
{"aliases": [], "authors": ["Sicheng Fan", "Rui Wan", "Yifei Leng", "Gaoning Liang", "Li Ling", "Yanyi Shang", "Dehan Kong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db52246d9f5f305046e6dd14f2ef15440c371678e1db0d05282c84a3667c6f10", "submitted_at": "2026-03-05T15:37:34Z", "updated_at": "2026-04-14T08:41:56Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2603.05295v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}], "id": "arxiv:2603.05295", "identifiers": {"arxiv": "2603.05295"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.05295"}], "name": "WebChain: A Large-Scale Human-Annotated Dataset of Real-World Web Interaction Traces", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebChain aligns human web-interaction trajectories with visual, structural and action data for agent training. It also introduces WebChainBench and a mid-training approach that separates spatial grounding from planning.", "topics": ["data-generation", "evaluation", "grounding", "planning"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2603.05295"}
{"aliases": [], "authors": ["Li Gu", "Zihuan Jiang", "Zhixiang Chi", "Huan Liu", "Ziqiang Wang", "Yuanhao Yu", "Glen Berseth", "Yang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "68350312fecc36630ecaeec18a203e692e072d31ff80f442662d202b43ead788", "submitted_at": "2026-03-08T03:08:11Z", "updated_at": "2026-03-08T03:08:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.07432v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2603.07432", "identifiers": {"arxiv": "2603.07432"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.07432"}], "name": "Generalization in Online Reinforcement Learning for Mobile Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AndroidWorld-Generalization with separate tests for unseen task instances, templates, and applications. A GRPO training system combines containerized environments with asynchronous trajectory collection to study these generalization regimes.", "topics": ["reinforcement-learning", "evaluation", "robustness"], "updated_at": "2026-03-08", "url": "https://arxiv.org/abs/2603.07432"}
{"aliases": [], "authors": ["Jiateng Liu", "Zhenhailong Wang", "Rushi Wang", "Bingxuan Li", "Jeonghwan Kim", "Aditi Tiwari", "Pengfei Yu", "Denghui Zhang", "Heng Ji"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "77eb68f31d616b8e542dbb41ada67f28bdc21f29f7da3f04a0077c0fa2d10196", "submitted_at": "2026-03-09T05:27:56Z", "updated_at": "2026-03-09T05:27:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.07978v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.07978", "identifiers": {"arxiv": "2603.07978"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.07978"}], "name": "OSExpert: Computer-Use Agents Learning Professional Skills via Exploration", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses GUI-based depth-first exploration to verify elementary software functions, assemble action primitives, and construct curricula of composed skills. OSExpert-Eval examines how the resulting procedural knowledge supports complex computer-use tasks.", "topics": ["planning", "data-generation", "evaluation"], "updated_at": "2026-03-09", "url": "https://arxiv.org/abs/2603.07978"}
{"aliases": [], "authors": ["Yuxiang Chai", "Shunye Tang", "Han Xiao", "Rui Liu", "Hongsheng Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7334e01b1fbf5cdd02c275c73d2a1b5c2931718d6019fcf6a0ea72b449704d94", "submitted_at": "2026-03-09T06:41:32Z", "updated_at": "2026-03-09T06:41:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.08013v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.08013", "identifiers": {"arxiv": "2603.08013"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.08013"}], "name": "PIRA-Bench: A Transition from Reactive GUI Agents to GUI-based Proactive Intent Recommendation Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "PIRA-Bench evaluates proactive intent recommendations from noisy GUI screenshot sequences containing interleaved tasks and user context. Its PIRF baseline combines memory and state tracking to maintain multiple task threads and identify opportunities to assist.", "topics": ["evaluation", "screen-understanding", "memory", "state-tracking", "human-in-the-loop"], "updated_at": "2026-03-09", "url": "https://arxiv.org/abs/2603.08013"}
{"aliases": [], "authors": ["Linxin Song", "Jieyu Zhang", "Huanxin Sheng", "Taiwei Shi", "Gupta Rahul", "Yang Liu", "Ranjay Krishna", "Jian Kang", "Jieyu Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "22b3001540ac8cc75117cdfb4f381462fe7df3fe7700a86e14b9f522803316ac", "submitted_at": "2026-03-10T19:17:22Z", "updated_at": "2026-03-10T19:17:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.10178v1"}}, "environments": ["desktop", "linux", "macos", "windows", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.10178", "identifiers": {"arxiv": "2603.10178"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.10178"}], "name": "Video-Based Reward Modeling for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds ExeVR-53k from execution-video, instruction and reward triplets, including negative instructions with step annotations. A reward model uses spatiotemporal token pruning to judge task completion from the instruction and execution keyframes.", "topics": ["reward-models", "video", "evaluation"], "updated_at": "2026-03-10", "url": "https://arxiv.org/abs/2603.10178"}
{"aliases": [], "authors": ["Syed Yusuf Ahmed", "Shiwei Feng", "Chanwoo Bae", "Calix Barrus Xiangyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8ad18d7fe815f5529fef663130b7ce8841763589e7f4319c0c7556ec2df1da03", "submitted_at": "2026-03-10T22:56:03Z", "updated_at": "2026-03-10T22:56:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.10268v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2603.10268", "identifiers": {"arxiv": "2603.10268"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.10268"}], "name": "SpecOps: A Fully Automated AI Agent Testing Framework in Real-World GUI Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "SpecOps divides testing of GUI-based agents into test generation, environment setup, execution and validation handled by specialist agents. Its evaluation spans five real-world agents, including web applications and browser extensions, and measures execution reliability and bug discovery.", "topics": ["automation", "evaluation", "planning"], "updated_at": "2026-03-10", "url": "https://arxiv.org/abs/2603.10268"}
{"aliases": [], "authors": ["Marta Sumyk", "Oleksandr Kosovan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "048dc7332372585872f5fd3da47a10bd999c7520b2999e45ea6f578b7c856400", "submitted_at": "2026-03-11T09:28:41Z", "updated_at": "2026-03-12T07:30:09Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.10577v2"}}, "environments": ["desktop", "macos", "windows", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.10577", "identifiers": {"arxiv": "2603.10577"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.10577"}], "name": "CUAAudit: Meta-Evaluation of Vision-Language Models as Auditors of Autonomous Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "CUAAudit examines vision-language models that judge computer-use task completion from an instruction and final environment state. Across macOS, Windows and Linux benchmarks, it measures accuracy, confidence calibration and disagreement between auditors.", "topics": ["evaluation", "observability", "robustness"], "updated_at": "2026-03-12", "url": "https://arxiv.org/abs/2603.10577"}
{"aliases": [], "authors": ["Xunzhuo Liu", "Bowei He", "Xue Liu", "Andy Luo", "Haichen Zhang", "Huamin Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "274deff6cacfb5f1b7b620a86abe7ef6496b673d929de39376a9ad42dd3c4c02", "submitted_at": "2026-03-13T09:21:25Z", "updated_at": "2026-03-13T09:21:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.12823v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.12823", "identifiers": {"arxiv": "2603.12823"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.12823"}], "name": "Adaptive Vision-Language Model Routing for Computer Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-03-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Adaptive VLM Routing estimates GUI-action difficulty and small-model confidence to select a model under a predicted reliability threshold. Grounding and agent-routing benchmarks assess the cost–accuracy trade-off, with prior interaction context used for warm agents.", "topics": ["efficiency", "grounding", "memory", "evaluation"], "updated_at": "2026-03-13", "url": "https://arxiv.org/abs/2603.12823"}
{"aliases": [], "authors": ["Yang Li", "Yuchen Liu", "Haoyu Lu", "Zhiqiang Xia", "Hongzhen Wang", "Kaiyang Han", "Changpeng Yang", "Jinyang Wu", "Jiaming Xu", "Runyu Shi", "Ying Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf6c3df80e405968d9ae2353639f3e039d3fcf8351d89db212bda038e3dfac9b", "submitted_at": "2026-03-16T09:45:33Z", "updated_at": "2026-03-16T09:45:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.15039v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.15039", "identifiers": {"arxiv": "2603.15039"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.15039"}], "name": "GUI-CEval: A Hierarchical and Comprehensive Chinese Benchmark for Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-CEval evaluates Chinese mobile GUI agents through a hierarchy of atomic abilities and application-level tasks collected on physical devices. Its diagnostics cover perception, planning, reflection, execution and post-action evaluation across a broad set of apps.", "topics": ["evaluation", "mobile-navigation", "self-reflection", "screen-understanding"], "updated_at": "2026-03-16", "url": "https://arxiv.org/abs/2603.15039"}
{"aliases": [], "authors": ["Mateusz Dziemian", "Maxwell Lin", "Xiaohan Fu", "Micha Nowak", "Nick Winter", "Eliot Jones", "Andy Zou", "Lama Ahmad", "Kamalika Chaudhuri", "Sahana Chennabasappa", "Xander Davies", "Lauren Deason", "Benjamin L. Edelman", "Tanner Emek", "Ivan Evtimov", "Jim Gust", "Maia Hamin", "Kat He", "Klaudia Krawiecka", "Riccardo Patana", "Neil Perry", "Troy Peterson", "Xiangyu Qi", "Javier Rando", "Zifan Wang", "Zihan Wang", "Spencer Whitman", "Eric Winsor", "Arman Zharmagambetov", "Matt Fredrikson", "Zico Kolter"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e7fe52247e01f331021013dbc3003b3b828e134d7706ec1f587210ad286eee1d", "submitted_at": "2026-03-16T14:49:36Z", "updated_at": "2026-03-16T14:49:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.15714v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}], "id": "arxiv:2603.15714", "identifiers": {"arxiv": "2603.15714"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:16Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.15714"}], "name": "How Vulnerable Are AI Agents to Indirect Prompt Injections? Insights from a Large-Scale Public Competition", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-03-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "This public red-teaming study evaluates indirect prompt injection that both changes agent behavior and hides the compromise from the final response. Its scenarios include computer use alongside tool calling and coding, with analyses of transfer and model-dependent vulnerability.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-03-16", "url": "https://arxiv.org/abs/2603.15714"}
{"aliases": [], "authors": ["Nathan Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e3b429d277ec776af18977a72aaeb95000a4fedf66aa89c0366ab02120d4cdbe", "submitted_at": "2026-03-18T04:41:16Z", "updated_at": "2026-03-18T04:41:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.17357v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.17357", "identifiers": {"arxiv": "2603.17357"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.17357"}], "name": "WebPII: Benchmarking Visual PII Detection for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebPII provides synthetic e-commerce screenshots with fine-grained annotations for sensitive information, including partially filled forms and transaction identifiers. It also trains WebRedact to detect visual PII as a privacy component for computer-use agents.", "topics": ["evaluation", "data-generation", "safety-security", "screen-understanding"], "updated_at": "2026-03-18", "url": "https://arxiv.org/abs/2603.17357"}
{"aliases": [], "authors": ["Yibo Shi", "Jungang Li", "Linghao Zhang", "Zihao Dongfang", "Biao Wu", "Sicheng Tao", "Yibo Yan", "Chenxi Qin", "Weiting Liu", "Zhixin Lin", "Hanqian Li", "Yu Huang", "Song Dai", "Yonghua Hei", "Yue Ding", "Xiang Li", "Shikang Wang", "Chengdong Xu", "Jingqi Liu", "Xueying Ma", "Zhiwen Zheng", "Xiaofei Zhang", "Bincheng Wang", "Nichen Yang", "Jie Wu", "Lihua Tian", "Chen Li", "Xuming Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0ce6da91e59ac57fa24ff991ca023b84448c54d477cacb2d6fded9bf527dd29d", "submitted_at": "2026-03-19T02:45:21Z", "updated_at": "2026-03-19T02:45:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.18429v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.18429", "identifiers": {"arxiv": "2603.18429"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.18429"}], "name": "AndroTMem: From Interaction Trajectories to Anchored Memory in Long-Horizon GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AndroTMem-Bench for Android tasks whose later actions depend on earlier UI states. Anchored State Memory retains causally linked intermediate states for subgoal-specific retrieval and evaluates them against full-history and summary-based memory.", "topics": ["memory", "long-horizon", "evaluation"], "updated_at": "2026-03-19", "url": "https://arxiv.org/abs/2603.18429"}
{"aliases": [], "authors": ["Haochen Zhao", "Shaoyang Cui"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0beb3fe342f080f0727181aba55fc1167d2512dd04542986448c5006b6b8213b", "submitted_at": "2026-03-19T11:14:45Z", "updated_at": "2026-03-19T11:14:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.18762v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2603.18762", "identifiers": {"arxiv": "2603.18762"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.18762"}], "name": "ClawTrap: A MITM-Based Red-Teaming Framework for Real-World OpenClaw Security Evaluation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents a man-in-the-middle network red-teaming framework for evaluating autonomous OpenClaw agents under live traffic manipulation.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2026-03-19", "url": "https://arxiv.org/abs/2603.18762"}
{"aliases": [], "authors": ["Zehao Li", "Zhenyu Wu", "Yibo Zhao", "Bowen Yang", "Jingjing Xie", "Zhaoyang Liu", "Zhoumianze Liu", "Kaiming Jin", "Jianze Liang", "Zonglin Li", "Feng Wu", "Bowen Zhou", "Zun Wang", "Zichen Ding"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ff31371580af6a0c43a6c26f076d5060ee06f974d3a6c1ade67348dc0af4859d", "submitted_at": "2026-03-19T17:47:47Z", "updated_at": "2026-03-19T17:47:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.19191v1"}}, "environments": ["android", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2603.19191", "identifiers": {"arxiv": "2603.19191"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.19191"}], "name": "OS-Themis: A Scalable Critic Framework for Generalist GUI Rewards", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents OS-Themis, a multi-agent critic that divides GUI trajectories into verifiable milestones and reviews their evidence before assigning an outcome. OGRBench evaluates GUI rewards across platforms, with AndroidWorld experiments examining training and trajectory filtering.", "topics": ["reward-models", "reinforcement-learning", "evaluation"], "updated_at": "2026-03-19", "url": "https://arxiv.org/abs/2603.19191"}
{"aliases": [], "authors": ["Taiyi Wang", "Sian Gooding", "Florian Hartmann", "Oriana Riva", "Edward Grefenstette"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5538933a0e3d11df6ff4d767849979703d83412599fbbb0f8b432c4d9850633b", "submitted_at": "2026-03-20T06:32:16Z", "updated_at": "2026-03-20T06:32:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.19685v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}], "id": "arxiv:2603.19685", "identifiers": {"arxiv": "2603.19685"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:16Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.19685"}], "name": "A Subgoal-driven Framework for Improving Long-Horizon LLM Agents", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-03-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "The framework separates inference-time subgoal planning from MiRA, a reinforcement-learning method that supplies milestone rewards. WebArena-Lite experiments evaluate these mechanisms for long-horizon web navigation.", "topics": ["planning", "reinforcement-learning", "reward-models", "long-horizon", "evaluation"], "updated_at": "2026-03-20", "url": "https://arxiv.org/abs/2603.19685"}
{"aliases": [], "authors": ["Rozain Shakeel", "Abdul Rahman Mohammad Ali", "Muneeb Mushtaq", "Tausifa Jan Saleem", "Tajamul Ashraf"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cfa6cd9aead44150fdc997654ded7253b2890e936b94e44be2fb33642b24d302", "submitted_at": "2026-03-20T14:43:53Z", "updated_at": "2026-03-20T14:43:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.19993v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T07:54:03Z", "role": "primary", "sha256": "13d5969845070cd2877dc51ae5300c2d623e5daef00c8a1552ffa46af3f721a5", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2401.05507%2C2406.07275%2C2409.16427%2C2410.00079%2C2503.07701%2C2503.22738%2C2505.15859%2C2505.17735%2C2505.23596%2C2507.05257%2C2507.06134%2C2507.21953%2C2507.23336%2C2508.06960%2C2508.08997%2C2509.02473%2C2509.23694%2C2510.09116%2C2510.14438%2C2511.18423%2C2512.12818%2C2512.13564%2C2601.10758%2C2601.12449%2C2601.21841%2C2603.01152%2C2603.19993%2C2603.20576%2C2604.04503%2C2604.17562%2C2604.23941%2C2605.16265%2C2606.06448%2C2606.06462%2C2606.29193%2C2607.01647%2C2607.13157&max_results=37"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T07:54:03Z", "role": "primary", "sha256": "13d5969845070cd2877dc51ae5300c2d623e5daef00c8a1552ffa46af3f721a5", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2401.05507%2C2406.07275%2C2409.16427%2C2410.00079%2C2503.07701%2C2503.22738%2C2505.15859%2C2505.17735%2C2505.23596%2C2507.05257%2C2507.06134%2C2507.21953%2C2507.23336%2C2508.06960%2C2508.08997%2C2509.02473%2C2509.23694%2C2510.09116%2C2510.14438%2C2511.18423%2C2512.12818%2C2512.13564%2C2601.10758%2C2601.12449%2C2601.21841%2C2603.01152%2C2603.19993%2C2603.20576%2C2604.04503%2C2604.17562%2C2604.23941%2C2605.16265%2C2606.06448%2C2606.06462%2C2606.29193%2C2607.01647%2C2607.13157&max_results=37"}], "id": "arxiv:2603.19993", "identifiers": {"arxiv": "2603.19993"}, "kind": "paper", "last_verified_at": "2026-10-16T07:54:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.19993"}], "name": "MedSPOT: A Workflow-Aware Sequential Grounding Benchmark for Clinical GUI", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2026-03-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Benchmarks sequential target localization in clinical software workflows, emphasizing how early mistakes affect subsequent interaction steps.", "topics": ["grounding", "evaluation"], "updated_at": "2026-03-20", "url": "https://arxiv.org/abs/2603.19993"}
{"aliases": [], "authors": ["Shoubin Yu", "Lei Shu", "Antoine Yang", "Yao Fu", "Srinivas Sunkara", "Maria Wang", "Jindong Chen", "Mohit Bansal", "Boqing Gong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3e4ed7391969359c2594e8b96b8a3a559ef98d2c84a10c1a2c39b3bd962b58c5", "submitted_at": "2026-03-23T19:47:58Z", "updated_at": "2026-03-23T19:47:58Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.22529v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.22529", "identifiers": {"arxiv": "2603.22529"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.22529"}], "name": "Ego2Web: A Web Agent Benchmark Grounded in Egocentric Videos", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Ego2Web pairs first-person videos with online tasks that require agents to connect the user’s physical context to web actions. It also evaluates an automatic task judge against human judgments and analyzes failures in video understanding and web execution.", "topics": ["evaluation", "video", "web-navigation", "planning"], "updated_at": "2026-03-23", "url": "https://arxiv.org/abs/2603.22529"}
{"aliases": [], "authors": ["Yutao Luo", "Haotian Zhu", "Shuchao Pang", "Zhigang Lu", "Tian Dong", "Yongbin Zhou", "Minhui Xue"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "37960f42162d2bff2fca5e71e350f1ef7c4d86e3f438fd777a88dc2280d3aabe", "submitted_at": "2026-03-24T09:51:43Z", "updated_at": "2026-03-24T09:51:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.23007v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.23007", "identifiers": {"arxiv": "2603.23007"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.23007"}], "name": "AgentRAE: Remote Action Execution through Notification-based Visual Backdoors against Screenshots-based Mobile GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-03-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "AgentRAE studies notification-based visual backdoors that redirect mobile GUI-agent actions. Its training separates sensitivity to icon differences from trigger/action association, with evaluations of clean behavior and targeted redirection.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": "2026-03-24", "url": "https://arxiv.org/abs/2603.23007"}
{"aliases": [], "authors": ["Akash Ghosh", "Tajamul Ashraf", "Rishu Kumar Singh", "Numan Saeed", "Sriparna Saha", "Xiuying Chen", "Salman Khan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8fb6a294616e1be46c5da05f60cda7d43c05d8a1ce71116c5e30cb94507cd56f", "submitted_at": "2026-03-25T10:25:48Z", "updated_at": "2026-03-25T10:25:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.24157v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}], "id": "arxiv:2603.24157", "identifiers": {"arxiv": "2603.24157"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:16Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.24157"}], "name": "CarePilot: A Multi-Agent Framework for Long-Horizon Computer Task Automation in Healthcare", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-03-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "CareFlow benchmarks long healthcare-software workflows, while CarePilot coordinates action proposals, execution checks and corrective feedback using short- and long-term memory. Its actor consumes visual interfaces and system state to select semantic actions.", "topics": ["automation", "planning", "memory", "evaluation", "long-horizon", "workflow"], "updated_at": "2026-03-25", "url": "https://arxiv.org/abs/2603.24157"}
{"aliases": [], "authors": ["Xiangru Jian", "Shravan Nayak", "Kevin Qinghong Lin", "Aarash Feizi", "Kaixin Li", "Patrice Bechard", "Spandana Gella", "Sai Rajeswar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "15237fda2f3799314009635017487749dbedc9b1a0ffdd6b56a8d28eaf01481d", "submitted_at": "2026-03-25T15:52:56Z", "updated_at": "2026-03-25T15:52:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.24440v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.24440", "identifiers": {"arxiv": "2603.24440"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.24440"}], "name": "CUA-Suite: Massive Human-annotated Video Demonstrations for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "CUA-Suite combines continuous expert desktop demonstrations in VideoCUA with the UI-Vision evaluation benchmark and GroundCUA grounding data. Screen recordings, cursor traces and layered annotations support study of temporal behavior and professional software workflows.", "topics": ["data-generation", "video", "evaluation", "grounding", "workflow"], "updated_at": "2026-03-25", "url": "https://arxiv.org/abs/2603.24440"}
{"aliases": [], "authors": ["Fanheng Kong", "Jingyuan Zhang", "Yang Yue", "Chenxi Sun", "Yang Tian", "Shi Feng", "Xiaocui Yang", "Daling Wang", "Yu Tian", "Jun Du", "Wenchong Zeng", "Han Li", "Kun Gai"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8414eb53c66f745fb074d9f2699f76ffea81cb55f98aacf9ce896401937f80c5", "submitted_at": "2026-03-26T09:27:29Z", "updated_at": "2026-03-26T09:27:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.25226v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.25226", "identifiers": {"arxiv": "2603.25226"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.25226"}], "name": "WebTestBench: Evaluating Computer-Use Agents towards End-to-End Automated Web Testing", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebTestBench for end-to-end website testing, separating checklist generation from defect detection. Its WebTester baseline evaluates test coverage, detection and sustained interaction across web application categories.", "topics": ["evaluation", "web-navigation", "automation"], "updated_at": "2026-03-26", "url": "https://arxiv.org/abs/2603.25226"}
{"aliases": [], "authors": ["Saelyne Yang", "Jaesang Yu", "Yi-Hao Peng", "Kevin Qinghong Lin", "Jae Won Cho", "Yale Song", "Juho Kim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b3dd3efca196fa3dd168431102dbb5f0f763e7d9e54b3d709ad2362a9c6f154", "submitted_at": "2026-03-26T19:37:53Z", "updated_at": "2026-03-26T19:37:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.25864v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2603.25864", "identifiers": {"arxiv": "2603.25864"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.25864"}], "name": "GUIDE: A Benchmark for Understanding and Assisting Users in Open-Ended GUI Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUIDE evaluates whether models can recognize a user’s behavior, infer intent and decide when to offer help during open-ended software tasks. It uses narrated screen recordings of novice users and studies the effect of additional user context on these assistance decisions.", "topics": ["evaluation", "human-in-the-loop", "video", "screen-understanding"], "updated_at": "2026-03-26", "url": "https://arxiv.org/abs/2603.25864"}
{"aliases": [], "authors": ["Deepak Akkil", "Mowafak Allaham", "Amal Raj", "Tamer Abuelsaad", "Ravi Kokku"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b4f03e65387e3aad99b8c4cef166a7a8b4a9e2954ad932b8daf566e30f156595", "submitted_at": "2026-03-30T21:27:28Z", "updated_at": "2026-03-30T21:27:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.29020v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.29020", "identifiers": {"arxiv": "2603.29020"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.29020"}], "name": "Emergence WebVoyager: Toward Consistent and Transparent Evaluation of (Web) Agents in The Wild", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Audits sources of ambiguity and operational variation in WebVoyager evaluation and introduces standardized guidance for task setup, failure handling, annotation and reporting. Emergence WebVoyager applies this protocol to a web-agent evaluation.", "topics": ["evaluation", "web-navigation"], "updated_at": "2026-03-30", "url": "https://arxiv.org/abs/2603.29020"}
{"aliases": [], "authors": ["Hongyi Nie", "Xunyuan Liu", "Yudong Bai", "Yaqing Wang", "Yang Liu", "Quanming Yao", "Zhen Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1c71b713efd6ea47e7c78f5dd51d06c5cfdba750529bfbcdfc8e1c69897cbcc9", "submitted_at": "2026-03-31T06:37:12Z", "updated_at": "2026-03-31T06:37:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.29318v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.29318", "identifiers": {"arxiv": "2603.29318"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.29318"}], "name": "PSPA-Bench: A Personalized Benchmark for Smartphone GUI Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-03-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "PSPA-Bench evaluates personalized smartphone GUI behavior using user-specific instructions and a process-aware assessment of interaction traces. Its comparison of GUI agents diagnoses weaknesses in adapting workflows and preferences rather than measuring generic task completion alone.", "topics": ["evaluation", "mobile-navigation"], "updated_at": "2026-03-31", "url": "https://arxiv.org/abs/2603.29318"}
{"aliases": [], "authors": ["Nan Wang", "Zhiwei Jin", "Chen Chen", "Haonan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ce7f70ff1f8c7d4058d8fdf00cb6cf1d61c0367a50ba76fc96d356bb07452c60", "submitted_at": "2026-04-01T13:33:27Z", "updated_at": "2026-04-01T13:33:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.00886v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}], "id": "arxiv:2604.00886", "identifiers": {"arxiv": "2604.00886"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:16Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.00886"}], "name": "PixelPrune: Pixel-Level Adaptive Visual Token Reduction via Predictive Coding", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent cross-domain VLM compression method with actual GUI-benchmark evaluation; the abstract does not establish a full computer-use policy.", "published_at": "2026-04-01", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "PixelPrune removes pixel-redundant patches before the vision encoder, reducing both encoding and downstream language-model work. The training-free method evaluates lossless and controlled-lossy compression on document and GUI tasks.", "topics": ["efficiency", "screen-understanding", "training-free", "evaluation"], "updated_at": "2026-04-01", "url": "https://arxiv.org/abs/2604.00886"}
{"aliases": [], "authors": ["Zhengyang Tang", "Ke Ji", "Xidong Wang", "Zihan Ye", "Xinyuan Wang", "Yiduo Guo", "Ziniu Li", "Chenxin Li", "Jingyuan Hu", "Shunian Chen", "Tongxu Luo", "Jiaxi Bi", "Zeyu Qin", "Shaobo Wang", "Xin Lai", "Pengyuan Lyu", "Junyi Li", "Can Xu", "Chengquan Zhang", "Han Hu", "Ming Yan", "Benyou Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3f137028b7608598fb1885a0bf1330e00c9404851571f856665aae1af1d9d63c", "submitted_at": "2026-04-01T14:50:50Z", "updated_at": "2026-04-02T13:51:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.00986v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2604.00986", "identifiers": {"arxiv": "2604.00986"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.00986"}], "name": "Do Phone-Use Agents Respect Your Privacy?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-04-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MyPhoneBench to verify whether phone-use agents minimize personal data disclosure when filling forms during benign mobile tasks.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2026-04-02", "url": "https://arxiv.org/abs/2604.00986"}
{"aliases": [], "authors": ["Mengzhou Wu", "Yuzhe Guo", "Yuan Cao", "Haochuan Lu", "Songhe Zhu", "Pingzhe Qu", "Xin Chen", "Kang Qin", "Zhongpu Wang", "Xiaode Zhang", "Xinyi Wang", "Wei Dai", "Gang Cao", "Yuetang Deng", "Zhi Gong", "Dezhi Ran", "Linyi Li", "Wei Yang", "Tao Xie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bac86c2ff2bf18feb3edfbe76285d33718d0c6d9564aff5a9d62f8778cdc5147", "submitted_at": "2026-02-11T17:02:38Z", "updated_at": "2026-02-11T17:02:38Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.02345v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.02345", "identifiers": {"arxiv": "2604.02345"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.02345"}], "name": "UI-Oceanus: Scaling GUI Agents with Synthetic Environmental Dynamics", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-02-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-Oceanus converts autonomous GUI exploration and execution feedback into supervision for predicting future interface states. It compares forward and inverse objectives during continual pretraining and examines downstream offline and online navigation as synthetic data scales.", "topics": ["world-models", "data-generation", "state-tracking", "evaluation"], "updated_at": "2026-02-11", "url": "https://arxiv.org/abs/2604.02345"}
{"aliases": [], "authors": ["Yuwen Zhai", "Runze Li", "Liang Wang", "Nian Shi", "Liwu Xu", "Wei Zhang", "Ran Lin", "Bo Xu", "Benlei Cui"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "856e9920aea7d821061b3258b31aa32fd37b5579d628f5507f45ba83f237b77b", "submitted_at": "2026-04-06T03:58:43Z", "updated_at": "2026-04-06T03:58:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.04399v1"}}, "environments": ["web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.04399", "identifiers": {"arxiv": "2604.04399"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.04399"}], "name": "GUIDE: Interpretable GUI Agent Evaluation via Hierarchical Diagnosis", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUIDE evaluates GUI trajectories by segmenting them into subtasks, diagnosing each segment, and aggregating structured findings into a task verdict. It tests diagnostic judgments on e-commerce, web-agent and Android-control traces rather than relying on one holistic trajectory score.", "topics": ["evaluation", "long-horizon", "state-tracking"], "updated_at": "2026-04-06", "url": "https://arxiv.org/abs/2604.04399"}
{"aliases": [], "authors": ["Chaoran Chen", "Zhiping Zhang", "Zeya Chen", "Eryue Xu", "Yinuo Yang", "Ibrahim Khalilov", "Simret A Gebreegziabher", "Yanfang Ye", "Ziang Xiao", "Yaxing Yao", "Tianshi Li", "Toby Jia-Jun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1411ee71ad8f65e4941af4ef538f52da6e9bf9c57fb537b6ecf37c8478609693", "submitted_at": "2026-04-06T17:57:16Z", "updated_at": "2026-09-12T06:05:52Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.04918v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.04918", "identifiers": {"arxiv": "2604.04918"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.04918"}], "name": "Comparing Human Oversight Strategies for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares four human oversight strategies during live web-agent sessions, examining exposure to problematic actions and opportunities to intervene. The analysis identifies timing and attention problems in control handoffs and upfront plan approval.", "topics": ["human-in-the-loop", "safety-security", "evaluation"], "updated_at": "2026-09-12", "url": "https://arxiv.org/abs/2604.04918"}
{"aliases": [], "authors": ["Rongqian Chen", "Yu Li", "Zeyu Fang", "Sizhe Tang", "Weidong Cao", "Tian Lan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "238aac264fb21022f7c546ae3aa3f4ea104c381f5d318f6d962f22fb00095216", "submitted_at": "2026-04-06T20:39:30Z", "updated_at": "2026-05-28T08:13:55Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.05157v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.05157", "identifiers": {"arxiv": "2604.05157"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.05157"}], "name": "IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains IntentScore on offline desktop GUI interactions to assess candidate actions using both their state relevance and planning intent. The reward model is evaluated as an action reranker for Agent S3 on OSWorld.", "topics": ["reward-models", "planning", "evaluation"], "updated_at": "2026-05-28", "url": "https://arxiv.org/abs/2604.05157"}
{"aliases": [], "authors": ["Pranjal Aggarwal", "Graham Neubig", "Sean Welleck"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fdd9cb6daad8832261ebe861fc1cb74429b819305e14ecfeb9698616013e6ac8", "submitted_at": "2026-04-07T17:38:15Z", "updated_at": "2026-04-07T17:38:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.06126v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2604.06126", "identifiers": {"arxiv": "2604.06126"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.06126"}], "name": "Gym-Anything: Turn any Software into an Agent Environment", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Gym-Anything, where a setup agent configures software environments and an independent auditor checks the resulting evidence. CUA-World supplies long-horizon tasks across professional applications, with experiments on trajectory distillation and test-time auditing.", "topics": ["data-generation", "sandbox", "long-horizon", "evaluation"], "updated_at": "2026-04-07", "url": "https://arxiv.org/abs/2604.06126"}
{"aliases": [], "authors": ["Corby Rosset", "Pratyusha Sharma", "Andrew Zhao", "Miguel Gonzalez-Fernandez", "Ahmed Awadallah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b452684b305c4f20ded686289b448fbe4f08bb3b34f515a43e2ce5d6dc4ef506", "submitted_at": "2026-04-05T06:26:05Z", "updated_at": "2026-04-05T06:26:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.06240v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.06240", "identifiers": {"arxiv": "2604.06240"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.06240"}], "name": "The Art of Building Verifiers for Computer Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a web-trajectory verifier built around structured rubrics, separate process and outcome judgments, and explicit treatment of uncontrollable failures. CUAVerifierBench supplies human labels for evaluating these choices and screenshot-based context management.", "topics": ["reward-models", "evaluation", "long-horizon"], "updated_at": "2026-04-05", "url": "https://arxiv.org/abs/2604.06240"}
{"aliases": [], "authors": ["Guruprasad Viswanathan Ramesh", "Asmit Nayak", "Basieem Siddique", "Kassem Fawaz"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "400d8a9f7aed4f150dd5198d9a1561d2df0454069495594ca165ad4254a1be59", "submitted_at": "2026-04-07T18:43:21Z", "updated_at": "2026-06-16T22:37:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.06367v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.06367", "identifiers": {"arxiv": "2604.06367"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.06367"}], "name": "WebSP-Eval: Evaluating Web Agents on Website Security and Privacy Tasks", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebSP-Eval evaluates agents carrying out website security and privacy tasks such as preference changes and session management. A browser extension controls account and initial state across runs, enabling analysis of failures on stateful UI controls.", "topics": ["safety-security", "evaluation", "web-navigation", "state-tracking"], "updated_at": "2026-06-16", "url": "https://arxiv.org/abs/2604.06367"}
{"aliases": [], "authors": ["Songze Li", "Xiaoke Guo", "Tianqi Liu", "Biao Yi", "Zhaoyan Gong", "Zhiqiang Liu", "Huajun Chen", "Wen Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b99643e43c5e20af088d6fa62152e03256c947af76a6b3c0abfc6f74a0f7d648", "submitted_at": "2026-04-08T12:12:09Z", "updated_at": "2026-10-02T06:30:10Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.06995v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}], "id": "arxiv:2604.06995", "identifiers": {"arxiv": "2604.06995"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:02Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.06995"}], "name": "What's Missing in Screen-to-Action? Towards a UI-in-the-Loop Paradigm for Multimodal GUI Reasoning", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces UILoop, which inserts explicit UI-element localization, function and usage reasoning between screen observation and action. UI Comprehension-Bench evaluates these element-level capabilities with three metrics.", "topics": ["screen-understanding", "grounding", "evaluation"], "updated_at": "2026-10-02", "url": "https://arxiv.org/abs/2604.06995"}
{"aliases": [], "authors": ["Wenkui Yang", "Chao Jin", "Haisu Zhu", "Weilin Luo", "Derek Yuen", "Kun Shao", "Junxian Duan", "Huaibo Huang", "Jie Cao", "Ran He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a5427eedff0e97f9c73101403754c38e8928ad71abab7b7eff5f136a30ee2205", "submitted_at": "2026-04-09T05:32:34Z", "updated_at": "2026-07-08T02:39:35Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.07831v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.07831", "identifiers": {"arxiv": "2604.07831"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.07831"}], "name": "Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element Injection", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study tests black-box GUI-grounding failures caused by strategically added, innocuous-looking UI elements. Its iterative overlay search is compared with random injection and repeated independent grounding trials to distinguish persistent misdirection from incidental clutter.", "topics": ["safety-security", "grounding", "robustness", "evaluation"], "updated_at": "2026-07-08", "url": "https://arxiv.org/abs/2604.07831"}
{"aliases": [], "authors": ["Yushi Feng", "Junye Du", "Qifan Wang", "Zizhan Ma", "Qian Niu", "Yutaka Matsuo", "Long Feng", "Lequan Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e53647528b9ee9d98f93ee665d0d886f9b400a129873e240704041df291f5749", "submitted_at": "2026-04-10T09:41:21Z", "updated_at": "2026-04-10T09:41:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.09155v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2604.09155", "identifiers": {"arxiv": "2604.09155"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.09155"}], "name": "CORA: Conformal Risk-Controlled Agents for Safeguarded Mobile GUI Automation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The annotation describes risk calibration without asserting an unconditional safety guarantee.", "published_at": "2026-04-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CORA, which calibrates a risk model to choose between executing and abstaining from mobile GUI actions, then diagnoses rejected actions to suggest interventions. Phone-Harm provides step-level safety labels for evaluating this safeguarding process.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2026-04-10", "url": "https://arxiv.org/abs/2604.09155"}
{"aliases": [], "authors": ["Wee Joe Tan", "Zi Rui Lucas Lim", "Shashank Durgad", "Karim Obegi", "Aiden Yiliu Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "445e520a60e0e3cc0c3c519e21bfa8eebfa33d4fe4320c437e67b94a02d84504", "submitted_at": "2026-02-25T18:59:42Z", "updated_at": "2026-04-14T20:54:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.09581v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.09581", "identifiers": {"arxiv": "2604.09581"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.09581"}], "name": "Avenir-UX: Automated UX Evaluation via Simulated Human Web Interaction with GUI Grounding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Avenir-UX combines grounded web interaction with simulated user profiles and structured usability measures to produce evaluations of website journeys. The paper describes its architecture and use of SUS, step-wise ease questions and think-aloud traces for automated UX reporting.", "topics": ["automation", "evaluation", "ui-design", "grounding"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2604.09581"}
{"aliases": [], "authors": ["Suhana Bedi", "Ryan Welch", "Ethan Steinberg", "Michael Wornow", "Taeil Matthew Kim", "Haroun Ahmed", "Peter Sterling", "Bravim Purohit", "Qurat Akram", "Angelic Acosta", "Esther Nubla", "Pritika Sharma", "Michael A. Pfeffer", "Sanmi Koyejo", "Nigam H. Shah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "55cedd4ad9760618fe9b3354e384f69b69b23b65f75837a3898484c9c6e55586", "submitted_at": "2026-04-10T22:33:39Z", "updated_at": "2026-04-10T22:33:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.09937v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.09937", "identifiers": {"arxiv": "2604.09937"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.09937"}], "name": "HealthAdminBench: Evaluating Computer-Use Agents on Healthcare Administration Tasks", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces HealthAdminBench with GUI environments for an electronic health record, payer portals and fax handling. Expert-defined administrative tasks are decomposed into verifiable subtasks to compare local progress with end-to-end workflow completion.", "topics": ["evaluation", "workflow", "industry"], "updated_at": "2026-04-10", "url": "https://arxiv.org/abs/2604.09937"}
{"aliases": [], "authors": ["Peng Yuan", "Yuyang Yin", "Yuxuan Cai", "Zheng Wei"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1db8d789fb8ba2d5e61ca24536670fd83491c305f532adfecf7193821f7eea52", "submitted_at": "2026-04-13T04:45:27Z", "updated_at": "2026-04-13T04:45:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.10988v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.10988", "identifiers": {"arxiv": "2604.10988"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.10988"}], "name": "WebForge: Breaking the Realism-Reproducibility-Scalability Trilemma in Browser Agent Benchmark", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents WebForge, a pipeline that plans, generates, refines and validates self-contained interactive websites. WebForge-Bench uses these environments to evaluate browser agents across domains and controlled difficulty levels.", "topics": ["data-generation", "evaluation", "web-navigation"], "updated_at": "2026-04-13", "url": "https://arxiv.org/abs/2604.10988"}
{"aliases": [], "authors": ["Fei Tang", "Zhiqiong Lu", "Boxuan Zhang", "Weiming Lu", "Jun Xiao", "Yueting Zhuang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "df56b38203bc5ec8d7e871eb2ad9fca9891a049df62b7d72796fe400eb8bce5e", "submitted_at": "2026-04-13T17:52:04Z", "updated_at": "2026-04-13T17:52:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.11784v1"}}, "environments": ["android", "harmonyos", "ios"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.11784", "identifiers": {"arxiv": "2604.11784"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.11784"}], "name": "ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents ClawGUI, a shared harness for GUI-agent reinforcement learning, benchmark evaluation and deployment on mobile devices. It combines parallel virtual and physical environments, process rewards, and hybrid CLI-GUI execution with persistent memory.", "topics": ["reinforcement-learning", "evaluation", "tool-integration", "mobile-navigation"], "updated_at": "2026-04-13", "url": "https://arxiv.org/abs/2604.11784"}
{"aliases": [], "authors": ["Yulin Chen", "Tri Cao", "Haoran Li", "Yue Liu", "Yibo Li", "Yufei He", "Le Minh Khoi", "Yangqiu Song", "Shuicheng Yan", "Bryan Hooi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "64d4323050a5ad0f7265231337eb22a4c0823cee41587099c8bdc5db6b320dea", "submitted_at": "2026-04-14T04:50:35Z", "updated_at": "2026-04-14T04:50:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.12284v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.12284", "identifiers": {"arxiv": "2604.12284"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.12284"}], "name": "WebAgentGuard: A Reasoning-Driven Guard Model for Detecting Prompt Injection Attacks in Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebAgentGuard separates prompt-injection detection from the operating web agent through a parallel multimodal guard. The guard is trained on synthetic visual/textual examples using reasoning-focused supervised fine-tuning followed by reinforcement learning.", "topics": ["safety-security", "data-generation", "imitation-learning", "reinforcement-learning", "evaluation"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2604.12284"}
{"aliases": [], "authors": ["Yangyue Wang", "Harshvardhan Sikka", "Yash Mathur", "Tony Zhou", "Jinu Nyachhyon", "Pranav Guruprasad"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5dc36ff0c9c4bd39de360824967e53c18fe57562cb8da1462d3e63c0dd348e71", "submitted_at": "2026-04-15T16:39:22Z", "updated_at": "2026-07-01T10:12:28Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.14262v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.14262", "identifiers": {"arxiv": "2604.14262"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.14262"}], "name": "GUI-Perturbed: Domain Randomization Reveals Systematic Brittleness in GUI Grounding Models", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUI-Perturbed, which varies screenshot appearance and instruction wording independently to diagnose grounding failures. Experiments on three related model architectures examine relational instructions, browser zoom and augmentation-based fine-tuning.", "topics": ["grounding", "robustness", "evaluation"], "updated_at": "2026-07-01", "url": "https://arxiv.org/abs/2604.14262"}
{"aliases": [], "authors": ["Pan Hao", "Rishi Selvakumaran", "Jacob Sun", "Qianwen Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bd3b5e5b9a607023c82150673255eaf2f90c22db86d3c049b2a2fd536aaef17f", "submitted_at": "2026-04-16T06:22:48Z", "updated_at": "2026-04-16T06:22:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.14668v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.14668", "identifiers": {"arxiv": "2604.14668"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.14668"}], "name": "Beyond Chat and Clicks: GUI Agents for In-Situ Assistance via Live Interface Transformation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "DOMSteer translates help requests and live page context into reversible DOM changes such as control highlighting, contextual tooltips and layout reorganization. The work evaluates in-place GUI assistance with interface tests, use cases and a comparative user study.", "topics": ["human-in-the-loop", "ui-design", "grounding", "evaluation"], "updated_at": "2026-04-16", "url": "https://arxiv.org/abs/2604.14668"}
{"aliases": [], "authors": ["Wenhao Wang", "Haoting Shi", "Mengying Yuan", "Yiquan Lin", "Panrong Tong", "Hanzhang Zhou", "Guangyi Liu", "Pengxiang Zhao", "Yue Wang", "Siheng Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1e2b5d7149adcbe634df0bfdc9c34ba436e51e6891df5c4836daea9795061164", "submitted_at": "2026-04-16T12:53:03Z", "updated_at": "2026-04-16T12:53:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.14956v1"}}, "environments": ["web", "desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.14956", "identifiers": {"arxiv": "2604.14956"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.14956"}], "name": "FedGUI: Benchmarking Federated GUI Agents across Heterogeneous Platforms, Devices, and Operating Systems", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces FedGUI, a federated GUI-learning benchmark built from six datasets spanning mobile, web and desktop interfaces. Its experiments separate platform, device, operating-system and data-source heterogeneity when evaluating collaborative training.", "topics": ["evaluation", "robustness"], "updated_at": "2026-04-16", "url": "https://arxiv.org/abs/2604.14956"}
{"aliases": [], "authors": ["Chao Jin", "Wenkui Yang", "Hao Sun", "Yuqi Liao", "Qianyi Jiang", "Kai Zhou", "Jie Cao", "Ran He", "Huaibo Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2b102c9363b1be77273c7fab6f7f5cde74157c646671c93b9503770fe60dd70c", "submitted_at": "2026-04-19T06:55:59Z", "updated_at": "2026-04-19T06:55:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.17284v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.17284", "identifiers": {"arxiv": "2604.17284"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.17284"}], "name": "HalluClear: Diagnosing, Evaluating and Mitigating Hallucinations in GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces HalluClear with a GUI hallucination taxonomy, expert-calibrated evaluation workflow and structured-reasoning post-training scheme. It evaluates how these components affect hallucinations, grounding and action fidelity across GUI agents.", "topics": ["robustness", "grounding", "evaluation"], "updated_at": "2026-04-19", "url": "https://arxiv.org/abs/2604.17284"}
{"aliases": [], "authors": ["Shiquan Zhang", "Tianyi Zhang", "Le Fang", "Simon D'Alfonso", "Hong Jia", "Vassilis Kostakos"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "176f4d68e2f5084bee1938ee1399978052cdce2d2250ef06c16ddbf49c977da1", "submitted_at": "2026-04-20T05:15:14Z", "updated_at": "2026-04-20T05:15:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.17817v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2604.17817", "identifiers": {"arxiv": "2604.17817"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.17817"}], "name": "Do LLMs Need to See Everything? A Benchmark and Study of Failures in LLM-driven Smartphone Automation using Screentext vs. Screenshots", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "DailyDroid compares text-only and text-plus-screenshot agents on everyday Android tasks across multiple apps and difficulty levels. Its failure analysis examines accessibility, observation modality and agent/application behavior.", "topics": ["evaluation", "mobile-navigation", "screen-understanding", "robustness"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2604.17817"}
{"aliases": [], "authors": ["Gonzalo Gonzalez-Pumariega", "Saaket Agashe", "Jiachen Yang", "Ang Li", "Xin Eric Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a508db9eba8a7db5b52273bff32a16189f5a8eed19d6563e0f84313188a0aad1", "submitted_at": "2026-04-20T05:59:04Z", "updated_at": "2026-04-20T05:59:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.17849v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.17849", "identifiers": {"arxiv": "2604.17849"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.17849"}], "name": "On the Reliability of Computer Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study examines repeated OSWorld executions to separate execution stochasticity, task-specification ambiguity and behavioral variation. Paired task-level tests show why one successful run is insufficient evidence of reliable computer use.", "topics": ["evaluation", "robustness"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2604.17849"}
{"aliases": [], "authors": ["Wentao Shi", "Yu Wang", "Yuyang Zhao", "Yuxin Chen", "Fuli Feng", "Xueyuan Hao", "Xi Su", "Qi Gu", "Hui Su", "Xunliang Cai", "Xiangnan He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "91a9cdb6c8cb124d2e6033f0decc85e0c73bd972984d973ff687bbdc2be21176", "submitted_at": "2026-04-20T13:23:38Z", "updated_at": "2026-04-20T13:23:38Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.18240v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2604.18240", "identifiers": {"arxiv": "2604.18240"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.18240"}], "name": "AJ-Bench: Benchmarking Agent-as-a-Judge for Environment-Aware Evaluation", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-04-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "AJ-Bench evaluates judge agents that actively gather evidence from search, data systems and graphical interfaces. Its annotated tasks and trajectories test information acquisition, state verification and process verification.", "topics": ["evaluation", "reward-models", "state-tracking"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2604.18240"}
{"aliases": [], "authors": ["Wenpeng Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8eb499729268ef287a9b9bf73275f526d47dabf2d7de399beaead2fe54cfd74e", "submitted_at": "2026-04-20T21:36:16Z", "updated_at": "2026-04-20T21:36:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.18860v1"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.18860", "identifiers": {"arxiv": "2604.18860"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.18860"}], "name": "Temporal UI State Inconsistency in Desktop GUI Agents: Formalizing and Defending Against TOCTOU Attacks on Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study examines UI changes between a desktop agent’s screenshot observation and action dispatch. Its pre-execution state checks address overlay and focus manipulation, while the reported DOM-injection blind spot limits the defense’s coverage.", "topics": ["safety-security", "robustness", "evaluation", "state-tracking"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2604.18860"}
{"aliases": [], "authors": ["Zhiyuan Peng", "Wei Tao", "Xin Yin", "Chenhao Ying", "Yuan Luo", "Yiwen Guo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "74ec7a508a815781cc7e721aa6c5e351f086f6cab46da48e22c2b58f52af8ce0", "submitted_at": "2026-04-21T17:59:16Z", "updated_at": "2026-09-29T10:11:48Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.19742v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2604.19742", "identifiers": {"arxiv": "2604.19742"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.19742"}], "name": "PlayCoder: Making LLM-Generated GUI Code Playable", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-04-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "PlayEval measures whether generated GUI applications remain logically playable, using PlayTester to execute task-oriented GUI playthroughs. PlayCoder closes the loop between repository-aware generation, interactive testing and targeted code repair.", "topics": ["evaluation", "automation", "ui-design", "workflow"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2604.19742"}
{"aliases": [], "authors": ["Qijun Han", "Haoqin Tu", "Zijun Wang", "Haoyue Dai", "Yiyang Zhou", "Nancy Lau", "Alvaro A. Cardenas", "Yuhui Xu", "Ran Xu", "Caiming Xiong", "Zeyu Zheng", "Huaxiu Yao", "Yuyin Zhou", "Cihang Xie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7e675476121ab3ef51a8627ccfcc86d76eddde73987ea882398f72e3cb31c5d9", "submitted_at": "2026-04-23T07:42:37Z", "updated_at": "2026-09-24T23:58:02Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.21375v3"}}, "environments": ["desktop", "linux", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}], "id": "arxiv:2604.21375", "identifiers": {"arxiv": "2604.21375"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:02Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.21375"}], "name": "VLAA-GUI: Knowing When to Stop, Recover, and Search, A Modular Framework for GUI Automation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "VLAA-GUI combines mandatory checks for visually supported completion, multi-tier loop recovery and on-demand workflow search. Optional coding and grounding agents support execution, with component ablations on Linux and Windows GUI tasks.", "topics": ["robustness", "self-reflection", "tool-integration", "evaluation"], "updated_at": "2026-09-24", "url": "https://arxiv.org/abs/2604.21375"}
{"aliases": [], "authors": ["Zheng Wu", "Yi Hua", "Zhaoyuan Huang", "Chenhao Xue", "Yijie Lu", "Pengzhou Cheng", "Zongru Wu", "Lingzhong Dong", "Gongshen Liu", "Xinghao Jiang", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "772420683208561df8493c9ed5d68b341e9e098f5729945b972f3555e976d799", "submitted_at": "2026-04-27T11:44:26Z", "updated_at": "2026-04-27T11:44:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.24348v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2604.24348", "identifiers": {"arxiv": "2604.24348"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.24348"}], "name": "OS-SPEAR: A Toolkit for the Safety, Performance,Efficiency, and Robustness Analysis of OS Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-04-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Provides an evaluation toolkit for joint safety, efficiency, and multimodal robustness analysis of operating-system GUI agents.", "topics": ["evaluation", "safety-security", "efficiency", "robustness"], "updated_at": "2026-04-27", "url": "https://arxiv.org/abs/2604.24348"}
{"aliases": [], "authors": ["Hongxin Li", "Xiping Wang", "Jingran Su", "Zheng Ju", "Yuntao Chen", "Qing Li", "Zhaoxiang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "65325912365effe8adbc39301a378d7526c60d1185691f3a89cb8acd4aa53c3f", "submitted_at": "2026-04-27T13:06:27Z", "updated_at": "2026-04-27T13:06:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.24441v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}], "id": "arxiv:2604.24441", "identifiers": {"arxiv": "2604.24441"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:02Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.24441"}], "name": "AutoGUI-v2: A Comprehensive Multi-Modal GUI Functionality Understanding Benchmark", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "AutoGUI-v2 evaluates interface functionality, grounding and predicted interaction outcomes using hierarchical regions derived through VLM-human annotation. Its tasks span multiple operating systems and expose differences between functional localization, captioning and transition-logic understanding.", "topics": ["screen-understanding", "grounding", "state-tracking", "evaluation", "data-generation"], "updated_at": "2026-04-27", "url": "https://arxiv.org/abs/2604.24441"}
{"aliases": [], "authors": ["Lawrence Keunho Jang", "Jing Yu Koh", "Daniel Fried", "Ruslan Salakhutdinov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db70a5736d04d26f10afe3d9a0cfb08955f65f69480a0345ff13b30b2e83df23", "submitted_at": "2026-04-27T20:05:41Z", "updated_at": "2026-04-27T20:05:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.24964v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.24964", "identifiers": {"arxiv": "2604.24964"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.24964"}], "name": "Odysseys: Benchmarking Web Agents on Realistic Long Horizon Tasks", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Odysseys, a live-web benchmark of long, multi-site tasks derived from browsing sessions. Graded task rubrics and a rubric-score-per-step measure evaluate both partial progress and trajectory efficiency.", "topics": ["evaluation", "long-horizon", "efficiency"], "updated_at": "2026-04-27", "url": "https://arxiv.org/abs/2604.24964"}
{"aliases": [], "authors": ["Enqi Liu", "Liyuan Pan", "Zhi Gao", "Yan Yang", "Chenrui Shi", "Yang Liu", "Jingrong Wu", "Qing Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "54b49f54319922268ba880a1c1e8a27ffa7597e228b18be04caa1a6952a6456a", "submitted_at": "2026-04-28T08:43:58Z", "updated_at": "2026-05-08T04:29:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.25380v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.25380", "identifiers": {"arxiv": "2604.25380"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.25380"}], "name": "Benchmarking and Improving GUI Agents in High-Dynamic Environments", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces DynamicGUIBench for applications whose interfaces change between actions, together with DynamicUI, an agent that processes screen recordings. The agent selects informative frames, filters inconsistent action reasoning and uses trajectory reflection for subsequent decisions.", "topics": ["video", "state-tracking", "evaluation"], "updated_at": "2026-05-08", "url": "https://arxiv.org/abs/2604.25380"}
{"aliases": [], "authors": ["Mengyao Du", "Han Fang", "Haokai Ma", "Jiahao Chen", "Kai Xu", "Quanjun Yin", "Ee-Chien Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d2610c2658221d7da0bd888bf7b143ca6c4175791058570f6562cfa0c8c3a946", "submitted_at": "2026-04-28T12:32:21Z", "updated_at": "2026-04-28T12:32:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.25562v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.25562", "identifiers": {"arxiv": "2604.25562"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.25562"}], "name": "SnapGuard: Lightweight Prompt Injection Detection for Screenshot-Based Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "SnapGuard detects screenshot-based prompt injection through complementary visual-stability and recovered textual cues. Its evaluation compares detection quality and inference cost across attack and benign-page settings, rather than establishing universal runtime prevention.", "topics": ["safety-security", "screen-understanding", "evaluation", "efficiency"], "updated_at": "2026-04-28", "url": "https://arxiv.org/abs/2604.25562"}
{"aliases": [], "authors": ["Alice Gao", "Weixi Tong", "Rishab Vempati", "Katharina Reinecke", "R. Benjamin Shapiro", "Tianyi Zhang", "Jason Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bb6d24f7dc38921da883c4850b37f1c7eb48a03de920356d4ec2e095dbb44b04", "submitted_at": "2026-04-28T18:04:11Z", "updated_at": "2026-04-28T18:04:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.26020v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.26020", "identifiers": {"arxiv": "2604.26020"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.26020"}], "name": "Training Computer Use Agents to Assess the Usability of Graphical User Interfaces", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "uxCUA learns to assess interface usability by choosing interaction flows, executing them and predicting a numerical usability score. Training uses interactive UIs with usability labels and human preferences, with evaluations on synthetic and real interfaces.", "topics": ["ui-design", "evaluation", "automation"], "updated_at": "2026-04-28", "url": "https://arxiv.org/abs/2604.26020"}
{"aliases": [], "authors": ["Chen Liang", "Xirui Jiang", "Naihao Deng", "Eytan Adar", "Anhong Guo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5b572e37eefab40a6a21fa4e4589324380ad9f66873a4da5746fc0f46aaeb1f6", "submitted_at": "2026-04-28T22:15:06Z", "updated_at": "2026-08-11T15:48:53Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.26148v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2604.26148", "identifiers": {"arxiv": "2604.26148"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.26148"}], "name": "Beyond Screenshots: Evaluating VLMs' Understanding of UI Animations", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-04-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "AniMINT evaluates how VLMs perceive, explain and interpret UI animations using densely annotated videos. The study distinguishes basic motion recognition from higher-level animation meaning and probes motion, context and perceptual cues.", "topics": ["evaluation", "screen-understanding", "video", "ui-design"], "updated_at": "2026-08-11", "url": "https://arxiv.org/abs/2604.26148"}
{"aliases": [], "authors": ["Fazle Elahi Faisal", "Qianhui Wu", "Baolin Peng", "Jianfeng Gao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "276d8a62192c0ad576a1d986bb27d3978963b93466a822858e3313957dd83a81", "submitted_at": "2026-04-29T22:57:35Z", "updated_at": "2026-04-29T22:57:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.27253v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.27253", "identifiers": {"arxiv": "2604.27253"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.27253"}], "name": "AutoSurfer -- Teaching Web Agents through Comprehensive Surfing, Learning, and Modeling", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "AutoSurfer explores website actions breadth-first, retaining page and trajectory knowledge to guide task synthesis and trajectory refinement. It evaluates the resulting training data by fine-tuning a web agent on WebArena and examining synthesized-task diversity.", "topics": ["web-navigation", "data-generation", "imitation-learning", "evaluation"], "updated_at": "2026-04-29", "url": "https://arxiv.org/abs/2604.27253"}
{"aliases": [], "authors": ["Jinchao Li", "Yunxin Li", "Chenrui Zhao", "Zhenran Xu", "Baotian Hu", "Min Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2120860b9c735466a1538ad26c7cda0bc436ab0c6204febdefa08ad66f0036b7", "submitted_at": "2026-04-30T12:13:27Z", "updated_at": "2026-04-30T12:13:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.27776v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.27776", "identifiers": {"arxiv": "2604.27776"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.27776"}], "name": "WindowsWorld: A Process-Centric Benchmark of Autonomous GUI Agents in Professional Cross-Application Environments", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. Desktop use is explicit; no additional operating-system claim is inferred solely from the benchmark name.", "published_at": "2026-04-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WindowsWorld for professional workflows spanning multiple desktop applications, with generated tasks refined through human review. Intermediate subgoals support process-oriented evaluation of cross-application reasoning and execution efficiency.", "topics": ["evaluation", "cross-app", "workflow"], "updated_at": "2026-04-30", "url": "https://arxiv.org/abs/2604.27776"}
{"aliases": [], "authors": ["Jackson Vonderhorst", "Kuangshi Ai", "Haichao Miao", "Shusen Liu", "Chaoli Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8e7b07a811df7e74a28f91a5b6c59c69b7ba066bd6c64452a299e678a3187733", "submitted_at": "2026-04-30T15:22:28Z", "updated_at": "2026-06-26T20:31:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.27996v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2604.27996", "identifiers": {"arxiv": "2604.27996"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.27996"}], "name": "Exploring LLM Agent Designs and Interaction Modalities for Scientific Visualization", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-04-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study compares specialized, computer-use and coding agents on scientific visualization workflows. It examines GUI, CLI, code and tool-call interaction modes, including how persistent memory changes performance across repeated trials.", "topics": ["science", "evaluation", "tool-integration", "memory"], "updated_at": "2026-06-26", "url": "https://arxiv.org/abs/2604.27996"}
{"aliases": [], "authors": ["Michito Takeshita", "Takuro Kawada", "Takumi Ohashi", "Shunsuke Kitada", "Hitoshi Iyatomi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "87ef518590e29328ce752791df974c9e2215eac656fcf43fed97dd3e847f3cf7", "submitted_at": "2026-05-01T10:16:00Z", "updated_at": "2026-05-01T10:16:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.00551v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.00551", "identifiers": {"arxiv": "2605.00551"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.00551"}], "name": "A11y-Compressor: A Framework for Enhancing the Efficiency of GUI Agent Observations through Visual Context Reconstruction and Redundancy Reduction", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "A11y-Compressor transforms linearized accessibility trees through modal detection, redundancy reduction and semantic structuring. Its OSWorld evaluation measures both observation-token reduction and task success.", "topics": ["screen-understanding", "efficiency", "grounding", "evaluation"], "updated_at": "2026-05-01", "url": "https://arxiv.org/abs/2605.00551"}
{"aliases": [], "authors": ["Sohom Datta", "Alex Nahapetyan", "William Enck", "Alexandros Kapravelos"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0ff44befcfdb97b67d1680c97784f9e9665d8d9879c093984f42d06c9516a342", "submitted_at": "2026-05-06T23:19:38Z", "updated_at": "2026-05-06T23:19:38Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.05509v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2605.05509", "identifiers": {"arxiv": "2605.05509"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.05509"}], "name": "WAAA! Web Adversaries Against Agentic Browsers", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-05-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates agentic browsers against classical web attacks and visual social-engineering lures beyond indirect prompt injection.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2026-05-06", "url": "https://arxiv.org/abs/2605.05509"}
{"aliases": [], "authors": ["Yilin Zhang", "Yingkai Hua", "Chunyu Wei", "Xin Wang", "Yueguo Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b57d784c70b50d2c25055c6d139c0a5acfc97da3a4effbd63536b72d6df38a69", "submitted_at": "2026-05-10T12:11:52Z", "updated_at": "2026-05-10T12:11:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.09497v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.09497", "identifiers": {"arxiv": "2605.09497"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.09497"}], "name": "Don't Click That: Teaching Web Agents to Resist Deceptive Interfaces", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "DUDE combines asymmetric hybrid-reward learning with summaries of deceptive-interface failures to guide safer web actions. Its RUC benchmark evaluates deception susceptibility alongside ordinary task performance across interface scenarios.", "topics": ["safety-security", "reinforcement-learning", "evaluation", "robustness"], "updated_at": "2026-05-10", "url": "https://arxiv.org/abs/2605.09497"}
{"aliases": [], "authors": ["Weikai Xu", "Kun Huang", "Yunren Feng", "Jiaxing Li", "Yuhan Chen", "Yuxuan Liu", "Zhizheng Jiang", "Heng Qu", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Xiaolin Hu", "Bo An"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4374ec1b7c44804bb78ea56bd0aa3214083419f37780d79c5fcfb68468727bbf", "submitted_at": "2026-05-11T10:49:31Z", "updated_at": "2026-05-22T05:43:30Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.10347v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.10347", "identifiers": {"arxiv": "2605.10347"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.10347"}], "name": "How Mobile World Model Guides GUI Agents?", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-05-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares mobile world models that predict state changes as delta text, full text, images or renderable code. Downstream Android benchmarks examine synthetic training trajectories and test-time guidance, including limits of post-action reflection.", "topics": ["world-models", "data-generation", "evaluation"], "updated_at": "2026-05-22", "url": "https://arxiv.org/abs/2605.10347"}
{"aliases": [], "authors": ["Chiyeong Heo", "Jaechang Kim", "Junhyuk Kwon", "Hoyoung Kim", "Dongmin Park", "Jonghyun Lee", "Jungseul Ok"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0bd75e6bf9d5a8185a2204cdd99d42f9d96c102f7297abcadb9f8425e9246349", "submitted_at": "2026-05-08T10:57:19Z", "updated_at": "2026-05-08T10:57:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.10966v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2605.10966", "identifiers": {"arxiv": "2605.10966"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.10966"}], "name": "MMTB: Evaluating Terminal Agents on Multimedia-File Tasks", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent computer-automation benchmark using terminal workflows, explicitly separated from graphical-interface control.", "published_at": "2026-05-08", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "MultiMedia-TerminalBench evaluates agents that use terminal tools to work with audio and video files. The Terminus-MM harness adds multimedia perception to study how accessible evidence changes executable workflows and task outcomes.", "topics": ["evaluation", "automation", "video", "tool-integration"], "updated_at": "2026-05-08", "url": "https://arxiv.org/abs/2605.10966"}
{"aliases": [], "authors": ["Di Wu", "Zixiang Ji", "Asmi Kawatkar", "Bryan Kwan", "Jia-Chen Gu", "Nanyun Peng", "Kai-Wei Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5893d1d9205776667a1e1d89720ec3baf55fafe80cc9c6201ca97088f44a10ac", "submitted_at": "2026-05-12T17:59:34Z", "updated_at": "2026-05-12T17:59:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.12493v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}], "id": "arxiv:2605.12493", "identifiers": {"arxiv": "2605.12493"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:23Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.12493"}], "name": "LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-05-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "LongMemEval-V2 asks memory systems to recover environment knowledge from long web-agent histories for question answering. Its AgentRunbook variants contrast indexed knowledge pools with coding-agent retrieval from stored trajectory files.", "topics": ["memory", "interaction-history", "state-tracking", "workflow", "evaluation"], "updated_at": "2026-05-12", "url": "https://arxiv.org/abs/2605.12493"}
{"aliases": [], "authors": ["Miaosen Zhang", "Xiaohan Zhao", "Zhihong Tan", "Zhou Huoshen", "Yijia Fan", "Yifan Yang", "Kai Qiu", "Bei Liu", "Justin Wagle", "Chenzhong Yin", "Mingxi Cheng", "Ji Li", "Qi Dai", "Chong Luo", "Xu Yang", "Xin Geng", "Baining Guo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1fb57ba3cd858513d1da58d64a7c85388f12e331c9ea9f89204be5f43e6fec38", "submitted_at": "2026-05-12T17:59:58Z", "updated_at": "2026-05-12T17:59:58Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.12501v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.12501", "identifiers": {"arxiv": "2605.12501"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.12501"}], "name": "Covering Human Action Space for Computer Use: Data Synthesis and Benchmark", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CUActSpot to evaluate clicks, drags, drawing and other screen interactions across GUI, text, tables, canvases and natural images. A renderer-based pipeline records geometry and screenshots, then generates matching instructions and action traces for training a grounding model.", "topics": ["grounding", "data-generation", "evaluation", "action-prediction"], "updated_at": "2026-05-12", "url": "https://arxiv.org/abs/2605.12501"}
{"aliases": [], "authors": ["Jiaping Lin", "Fei Shen", "Junzhe Li", "Ping Nie", "Fei Yu", "Ming Li", "Haizhou Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cb60b76fca3a8825030b1a31a7b97ff31e06f30ac3852f243efe440f965e5578", "submitted_at": "2026-05-10T07:04:07Z", "updated_at": "2026-05-10T07:04:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.12549v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.12549", "identifiers": {"arxiv": "2605.12549"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.12549"}], "name": "What Happens Before Decoding? Prefill Determines GUI Grounding in VLMs", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-05-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Re-Prefill adds an attention-guided second prefill pass that revisits candidate screen regions before coordinate generation. The study evaluates this training-free intervention across GUI-grounding models and benchmarks.", "topics": ["grounding", "training-free", "screen-understanding", "evaluation"], "updated_at": "2026-05-10", "url": "https://arxiv.org/abs/2605.12549"}
{"aliases": [], "authors": ["Kean Shi", "Zihang Li", "Tianyi Ma", "Zengji Tu", "Jialong Wu", "Xinbo Xu", "Qingyao Yang", "Ruoyu Wu", "Weichu Xie", "Ming Wu", "Jason Zeng", "Michael Heinrich", "Elvis Zhang", "Liang Chen", "Kuan Li", "Baobao Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c4ee2439dd31a375b6f64cbc7f1d0ff75c87e1346b1627a53db145ae42682f8a", "submitted_at": "2026-05-15T09:35:15Z", "updated_at": "2026-05-24T16:35:56Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.15777v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "topics"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "8cfab6fca80090adb47c4ef634f89eedd49a6204dbdc2b5b8e0d6f04c32db289", "supports": ["summary", "environments", "scope", "notes"], "url": "https://raw.githubusercontent.com/UniPat-AI/SaaS-Bench/HEAD/README.md"}], "id": "arxiv:2605.15777", "identifiers": {"arxiv": "2605.15777"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.15777"}], "name": "SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows?", "notes": "The complete retained official abstract and cached official README were reviewed. The README establishes browser execution against running SaaS applications; no repository code was executed and benchmark results were not independently reproduced.", "published_at": "2026-05-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "SaaS-Bench evaluates browser agents on professional workflows across self-hosted applications. Weighted checkpoints measure task completion and partial progress, with application-state checks and configured content judges documented in the official harness.", "topics": ["web-navigation", "evaluation", "long-horizon", "workflow", "cross-app", "state-tracking"], "updated_at": "2026-05-24", "url": "https://arxiv.org/abs/2605.15777"}
{"aliases": [], "authors": ["Haoren Zhao", "Tianyi Chen", "Zhen Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "671f33dd87a7ebf9233494ef6fa0efd9c9772a3951f3c218bf4de822825f7af2", "submitted_at": "2026-05-13T02:48:52Z", "updated_at": "2026-05-13T02:48:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.16402v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.16402", "identifiers": {"arxiv": "2605.16402"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.16402"}], "name": "WinDeskGround: A Benchmark for Robust GUI Grounding in Complex Multi-Window Desktop Environments", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "WinDeskGround synthesizes multi-window desktop layouts with controlled occlusion, density and semantic similarity to test GUI grounding robustness. Its instruction-target benchmark measures how localization changes when interfaces depart from simple single-layer scenes.", "topics": ["grounding", "evaluation", "robustness", "data-generation"], "updated_at": "2026-05-13", "url": "https://arxiv.org/abs/2605.16402"}
{"aliases": [], "authors": ["Shilong Jin", "Lanjun Wang", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "52c1bf4c614ac05a55f2c2c66d3ac4eeb1bf31743814a45eb9f0ce064b6baaab", "submitted_at": "2026-05-16T08:51:57Z", "updated_at": "2026-05-16T08:51:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.16883v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.16883", "identifiers": {"arxiv": "2605.16883"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.16883"}], "name": "SE-GA: Memory-Augmented Self-Evolution for GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-05-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "SE-GA retrieves hierarchical memories during inference and uses the collected experience in a separate policy-improvement pipeline. Its evaluation includes offline grounding/action tasks and online Android interaction, which measure different capabilities.", "topics": ["memory", "interaction-history", "mobile-navigation", "grounding", "evaluation"], "updated_at": "2026-05-16", "url": "https://arxiv.org/abs/2605.16883"}
{"aliases": [], "authors": ["Athar Parvez", "Muhammad Jawad Mufti", "Muqaddas Gull", "Omar Hammad"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ff95e0b071508592190ba236741c1c01200dd38fdc74c31b2cc4d1bf8143ae9c", "submitted_at": "2026-05-17T21:14:32Z", "updated_at": "2026-05-17T21:14:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.17656v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}], "id": "arxiv:2605.17656", "identifiers": {"arxiv": "2605.17656"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:23Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.17656"}], "name": "MUIAnno: An Expert-Annotated Dataset and Evaluation Benchmark for Mobile UI Understanding", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-05-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "MUIAnno provides expert labels for elements in manually explored mobile-app screens, with a web annotation tool that emits structured JSON. It also supplies UI-element detection baselines for the dataset.", "topics": ["screen-understanding", "grounding", "evaluation", "data-generation"], "updated_at": "2026-05-17", "url": "https://arxiv.org/abs/2605.17656"}
{"aliases": [], "authors": ["Jingjing Liu", "Ziye Huang", "Zihao Cheng", "Zeming Liu", "Jiahong Wu", "Yuhang Guo", "Kehai Chen", "Yunhong Wang", "Haifeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "017909a697c6c661d3486d960042f4d691f70a660a60083e70512de34690cb51", "submitted_at": "2026-05-18T08:36:04Z", "updated_at": "2026-05-18T08:36:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.18048v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.18048", "identifiers": {"arxiv": "2605.18048"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.18048"}], "name": "DocOS: Towards Proactive Document-Guided Actions in GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "DocOS benchmarks agents that search online documentation and translate procedural instructions into GUI actions in interactive web environments. It separates failures to retrieve useful procedures from failures to ground retrieved instructions into executable actions.", "topics": ["web-navigation", "grounding", "evaluation", "planning"], "updated_at": "2026-05-18", "url": "https://arxiv.org/abs/2605.18048"}
{"aliases": [], "authors": ["Ziyun Zeng", "Hang Hua", "Bocheng Zou", "Mu Cai", "Rogerio Feris", "Jiebo Luo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d96c8844274f79adea167169a0853e3a7cd82fee4a8a666913f97898a0836e7c", "submitted_at": "2026-05-18T16:57:36Z", "updated_at": "2026-05-18T16:57:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.18652v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.18652", "identifiers": {"arxiv": "2605.18652"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.18652"}], "name": "MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MementoGUI learns a controller for selecting, compressing and retrieving textual and region-level visual memories while leaving the GUI-agent backbone fixed. It also converts computer-use traces into controller training data and evaluates semantic actions, progress and memory consistency on long-horizon tasks.", "topics": ["memory", "interaction-history", "long-horizon", "data-generation", "evaluation"], "updated_at": "2026-05-18", "url": "https://arxiv.org/abs/2605.18652"}
{"aliases": [], "authors": ["Felix Henry", "Lihan Lin", "Jiangyou Zhu", "Yangfan", "Bingqian Zhang", "Min Chen", "Shiyu Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "99e8e0039ec99ef8f6a6ef3900cb74a62ddd5ae4f8df30629110969290028e6a", "submitted_at": "2026-04-03T08:57:06Z", "updated_at": "2026-09-28T02:42:06Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.18758v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.18758", "identifiers": {"arxiv": "2605.18758"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.18758"}], "name": "OmniGUI: Benchmarking GUI Agents in Omni-Modal Smartphone Environments", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "OmniGUI evaluates step-level action prediction from interleaved smartphone screenshots, synchronous audio and video. Expert demonstrations carry multimodal-dependency annotations, and initial baselines use native multimodal models as agent proxies to study temporal and audio-related failures.", "topics": ["evaluation", "action-prediction", "video", "mobile-navigation"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2605.18758"}
{"aliases": [], "authors": ["Jinbiao Wei", "Qianran Ma", "Yilun Zhao", "Xiao Zhou", "Kangqi Ni", "Guo Gan", "Arman Cohan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0a93d880793018d1b47a39e0487e3be76550fd2b49de9535b2913e9f08159074", "submitted_at": "2026-05-19T12:40:29Z", "updated_at": "2026-05-19T12:40:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.19769v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.19769", "identifiers": {"arxiv": "2605.19769"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.19769"}], "name": "OpenComputer: Verifiable Software Worlds for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-05-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OpenComputer, which builds desktop tasks around application-state verifiers and refines those verifiers using execution feedback. An evaluation harness records trajectories and assigns auditable partial credit across professional software workflows.", "topics": ["evaluation", "data-generation", "observability"], "updated_at": "2026-05-19", "url": "https://arxiv.org/abs/2605.19769"}
{"aliases": [], "authors": ["Guohong Liu", "Jialei Ye", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Yunxin Liu", "Yuanchun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "eb40c8bf005d56139d263e1b5dc4553ddcd68522124686b7f0865e6d8d1725ad", "submitted_at": "2026-05-24T16:33:14Z", "updated_at": "2026-06-18T02:24:07Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.25160v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.25160", "identifiers": {"arxiv": "2605.25160"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.25160"}], "name": "ScaleWoB: Guiding GUI Agents with Coding Agents via Large-Scale Environmental Synthesis", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. Mobile, desktop and automotive interfaces are simulated as webpages; this does not assert native deployment on those platforms.", "published_at": "2026-05-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces ScaleWoB, which synthesizes interactive GUI environments as backend-free webpages with verifiable rewards. Its mobile-oriented benchmark evaluates agent execution and compares selected synthetic tasks with real-app behavior.", "topics": ["data-generation", "evaluation", "sandbox"], "updated_at": "2026-06-18", "url": "https://arxiv.org/abs/2605.25160"}
{"aliases": [], "authors": ["Yusong Lin", "Xinyuan Liang", "Haiyang Wang", "Qipeng Gu", "Siqi Cheng", "Jiangui Chen", "Shuzhe Wu", "Feiyang Pan", "Lue Fan", "Sanyuan Zhao", "Dandan Tu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "27999553928305cfa4223011feb3f9c2decda8a861e5800865537a81c7c08832", "submitted_at": "2026-05-25T17:50:04Z", "updated_at": "2026-05-25T17:50:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.26086v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}], "id": "arxiv:2605.26086", "identifiers": {"arxiv": "2605.26086"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:23Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.26086"}], "name": "Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-05-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Claw-Anything benchmarks persistent personal assistance across activity histories, backend services and integrated GUI/CLI interaction. Its simulated event histories introduce irrelevant or conflicting context, and a separate generation pipeline supplies training environments.", "topics": ["evaluation", "interaction-history", "long-horizon", "robustness", "data-generation", "workflow"], "updated_at": "2026-05-25", "url": "https://arxiv.org/abs/2605.26086"}
{"aliases": [], "authors": ["Dingbang Wu", "Rui Hao", "Haiyang Wang", "Shuzhe Wu", "Han Xiao", "Zhenghong Li", "Bojiang Zhou", "Zheng Ju", "Zichen Liu", "Lue Fan", "Zhaoxiang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "145c7012fb15a85b203924ecd559c8b20d18b51bf79805a6751fec75d6f68dec", "submitted_at": "2026-05-25T17:59:49Z", "updated_at": "2026-09-23T17:59:57Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2605.26114v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.26114", "identifiers": {"arxiv": "2605.26114"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.26114"}], "name": "MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The simulator is browser-hosted; the abstract does not identify a native mobile operating system.", "published_at": "2026-05-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MobileGym, a browser-hosted simulation of mobile app interactions whose configurable JSON state supports deterministic judging and parallel rollouts. MobileGym-Bench adds parameterized tasks, with experiments examining online training and transfer to real devices.", "topics": ["sandbox", "evaluation", "reinforcement-learning"], "updated_at": "2026-09-23", "url": "https://arxiv.org/abs/2605.26114"}
{"aliases": [], "authors": ["JunJia Guo", "Yuhang Yao", "Chang Li", "Zimeng Pan", "Jiawei", "Zhou", "Jingdi Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "64f8e09fc62f080a75aec1928b82187063cfc3ad8467290608347622410b6460", "submitted_at": "2026-05-22T20:29:12Z", "updated_at": "2026-09-29T00:12:17Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2605.26144v4"}}, "environments": ["web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}], "id": "arxiv:2605.26144", "identifiers": {"arxiv": "2605.26144"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:23Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.26144"}], "name": "VISTA: An End-to-End Benchmark for Visual Spec-to-Web-App Coding Agents", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent GUI-application generation benchmark with real component interaction probes; its main evaluated agents are application coders, not general GUI task executors.", "published_at": "2026-05-22", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "VISTA evaluates coding agents that turn design handoffs into working web and Android applications. Its evaluator locates annotated interactive components and probes their behavior, keeping visual localization distinct from functional correctness.", "topics": ["evaluation", "ui-design", "automation"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2605.26144"}
{"aliases": [], "authors": ["Yifan Sui", "Xin Huang", "Hongbing Li", "Fang Xu", "Jiahe Lv", "Haolong Yan", "Yeqing Shen", "Litao Liu", "Zhimin Fan", "Ziyang Meng", "Jia Wang", "Junbo Qi", "Kaijun Tan", "Zheng Ge", "Xiangyu Zhang", "Daxin Jiang", "Osamu Yoshie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0c89b8b5ef5aa17fe747201285104544648449804cc2229a77577d217763f497", "submitted_at": "2026-05-26T23:19:42Z", "updated_at": "2026-05-26T23:19:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.27761v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.27761", "identifiers": {"arxiv": "2605.27761"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.27761"}], "name": "AndroidDaily: A Verifiable Benchmark for Mobile GUI Agents on Real-World Closed-Source Applications", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "AndroidDaily evaluates mobile GUI agents on daily workflows in closed-source Android apps. Its GRADE evaluator follows visual trajectories against operational obligations, output-quality requirements and negative constraints, producing process-level diagnostics without hidden app state.", "topics": ["mobile-navigation", "evaluation", "long-horizon", "robustness"], "updated_at": "2026-05-26", "url": "https://arxiv.org/abs/2605.27761"}
{"aliases": [], "authors": ["Ruoqi Guo", "Yi Liu", "Gelei Deng", "Yiheng Xiong", "Yuekang Li", "Ying Zhang", "Leo Yu Zhang", "Lida Zhao", "Ji Jie", "Yuxiao Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4f39433695b9574b275fc844ca6b10a43baf66aed57048bf841f41bf08d4ac9a", "submitted_at": "2026-05-27T08:09:53Z", "updated_at": "2026-05-27T08:09:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.28116v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.28116", "identifiers": {"arxiv": "2605.28116"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.28116"}], "name": "MIRAGE: Context-Aware Prompt Injection against Mobile GUI Agents via User-Generated Content", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "MIRAGE constructs mobile screenshot injections in ordinary user-content regions through localization, contextual payload generation and realism curation. Its evaluation separates visual realism from attack effectiveness against pixel-observing GUI agents.", "topics": ["safety-security", "data-generation", "evaluation", "robustness"], "updated_at": "2026-05-27", "url": "https://arxiv.org/abs/2605.28116"}
{"aliases": [], "authors": ["Yixu Huang", "Bo Li", "Na Li", "Zhe Wang", "Kaijie Chen", "Haonan Ge", "Qingyi Si", "Yuanzhe Shen", "Ruihan Yang", "Guangjing Wang", "Hongcheng Guo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4e03e0ef06c6610eaefeec08b7891a1ab4d5073c0793650d4845735259978723", "submitted_at": "2026-05-27T10:08:48Z", "updated_at": "2026-10-02T05:26:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.28258v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.28258", "identifiers": {"arxiv": "2605.28258"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.28258"}], "name": "GUI Agents for Continual Game Generation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced. GUI agents actually load and play browser-game builds, providing an executable interface-testing signal for game-code generation; the central application remains game development.", "published_at": "2026-05-27", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "PlaytestArena evaluates generated browser games using expected in-play behaviors adjudicated by GUI agents. Play2Code pairs a coding agent with a rubric-blind GUI playtester whose interaction traces and feedback guide iterative refinement, with a separate final judge.", "topics": ["evaluation", "automation", "workflow"], "updated_at": "2026-10-02", "url": "https://arxiv.org/abs/2605.28258"}
{"aliases": [], "authors": ["Yanqiu Zhao", "Dongying Zheng", "Kaibo Huang", "Yukun Wei", "Zhongliang Yang", "Linna Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6eb9b67f7e51edecfbe57af47b6bad445c05129c11c377d32ea84369b430f895", "submitted_at": "2026-05-27T15:51:22Z", "updated_at": "2026-05-29T04:03:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.28646v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.28646", "identifiers": {"arxiv": "2605.28646"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.28646"}], "name": "MaskClaw: Edge-Side Personalized Privacy Arbitration for GUI Agents with Behavior-Driven Skill Evolution", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "MaskClaw makes local Allow, Mask or Ask decisions before GUI screenshots leave a trusted environment, using visual evidence and personalized policy memory. Its benchmark and designed evolution scenarios test how corrections and edits can become privacy skills checked by a sandbox gate.", "topics": ["safety-security", "memory", "human-in-the-loop", "evaluation"], "updated_at": "2026-05-29", "url": "https://arxiv.org/abs/2605.28646"}
{"aliases": [], "authors": ["Yuxuan Liu", "Xin Lai", "Junyi Li", "Pengyuan Lyu", "Jason", "Yiduo Guo", "Zhengyao Fang", "Yang Ding", "Yi Zhang", "Weinong Wang", "Huawen Shen", "Xingran Zhou", "Liang Wu", "Fei Tang", "Sunqi Fan", "Shangpin Peng", "Zheng Ruan", "Anran Zhang", "Chengquan Zhang", "Han Hu", "Benyou Wang", "Ji-Rong Wen", "Rui Yan", "Zhengyang Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a41cb1a93e6aa6ba8fbc7dca55415577ab5880a6592e9b0bee5b12b5a7732cbf", "submitted_at": "2026-05-28T07:14:15Z", "updated_at": "2026-09-28T15:27:12Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2605.29486v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}], "id": "arxiv:2605.29486", "identifiers": {"arxiv": "2605.29486"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.29486"}], "name": "PhoneWorld: From Real-App Trajectories to Dynamic and Verifiable Environments for Phone-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-05-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces PhoneWorld, which converts observed app pages, transitions and state changes into executable Android environments through a build-inspect-repair process. Generated tasks and programmatic verifiers support benchmarking, trajectory collection and online training.", "topics": ["data-generation", "sandbox", "reinforcement-learning", "evaluation"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2605.29486"}
{"aliases": [], "authors": ["Wenkai Wang", "Tao Xiong", "Jingchen Ni", "Yunpeng Bao", "Xiyun Li", "Tianqi Liu", "Hongcan Guo", "Zilong Huang", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf0fcafeb2d40e489ec80784a3541114b02c6a9077a330265d5c49775eca4582", "submitted_at": "2026-06-02T03:42:34Z", "updated_at": "2026-06-02T03:42:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.03103v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}], "id": "arxiv:2606.03103", "identifiers": {"arxiv": "2606.03103"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:23Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.03103"}], "name": "DeskCraft: Benchmarking Desktop Agents on Professional Workflows and Human-in-the-Loop Collaboration", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-06-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "DeskCraft evaluates professional desktop workflows and human-agent collaboration through mid-execution clarification, interruption and post-completion feedback. Its tasks cover creative applications and expose difficulties in long-horizon execution and proactive clarification.", "topics": ["evaluation", "human-in-the-loop", "long-horizon", "workflow"], "updated_at": "2026-06-02", "url": "https://arxiv.org/abs/2606.03103"}
{"aliases": [], "authors": ["Jia Yu", "Zilong Wang", "Xinyang Jiang", "Dongsheng Li", "Shuo Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b23c12cafdd2d4a3ef63027c914ed3894c61c555edd89468c56ae40c3ec6cea9", "submitted_at": "2026-06-02T06:02:35Z", "updated_at": "2026-06-02T06:02:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.03203v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.03203", "identifiers": {"arxiv": "2606.03203"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.03203"}], "name": "MedCUA-Bench: A Screenshot-Only Benchmark for Clinical Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "MedCUA-Bench evaluates screenshot-only clinical interface workflows using paired intent- and step-level goals. It combines reconstructed medical interfaces with completion and clinical-safety checks, and separately reports results on a real OpenEMR system.", "topics": ["evaluation", "safety-security", "science", "workflow"], "updated_at": "2026-06-02", "url": "https://arxiv.org/abs/2606.03203"}
{"aliases": [], "authors": ["Jiashu Yao", "Heyan Huang", "Daiqing Wu", "Wangke Chen", "Huaxi Ai", "Haoyu Wen", "Zeming Liu", "Yuhang Guo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4d5a90333c4f6be0b8d203a2f5938f7e7eefcb09fbe7ea47b85eba8c7c1c422a", "submitted_at": "2026-06-03T10:25:46Z", "updated_at": "2026-06-03T10:25:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.04701v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.04701", "identifiers": {"arxiv": "2606.04701"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.04701"}], "name": "Benchmarking Living-Screen-Native GUI Agents on Short-Video Platforms", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "LivingScreen evaluates agents in a browser-based short-video environment where observing content consumes time and affects subsequent actions. Its task suite scores accuracy together with information efficiency to diagnose over- and under-observation on continuously changing screens.", "topics": ["evaluation", "video", "efficiency", "screen-understanding"], "updated_at": "2026-06-03", "url": "https://arxiv.org/abs/2606.04701"}
{"aliases": [], "authors": ["Nicholas Saban"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7b649c0cef63bc15558aa4498a65043644c109faaf6403b610c2132fd3f5706d", "submitted_at": "2026-06-03T01:21:59Z", "updated_at": "2026-06-03T01:21:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.05233v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.05233", "identifiers": {"arxiv": "2606.05233"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.05233"}], "name": "Domain-Conditioned Safety in Frontier Computer-Using Agents: A 793-Episode Browser Benchmark, a Coding-Domain Cross-Reference, and a Reproducibility Audit of Recent Red-Teaming", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CUA-HandCrafted, a browser-agent red-teaming benchmark that varies attack templates, tasks and system prompts. Its domain comparison and reproducibility audit caution against transferring browser-surface safety conclusions to coding agents or unreleased optimized injections.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-06-03", "url": "https://arxiv.org/abs/2606.05233"}
{"aliases": [], "authors": ["Nathan Bout", "Maxime Langevin", "Ronan Riochet"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "27ffcb10d966946de9fa2353c4b96798e5f090bdfdc5610f5e8a5c6233cf70ea", "submitted_at": "2026-06-04T15:57:29Z", "updated_at": "2026-06-04T15:57:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.06322v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}], "id": "arxiv:2606.06322", "identifiers": {"arxiv": "2606.06322"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.06322"}], "name": "DragOn: A Benchmark and Dataset for Drag-Based GUI Interactions", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "DragOn supplies drag-grounding training data and a held-out benchmark for highlighting text, selecting cells, resizing elements and moving sliders. It evaluates existing vision-language models and a model fine-tuned on the drag tasks.", "topics": ["grounding", "data-generation", "evaluation", "screen-understanding"], "updated_at": "2026-06-04", "url": "https://arxiv.org/abs/2606.06322"}
{"aliases": [], "authors": ["Victor Muryn", "Maksym Shamrai", "Sofiia Mazepa", "Yehor Khodysko"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fd66adb4e99df2348f59e238d219d48cb6e758e694839367d2b9a64675429d87", "submitted_at": "2026-06-04T14:01:32Z", "updated_at": "2026-06-04T14:01:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.06560v1"}}, "environments": ["desktop", "macos"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.06560", "identifiers": {"arxiv": "2606.06560"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.06560"}], "name": "MacArena: Benchmarking Computer Use Agents on an Online macOS Environment", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "MacArena runs computer-use tasks through Apple’s virtualization framework on Apple Silicon, combining task ports with macOS-native workloads. It distinguishes the ported and native subsets when examining agent performance and cross-platform transfer.", "topics": ["evaluation", "robustness"], "updated_at": "2026-06-04", "url": "https://arxiv.org/abs/2606.06560"}
{"aliases": [], "authors": ["Haojie Hao", "Longkun Hao", "Yihang Lou", "Yan Bai", "Zhenyang Li", "Zhichao Yang", "Dongshuo Huang", "Hongyu Lin", "Lanqing Hong", "Jiakai Wang", "Xianglong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2d530152b210a69cb458fecb036c10621c72daccfbcac7c8306c7e248c87cf14", "submitted_at": "2026-06-05T08:17:28Z", "updated_at": "2026-06-12T07:28:19Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.07027v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.07027", "identifiers": {"arxiv": "2606.07027"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.07027"}], "name": "StainFlow: Entity-Stain Tracking and Evidence Linking for Process Rewards in GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "StainFlow tracks visually verifiable task entities through GUI trajectories to identify phases and retrieve focused evidence windows. Its process-reward model is tested for both online policy learning and trajectory-completion judgments.", "topics": ["reward-models", "state-tracking", "reinforcement-learning", "evaluation"], "updated_at": "2026-06-12", "url": "https://arxiv.org/abs/2606.07027"}
{"aliases": [], "authors": ["Wanli Li", "Bowen Zhou", "Yunyao Yu", "Zhou Xu", "Yifan Yang", "Dongsheng Li", "Caihua Shan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73b751efd63d98fc81166a1a80cb0c162d68bf1ec4ecca0a1a1a7717ad62e680", "submitted_at": "2026-06-08T12:39:23Z", "updated_at": "2026-07-06T14:25:02Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2606.09426v3"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.09426", "identifiers": {"arxiv": "2606.09426"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.09426"}], "name": "WeaveBench: A Long-Horizon, Real-World Benchmark for Computer-Use Agents with Hybrid Interfaces", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "WeaveBench requires agents to combine GUI, command-line and code operations within long tasks on an Ubuntu desktop. A trajectory-aware judge checks deliverables and interaction evidence, including shortcut behavior that outcome-only grading can miss.", "topics": ["evaluation", "long-horizon", "workflow", "tool-integration"], "updated_at": "2026-07-06", "url": "https://arxiv.org/abs/2606.09426"}
{"aliases": [], "authors": ["Bojie Rong", "Zheyu Shen", "Qiaoping Wang", "Pengfei Kang", "Yang Xu", "Yawen Wei", "Hanyu Wu", "Zhi Zhao", "Leihao Pei", "Linquan Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0a852c77650fa54f234794c6cd27777f4e7219b8389b509d807e27db16138a1b", "submitted_at": "2026-06-08T12:55:42Z", "updated_at": "2026-06-08T12:55:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.09447v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.09447", "identifiers": {"arxiv": "2606.09447"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.09447"}], "name": "AliyunConsoleAgent: Training Web Agents in Real-World Cloud Environments via Distillation and Reinforcement Learning", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "AliyunConsoleAgent trains browser agents to verify documented procedures against changing cloud-console UIs through trajectory distillation and online GRPO. Provisioned rollout environments and backend audit-log checks support training and outcome evaluation.", "topics": ["web-navigation", "imitation-learning", "reinforcement-learning", "evaluation"], "updated_at": "2026-06-08", "url": "https://arxiv.org/abs/2606.09447"}
{"aliases": [], "authors": ["Lawrence Keunho Jang", "Mareks Woodside", "Geronimo Carom", "Andrew Keunwoo Jang", "Jing Yu Koh", "Ruslan Salakhutdinov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "495420bd87ba06ae2ec1763865b1145271e964b4bd40145c8e591653a1581c42", "submitted_at": "2026-06-08T17:27:13Z", "updated_at": "2026-06-08T17:27:13Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.09764v1"}}, "environments": ["ios"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}], "id": "arxiv:2606.09764", "identifiers": {"arxiv": "2606.09764"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:02Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.09764"}], "name": "iOSWorld: A Benchmark for Personally Intelligent Phone Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a native iOS simulator benchmark with persistent user data, multi-app tasks, and personalization-dependent phone workflows.", "topics": ["evaluation", "cross-app", "memory"], "updated_at": "2026-06-08", "url": "https://arxiv.org/abs/2606.09764"}
{"aliases": [], "authors": ["Liya Zhu", "Jingzhe Ding", "Jian Zhang", "Jianbo Xue", "Shihao Liang", "Ge Zhang", "Yi Zhu", "Duju Zeng", "Xiang Gao", "Qingshui Gu", "Mailun Gao", "Huimin Che", "Yan Zhao", "Peiheng Zhou", "Haojun Wang", "Chaobo Xian", "Lili Le", "Chi Wu", "Yiwei Liu", "Shengda Long", "Jiale Yang", "Fangzhi Xu", "Sijin Wu", "Haodong Duan", "Chao He", "Zhaojian Li", "Minchao Wang", "Huan Zhou", "Jiani Hou", "Chuqian Yu", "Weiran Shi", "Hongwan Gao", "Jiamin Chen", "Guanhong Chen", "Tingqin Luo", "Kaiyuan Zhang", "Zhixin Yao", "Qing Hua", "Yuhao Jiang", "Jin Chen", "Pu Chen", "Zhenyu Hu", "Xingyu Li", "Zhengxuan Jiang", "Meng Cao", "Tianfeng Long", "Haozhe Wang", "Mingzhang Wang", "Yichen Zhang", "Yiming Dai", "Chenchen Zhang", "Jiaying Wang", "Xinying Liu", "Xingzu Liu", "Lingling Zhang", "Xinjie Chen", "Yujia Qin", "Wangchunshu Zhou", "Zhiyong Wu", "Yang Liu", "Jiaheng Liu", "Lei Zhang", "Shen Yan", "Wenhao Huang", "Zaiyuan Wang", "Xiaolong Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "037a3197137d27318320a3fd23fb5e606a1aa748133e57213cd02fd10aad4c4d", "submitted_at": "2026-06-09T16:10:16Z", "updated_at": "2026-07-17T12:22:12Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2606.11042v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.11042", "identifiers": {"arxiv": "2606.11042"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.11042"}], "name": "Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional Fields", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Workflow-GYM evaluates long GUI workflows in professional software rather than isolated interface actions. Its analysis examines failures such as omitted stages, accumulated errors and goal drift during specialized end-to-end work.", "topics": ["evaluation", "long-horizon", "workflow", "robustness"], "updated_at": "2026-07-17", "url": "https://arxiv.org/abs/2606.11042"}
{"aliases": [], "authors": ["Jaewoo Lee", "Zaid Khan", "Archiki Prasad", "Justin Chih-Yao Chen", "Supriyo Chakraborty", "Kartik Balasubramaniam", "Sambit Sahu", "Elias Stengel-Eskin", "Hyunji Lee", "Mohit Bansal"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a8ed399c8b489a80c5313a00232d2ecca0fbc8228b1f56745c02d4fedb6f6908", "submitted_at": "2026-06-09T16:39:10Z", "updated_at": "2026-06-09T16:39:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.11078v1"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.11078", "identifiers": {"arxiv": "2606.11078"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.11078"}], "name": "A History-Aware Visually Grounded Critic for Computer Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "HiViG combines a compact history of completed actions with a visually grounded critic that checks proposed coordinates before execution. Its evaluations separate history-related planning failures from errors caused by targeting the wrong screen element.", "topics": ["interaction-history", "memory", "grounding", "reward-models", "long-horizon", "evaluation"], "updated_at": "2026-06-09", "url": "https://arxiv.org/abs/2606.11078"}
{"aliases": [], "authors": ["Jiaxin Ai", "Tao Hu", "Xuemeng Yang", "Shu Zou", "Hairong Zhang", "Daocheng Fu", "Yu Yang", "Hongbin Zhou", "Nianchen Deng", "Pinlong Cai", "Zhongyuan Wang", "Botian Shi", "Kaipeng Zhang", "Licheng Wen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "edf7e63803efba89551aa297bff1da9b0babf8ec898a2971fa6c8a20cc58dea8", "submitted_at": "2026-06-11T11:53:32Z", "updated_at": "2026-06-30T03:31:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.13239v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.13239", "identifiers": {"arxiv": "2606.13239"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.13239"}], "name": "ComAct: Reframing Professional Software Manipulation via COM-as-Action Paradigm", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-11", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "ComAct proposes COM program synthesis as an alternative interface for manipulating professional software and introduces ComCADBench with GUI-based comparison agents. ComActor adds a staged training and correction approach, supported by a Windows-container training platform.", "topics": ["automation", "evaluation", "tool-integration", "science"], "updated_at": "2026-06-30", "url": "https://arxiv.org/abs/2606.13239"}
{"aliases": [], "authors": ["Zihao Wang", "Yiming Li", "Yutong Wu", "Kangjie Chen", "Zheyu Liu", "Fok Kar Wai", "Pin-Yu Chen", "Vrizlynn L. L. Thing", "Bo Li", "Dacheng Tao", "Tianwei Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "88b7e2ee8717661e4f8b417916e35b844171ecd8a8a2add657cf341c8f759536", "submitted_at": "2026-06-11T14:12:43Z", "updated_at": "2026-07-27T14:19:24Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.13385v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.13385", "identifiers": {"arxiv": "2606.13385"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.13385"}], "name": "Who Pays the Price? Stakeholder-Centric Prompt Injection Benchmarking for Real-world Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "StakeBench evaluates prompt-injection harm to different stakeholders in shopping web-agent workflows. Its executable cases separate attack objectives and assess both outcomes and execution processes to reveal how the same attack pattern can affect different parties.", "topics": ["evaluation", "safety-security", "web-navigation"], "updated_at": "2026-07-27", "url": "https://arxiv.org/abs/2606.13385"}
{"aliases": [], "authors": ["Seoyoung Choi", "Minseok Ko", "Hyunseok Lee", "Kunwoong Kim", "Woomin Song", "Chanseok Jeon", "Jinwoo Shin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1fb87da27071ab21519599b7d16359a38118aa358f620de94e66b15cad040828", "submitted_at": "2026-06-12T04:39:01Z", "updated_at": "2026-06-12T04:39:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.14106v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.14106", "identifiers": {"arxiv": "2606.14106"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.14106"}], "name": "Naive Visual Memory is Not Enough: A Failure-Mode Study of GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study finds that full-image visual memory can reduce some state errors while worsening action-level failures. Its AGMem alternative stores local crops associated with successful actions or recovery and evaluates their use in desktop GUI tasks.", "topics": ["memory", "interaction-history", "grounding", "robustness", "evaluation"], "updated_at": "2026-06-12", "url": "https://arxiv.org/abs/2606.14106"}
{"aliases": [], "authors": ["Chenxin Li", "Zhengyao Fang", "Zhengyang Tang", "Pengyuan Lyu", "Xingran Zhou", "Xin Lai", "Fei Tang", "Liang Wu", "Yiduo Guo", "Weinong Wang", "Junyi Li", "Yi Zhang", "Yang Ding", "Huawen Shen", "Sunqi Fan", "Shangpin Peng", "Zheng Ruan", "Anran Zhang", "Benyou Wang", "Chengquan Zhang", "Han Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "280ec12bcb246f1f07cf3627011945e8a9ff4afed93b1fb2937f2d6b1bc1323b", "submitted_at": "2026-06-12T15:01:32Z", "updated_at": "2026-06-12T15:01:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.14832v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.14832", "identifiers": {"arxiv": "2606.14832"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.14832"}], "name": "PhoneHarness: Harnessing Phone-Use Agents through Mixed GUI, CLI, and Tool Actions", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "PhoneHarness routes phone workflows across GUI actions, device-side commands and host tools while retaining auditable execution traces. Its accompanying benchmark checks observable task side effects, separating mixed-action execution from plausible final-answer generation.", "topics": ["mobile-navigation", "tool-integration", "evaluation", "workflow"], "updated_at": "2026-06-12", "url": "https://arxiv.org/abs/2606.14832"}
{"aliases": [], "authors": ["Sina Hajimiri", "Masih Aminbeidokhti", "Jose Dolz", "Ismail Ben Ayed", "Issam H. Laradji", "Spandana Gella", "Nicolas Gontier"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "50f1acd5727f84946cf95d181f270dbc4f46f15d24339e6cb257fea1b23b21a0", "submitted_at": "2026-06-12T23:30:14Z", "updated_at": "2026-08-30T17:49:13Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.15017v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.15017", "identifiers": {"arxiv": "2606.15017"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.15017"}], "name": "Are Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares online skill and memory augmentations with a vanilla web agent given the same total inference-token budget. Across WebArena and WorkArena settings, it studies when module overhead removes aggregate gains and how rerun variability changes conclusions.", "topics": ["memory", "efficiency", "evaluation", "robustness"], "updated_at": "2026-08-30", "url": "https://arxiv.org/abs/2606.15017"}
{"aliases": [], "authors": ["Mina Mohammadmirzaei", "Jeffrey Flanigan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7bf0297fba942441b19bcfeba147838bdef888b428f1150c2462d8bf3d986882", "submitted_at": "2026-06-13T00:32:24Z", "updated_at": "2026-10-09T19:45:15Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2606.15034v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.15034", "identifiers": {"arxiv": "2606.15034"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.15034"}], "name": "OSGuard: A Benchmark for Safety in Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates action-level guardrails and full computer tasks with environment-dependent safety constraints, separating safe completion from nominal success.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2606.15034"}
{"aliases": [], "authors": ["Y. H. Zhou", "Z. M. Ma", "Y. J. Zhou", "Y. T. Li", "H. X. Xiang", "Y. M. Cheng", "T. L. Chen", "K. J. Zhang", "Z. H. Nan", "J. H. Ni", "Z. Wu", "Q. Y. Pan", "S. Zhang", "S. Cheng", "M. Y. Luo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6311b22ddb2262c74efbe79acc8856335f6870aa24197c507f3f7b048d6f7901", "submitted_at": "2026-06-15T12:53:55Z", "updated_at": "2026-06-15T12:53:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.16659v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.16659", "identifiers": {"arxiv": "2606.16659"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.16659"}], "name": "FraudSMSWalker: Benchmarking Agentic Large Language Models for SMS-to-Webpage Fraud Detection", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "FraudSMSWalker tests browser agents on bilingual SMS-to-webpage fraud judgments while hiding URL and reputation shortcuts. Its controlled chains include difficult benign login/payment cases and evaluate whether classifications remain supported by the visible message and page evidence.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-06-15", "url": "https://arxiv.org/abs/2606.16659"}
{"aliases": [], "authors": ["Lawrence Keunho Jang", "Andrew Keunwoo Jang", "Jing Yu Koh", "Ruslan Salakhutdinov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "acc529f96d6df52e3560f66df8e0630dabdae3ab30cadc8b59123f347eba187a", "submitted_at": "2026-06-15T14:08:09Z", "updated_at": "2026-06-15T14:08:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.16748v1"}}, "environments": ["desktop", "linux", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.16748", "identifiers": {"arxiv": "2606.16748"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.16748"}], "name": "MyPCBench: A Benchmark for Personally Intelligent Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "MyPCBench tests personal-assistant tasks on a Linux desktop with simulated web applications and data for a single fictional persona. A shared computer-and-shell interface enables evaluation of cross-application workflows that depend on personal context and account-like state.", "topics": ["evaluation", "cross-app", "workflow"], "updated_at": "2026-06-15", "url": "https://arxiv.org/abs/2606.16748"}
{"aliases": [], "authors": ["Anqi Zou", "Han Deng", "Chengyu Zhang", "Junquan Hu", "Yu Wang", "Yuxiang Xing", "Aokai Zhang", "Hanling Zhang", "Zhaoyang Liu", "Ben Fei", "Zhihui Wang", "Wanli Ouyang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7487f31097bc8de90dfbe836051d47f407c26e31a42bd751bedfd38cf4f1742c", "submitted_at": "2026-06-15T14:42:33Z", "updated_at": "2026-06-15T14:42:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.16802v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.16802", "identifiers": {"arxiv": "2606.16802"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.16802"}], "name": "LabOSBench: Benchmarking Computer Use Agents for Scientific Instrument Control", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "LabOSBench evaluates GUI agents on browser-based scientific instrument simulators with configurable tasks and execution-based checks. Its subtask and end-to-end tests examine feedback-driven adjustment and long workflows without controlling physical instruments.", "topics": ["evaluation", "science", "long-horizon"], "updated_at": "2026-06-15", "url": "https://arxiv.org/abs/2606.16802"}
{"aliases": [], "authors": ["Jaehun Jung", "Ximing Lu", "Brandon Cui", "Muhammad Khalifa", "Shaokun Zhang", "Hao Zhang", "Jin Xu", "Amala Sanjay Deshmukh", "Karan Sapra", "Andrew Tao", "Yejin Choi", "Jan Kautz", "Mingjie Liu", "Yi Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d7e6e0b66f2923cb66ab7aa404bedf9f4e1c73ae00df2e9fbfa7a5c3795cac4c", "submitted_at": "2026-06-15T22:04:11Z", "updated_at": "2026-06-15T22:04:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.17321v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.17321", "identifiers": {"arxiv": "2606.17321"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.17321"}], "name": "ProCUA-SFT Technical Report", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "ProCUA-SFT generates tasks and synthetic trajectories on live desktops seeded with document content, checking task preconditions before rollout. The resulting step-prefix examples are formatted to match inference context and used for supervised GUI-agent fine-tuning.", "topics": ["data-generation", "imitation-learning", "evaluation"], "updated_at": "2026-06-15", "url": "https://arxiv.org/abs/2606.17321"}
{"aliases": [], "authors": ["Jingyuan Huang", "Zuming Huang", "Yucheng Shi", "Tianze Yang", "Xiaoming Zhai", "Wei Chu", "Ninghao Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "49b935d8993be0d03ae778375f908b68b3ec3c614a7406b6b3f6a00f2a89ef1f", "submitted_at": "2026-06-16T16:02:41Z", "updated_at": "2026-06-17T08:58:52Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.18101v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.18101", "identifiers": {"arxiv": "2606.18101"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.18101"}], "name": "Trust the Right Teacher: Quality-Aware Self-Distillation for GUI Grounding", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Proposes quality-aware self-distillation for GUI coordinate generation under student-produced prefixes. A correctness-aware gate suppresses coordinate-token supervision that cannot reach the target box, while teacher-probability scaling adjusts the remaining signal strength.", "topics": ["grounding", "imitation-learning", "evaluation"], "updated_at": "2026-06-17", "url": "https://arxiv.org/abs/2606.18101"}
{"aliases": [], "authors": ["Yujin Zhang", "Daye Nam"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c52695e860206d02735839853416965ec3d42fac134ebfa0e8f2f5f5ea043914", "submitted_at": "2026-06-17T04:13:20Z", "updated_at": "2026-06-17T04:13:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.18671v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.18671", "identifiers": {"arxiv": "2606.18671"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.18671"}], "name": "HANSEL: Extracting Breadcrumbs from Web Agent Trajectories for Interactive Verification", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "HANSEL extracts evidence pages and snippets from browser-agent traces and preserves relevant page state in interactive verification views. It flags untraceable answers and evaluates evidence selection and human verification effort separately.", "topics": ["human-in-the-loop", "observability", "recording", "evaluation"], "updated_at": "2026-06-17", "url": "https://arxiv.org/abs/2606.18671"}
{"aliases": [], "authors": ["Li Gu", "Zihuan Jiang", "Linqiang Guo", "Zhixiang Chi", "Ziqiang Wang", "Huan Liu", "Yuanhao Yu", "Tse-Hsun Chen", "Yang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9bf8b07e08b275d90fcbfc63d867f54f5579509132cc9cf694d01447b7f91b78", "submitted_at": "2026-06-16T02:36:22Z", "updated_at": "2026-06-16T02:36:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.19388v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.19388", "identifiers": {"arxiv": "2606.19388"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.19388"}], "name": "Beyond the GUI Paradigm: Do Mobile Agents Need the Phone Screen?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-16", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "This study compares general coding agents using mobile command-line access with reproducible GUI agents on AndroidWorld and MobileWorld. It also supplies oracle CLI analyses and a separate task suite for operations such as bulk changes, filtering and hidden device state.", "topics": ["evaluation", "automation", "cross-app", "tool-integration"], "updated_at": "2026-06-16", "url": "https://arxiv.org/abs/2606.19388"}
{"aliases": [], "authors": ["Guangyi Liu", "Pengxiang Zhao", "Gao Wu", "Yiwen Yin", "Mading Li", "Liang Liu", "Congxiao Liu", "Zhang Qi", "Mengyan Wang", "Liang Guo", "Jiangning Zhang", "Yong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cab809618ea4e218da26b4ae11bf937859c3a4bc240c91dd27b3a0d333737069", "submitted_at": "2026-06-18T08:29:33Z", "updated_at": "2026-08-07T03:54:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.19930v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.19930", "identifiers": {"arxiv": "2606.19930"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.19930"}], "name": "MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileForge couples real-app exploration and generated tasks with hierarchical feedback for mobile-agent adaptation. Its optimizer turns trajectory outcomes, step feedback and corrective hints into contextualized GRPO updates, with evaluation on AndroidWorld and a GUI-only MobileWorld split.", "topics": ["mobile-navigation", "reinforcement-learning", "data-generation", "evaluation"], "updated_at": "2026-08-07", "url": "https://arxiv.org/abs/2606.19930"}
{"aliases": [], "authors": ["Yuexing Hao", "Xiaomin Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b370491b37d5bf25c624f2bac654e438459c8084b5da292fb18a0bf9142adaa2", "submitted_at": "2026-06-18T15:25:42Z", "updated_at": "2026-06-18T15:25:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.20363v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.20363", "identifiers": {"arxiv": "2606.20363"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.20363"}], "name": "Automating SKILL.md Generation for Computer-Using Agents via Interaction Trajectory Mining", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Segments and clusters GUI trajectories into inspectable skills, then trains a skill-aware policy from those annotations. Its diagnostic results distinguish readable source-domain clusters from reliable transfer and report limited or absent downstream gains.", "topics": ["memory", "data-generation", "reinforcement-learning", "evaluation"], "updated_at": "2026-06-18", "url": "https://arxiv.org/abs/2606.20363"}
{"aliases": [], "authors": ["Daniel Beechey", "Derek Yuen", "Jianheng Liu", "Dezhao Luo", "Tiantian He", "Weilin Luo", "Jun Wang", "Kun Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9e761b20caa6882605ecfd6cd001aeda4a2d36b2e702eea9a6d70451fc0ea9eb", "submitted_at": "2026-05-26T13:32:14Z", "updated_at": "2026-05-26T13:32:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.20622v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:00:51.952206+00:00", "role": "primary", "sha256": "5cd7fb95d5ee8e8c473e0c9d5b892bf50ee425604c6a320e8e9a12eff17120a1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:00:51.952206+00:00", "role": "primary", "sha256": "5cd7fb95d5ee8e8c473e0c9d5b892bf50ee425604c6a320e8e9a12eff17120a1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:14:00.181958+00:00", "role": "primary", "sha256": "e4711466886372760daa83384cb4d129d54fdcfe320be6e91fb7b4971d1a61de", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/pdf/2606.20622v1"}], "id": "arxiv:2606.20622", "identifiers": {"arxiv": "2606.20622"}, "kind": "paper", "last_verified_at": "2026-10-18T10:00:51.952206+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.20622"}], "name": "Darwin Mobile Agent: A Roadmap for Self-Evolution", "notes": "The implemented stage uses selected tasks, an external judge and heuristic history management. Autonomous curricula, verification and memory are roadmap directions, not completed capabilities established by this record. No runtime or results were reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (2 Agent-Environment Interface and Policy Optimisation; 3 Scalable Asynchronous Mobile Environment; 3.2 Persistent State Management; 3.3 Task Lifecycle; 4.2 Rollout and Optimisation; 4.3 Anatomy of the Self-Evolving Agent); not an entire-paper reading or reproduction claim.", "published_at": "2026-05-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes asynchronous reinforcement-learning infrastructure using parallel cloud-hosted Android instances, explicit task lifecycles, and persistent-state management for mobile GUI interaction.", "topics": ["reinforcement-learning", "mobile-navigation", "evaluation"], "updated_at": "2026-05-26", "url": "https://arxiv.org/abs/2606.20622"}
{"aliases": [], "authors": ["Xuelong Dai", "Jianyu Ma", "Boyang Ma", "Biwei Yan", "Yijun Yang", "Yue Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fea441692131ec677e4a9ae54f9ecef1f31d422e20884b496b124e99392e59c1", "submitted_at": "2026-06-16T15:31:33Z", "updated_at": "2026-10-15T04:29:19Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.20717v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.20717", "identifiers": {"arxiv": "2606.20717"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.20717"}], "name": "MIRAGE: Stealthy Visual Prompt Injection for Vulnerability Detection in Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "MIRAGE evaluates targeted next-action hijacking from adversarial images restricted to legitimate third-party webpage regions. It combines diffusion-based image generation with execution-aware optimization and tests the constrained visual-injection setting on web-agent frameworks.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": "2026-10-15", "url": "https://arxiv.org/abs/2606.20717"}
{"aliases": [], "authors": ["Ahmed Awadallah", "Sahil Gupta", "Yash Lara", "Yadong Lu", "Hussein Mozannar", "Akshay Nambi", "Zach Nussbaum", "Yash Pandya", "Aravind Rajeswaran", "Corby Rosset", "Alexey Taymanov", "Luiz do Valle", "Vibhav Vineet", "Spencer Whitehead", "Andrew Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "414eced85da9f76828e4398b977f9b8dac77236c97d1594108c8f5dfc3be7cbb", "submitted_at": "2026-06-18T17:53:03Z", "updated_at": "2026-07-22T17:47:01Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.20785v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.20785", "identifiers": {"arxiv": "2606.20785"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.20785"}], "name": "Fara-1.5: Scalable Learning Environments for Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds browser-agent training trajectories with live and synthetic environments, solvers, user simulation, and complementary correctness and efficiency verifiers.", "topics": ["data-generation", "web-navigation", "evaluation"], "updated_at": "2026-07-22", "url": "https://arxiv.org/abs/2606.20785"}
{"aliases": [], "authors": ["Dayeon Kang", "Hyejun Jeong", "Jade Sheffey", "Pubali Datta", "Amir Houmansadr"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8b2c6080f0de11c06fe73c30195c7c8529e07de283dc45fa47709072846d385a", "submitted_at": "2026-06-18T20:07:39Z", "updated_at": "2026-06-18T20:07:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.20910v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.20910", "identifiers": {"arxiv": "2606.20910"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.20910"}], "name": "Whose Agent Are You? Multi-Layer Fingerprinting and Attribution of Autonomous Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds a logging framework that combines network/protocol fingerprints with browser interaction behavior to attribute web-agent traffic. A decision-tree classifier is evaluated across six agent implementations, human browsing and conventional crawlers.", "topics": ["safety-security", "observability", "evaluation"], "updated_at": "2026-06-18", "url": "https://arxiv.org/abs/2606.20910"}
{"aliases": [], "authors": ["Vincent Siu", "Manasi Sharma", "Dawn Song", "Daniel Yue Zhang", "Chenguang Wang", "Ying Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "69e668a9491a31e8d8849bab8cc63eede9878fff10ccb0ae0a211610e90819b3", "submitted_at": "2026-06-19T18:00:14Z", "updated_at": "2026-08-19T17:29:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.21654v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.21654", "identifiers": {"arxiv": "2606.21654"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.21654"}], "name": "ChainWorld: Composing Long-Horizon Desktop Workloads from Atomic OSWorld Tasks", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "ChainWorld composes compatible atomic OSWorld tasks into multi-objective desktop chains while preserving their evaluators. It compares giving the full sequence at once with revealing goals turn by turn, exposing different state and session-management failures.", "topics": ["evaluation", "long-horizon", "state-tracking", "workflow"], "updated_at": "2026-08-19", "url": "https://arxiv.org/abs/2606.21654"}
{"aliases": [], "authors": ["Yikun Fu", "Bowen Fu", "Zhenyu Wu", "Shuang Cheng", "Xiaowei Sun", "Bowen Yang", "Zehao Li", "Yibo Zhao", "Zichen Ding", "Zhoumianze Liu", "Shijie Wang", "Biqing Qi", "Bowen Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6dc52f58bb2237d12041fa8b994b20bd1e41290aaab0c70d2c5dd79fcaca37b6", "submitted_at": "2026-06-21T15:32:57Z", "updated_at": "2026-06-21T15:32:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.22557v1"}}, "environments": ["desktop", "macos"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}], "id": "arxiv:2606.22557", "identifiers": {"arxiv": "2606.22557"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.22557"}], "name": "MacAgentBench: Benchmarking AI Agents on Real-World macOS Desktop", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces macOS tasks spanning GUI and command-line interaction, with deterministic checks and intermediate-progress scoring for application workflows.", "topics": ["evaluation", "cross-app", "tool-integration"], "updated_at": "2026-06-21", "url": "https://arxiv.org/abs/2606.22557"}
{"aliases": [], "authors": ["Yincheng Zhou", "Athena Zhuoming Zhong", "Shijie Zhang", "Kevin Zhang", "Teresa Xiaotao Shang", "Shanghang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fbd8892e06a5e847c16aa3101b98b3d669a783b84ee8e3d4d4d810754914028b", "submitted_at": "2026-06-22T07:28:55Z", "updated_at": "2026-06-22T07:28:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.22948v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.22948", "identifiers": {"arxiv": "2606.22948"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.22948"}], "name": "ENVS: Environment-Native Verified Search for Long-Horizon GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "ENVS branches over GUI actions in live desktop VMs, verifies successful leaves, and converts them into balanced step-level training supervision. It also introduces OSWorld-Noisy to test recovery from dynamic interruptions while retaining the underlying tasks.", "topics": ["data-generation", "reinforcement-learning", "long-horizon", "robustness", "evaluation"], "updated_at": "2026-06-22", "url": "https://arxiv.org/abs/2606.22948"}
{"aliases": [], "authors": ["Anmol Goel", "Iryna Gurevych"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "da8190bc5afc0bae400fc8ab5e7d80a7f34488c3400dce4477ca61635b19d5b7", "submitted_at": "2026-06-22T11:36:58Z", "updated_at": "2026-09-10T18:49:54Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.23189v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.23189", "identifiers": {"arxiv": "2606.23189"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.23189"}], "name": "Capable but Careless: Do Computer-Use Agents Follow Contextual Integrity?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "AgentCIBench evaluates whether computer-use agents disclose information in ways that violate its original context. Deterministically scored scenarios test nearby sensitive UI content, ambiguous requests and recipient mismatches, including end-to-end agent execution.", "topics": ["evaluation", "safety-security", "cross-app"], "updated_at": "2026-09-10", "url": "https://arxiv.org/abs/2606.23189"}
{"aliases": [], "authors": ["Marta Sumyk", "Oleksandr Kosovan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0aeb8c28b570133f575c2f336b6dab0dcbecdfaec78549115407b21722816e8a", "submitted_at": "2026-06-23T12:46:29Z", "updated_at": "2026-06-23T12:46:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.24515v1"}}, "environments": ["desktop", "macos", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.24515", "identifiers": {"arxiv": "2606.24515"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.24515"}], "name": "Reinforcement Learning for Computer-Use Agents with Autonomous Evaluation", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "The framework uses a vision-language judge’s final-screen assessment as a terminal reward for GUI reinforcement learning. It models evaluator noise and corrects the reward estimator used by PPO across desktop benchmarks.", "topics": ["reinforcement-learning", "reward-models", "evaluation", "robustness"], "updated_at": "2026-06-23", "url": "https://arxiv.org/abs/2606.24515"}
{"aliases": [], "authors": ["Jiachen Qian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a3d0341f41a8ecba57fb7ed9e4b266bb4a176639c742be2b4fc98555754f5744", "submitted_at": "2026-06-23T12:57:02Z", "updated_at": "2026-06-23T12:57:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.24525v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.24525", "identifiers": {"arxiv": "2606.24525"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.24525"}], "name": "VisCritic: Visual State Comparison as Process Reward for GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "VisCritic compares before/after screenshots through change-aware visual features and an action-conditioned critic. Weakly supervised samples from existing traces train judgments of action success, progress and error type for GUI-agent verification.", "topics": ["reward-models", "state-tracking", "data-generation", "evaluation"], "updated_at": "2026-06-23", "url": "https://arxiv.org/abs/2606.24525"}
{"aliases": [], "authors": ["Xiao Zhou", "Siyue Zhang", "Yilun Zhao", "Jinbiao Wei", "Tingyu Song", "Arman Cohan", "Chen Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "77f777f05b1c169ee4ad7ae7d77e9d1fc80a0696e9a1f6eefb5028a5c3e6fe5d", "submitted_at": "2026-06-22T17:05:07Z", "updated_at": "2026-06-22T17:05:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.24551v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.24551", "identifiers": {"arxiv": "2606.24551"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.24551"}], "name": "GUI vs. CLI: Execution Bottlenecks in Screen-Only and Skill-Mediated Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "This benchmark matches desktop goals, initial states and final-state verifiers across screen-only GUI and skill-mediated CLI agents. Its experiments separate graphical execution failures from missing CLI skill coverage, including a verifier-guided skill-augmentation condition.", "topics": ["evaluation", "tool-integration", "robustness", "workflow"], "updated_at": "2026-06-22", "url": "https://arxiv.org/abs/2606.24551"}
{"aliases": [], "authors": ["Divake Kumar", "Sina Tayebati", "Devashri Naik", "Amanda Sofie Rios", "Nilesh Ahuja", "Omesh Tickoo", "Ranganath Krishnan", "Amit Ranjan Trivedi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "628d08bdbf23ae9e599bf767f0503cecedcde72dfbfcc183f9ea9f044c6f2d44", "submitted_at": "2026-06-24T12:34:28Z", "updated_at": "2026-09-29T20:55:51Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.25760v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.25760", "identifiers": {"arxiv": "2606.25760"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.25760"}], "name": "Uncertainty Quantification for Computer-Use Agents: A Benchmark across Vision-Language Models and GUI Grounding Datasets", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Argus benchmarks post-hoc uncertainty methods for single-step GUI grounding across open-weight and closed-source observation regimes. It tests calibration, rejection, error severity and conformal click regions, finding that uncertainty rankings transfer less reliably across model/interface regimes than across datasets for one model.", "topics": ["grounding", "evaluation", "robustness", "safety-security"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2606.25760"}
{"aliases": [], "authors": ["Shoufa Chen", "Luyuan Wang", "Xuan Yang", "Zhiheng Liu", "Yuren Cong", "Yuanfeng Ji", "Feiyan Zhou", "Xiaohui Zhang", "Fanny Yang", "Belinda Zeng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6164c42541650e5120adc881b19524cb3b268f08d6eb77fbe91d8c1a02823786", "submitted_at": "2026-06-26T17:59:51Z", "updated_at": "2026-06-26T17:59:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.28480v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:28Z", "role": "primary", "sha256": "7eaa9a5430d3a45049e612a117da345c1daafc54277a35a7cf56b42fbb1e8618", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.05646%2C2606.05670%2C2606.06556%2C2606.06891%2C2606.07367%2C2606.07594%2C2606.07682%2C2606.07723%2C2606.07867%2C2606.10394%2C2606.10525%2C2606.11290%2C2606.12203%2C2606.12736%2C2606.13040%2C2606.13368%2C2606.13578%2C2606.13669%2C2606.13802%2C2606.13898%2C2606.14397%2C2606.15057%2C2606.15300%2C2606.15441%2C2606.17162%2C2606.18425%2C2606.19336%2C2606.19409%2C2606.19704%2C2606.20950%2C2606.21033%2C2606.21565%2C2606.23521%2C2606.23654%2C2606.24775%2C2606.25084%2C2606.26080%2C2606.27406%2C2606.27634%2C2606.28430%2C2606.28480%2C2606.29955%2C2606.30616%2C2607.00692%2C2607.01084%2C2607.02032%2C2607.02448%2C2607.02490%2C2607.02577%2C2607.02911&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:28Z", "role": "primary", "sha256": "7eaa9a5430d3a45049e612a117da345c1daafc54277a35a7cf56b42fbb1e8618", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2606.05646%2C2606.05670%2C2606.06556%2C2606.06891%2C2606.07367%2C2606.07594%2C2606.07682%2C2606.07723%2C2606.07867%2C2606.10394%2C2606.10525%2C2606.11290%2C2606.12203%2C2606.12736%2C2606.13040%2C2606.13368%2C2606.13578%2C2606.13669%2C2606.13802%2C2606.13898%2C2606.14397%2C2606.15057%2C2606.15300%2C2606.15441%2C2606.17162%2C2606.18425%2C2606.19336%2C2606.19409%2C2606.19704%2C2606.20950%2C2606.21033%2C2606.21565%2C2606.23521%2C2606.23654%2C2606.24775%2C2606.25084%2C2606.26080%2C2606.27406%2C2606.27634%2C2606.28430%2C2606.28480%2C2606.29955%2C2606.30616%2C2607.00692%2C2607.01084%2C2607.02032%2C2607.02448%2C2607.02490%2C2607.02577%2C2607.02911&max_results=50"}], "id": "arxiv:2606.28480", "identifiers": {"arxiv": "2606.28480"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.28480"}], "name": "TUA-Bench: A Benchmark for General-Purpose Terminal-Use Agents", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent general computer-operation benchmark: the evaluated interface is a terminal; GUI benchmarks are a contrast, not its execution modality.", "published_at": "2026-06-26", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "TUA-Bench evaluates terminal-use agents on general digital tasks, including document editing, email, web information seeking and specialist software workflows. Tasks run in real terminals with deterministic setup and execution-based scoring.", "topics": ["evaluation", "automation"], "updated_at": "2026-06-26", "url": "https://arxiv.org/abs/2606.28480"}
{"aliases": [], "authors": ["Bojie Li", "Noah Shi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "54de4e93f4f3e39d904b4661a193765e02c29f8ec7787264aee5616579c32210", "submitted_at": "2026-06-28T15:59:31Z", "updated_at": "2026-06-28T15:59:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.29472v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.29472", "identifiers": {"arxiv": "2606.29472"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.29472"}], "name": "Agent-Computer Observation Interfaces Enable Dynamic Computer Use", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "The Agent-Computer Observation Interface separates continuous perception from discrete computer actions using captured frames, audio transcription and persistent visual narration. DynaCU-Bench tests these components on dynamic browser tasks and static controls, with model-specific ablations showing that additional observations can also hurt.", "topics": ["evaluation", "video", "screen-understanding", "memory", "training-free"], "updated_at": "2026-06-28", "url": "https://arxiv.org/abs/2606.29472"}
{"aliases": [], "authors": ["Mengqi Yuan", "Zilong Zhou", "Xinzhuang Xiong", "Weiming Wu", "Jiayang Sun", "Jiamin Song", "Kaiqian Cui", "Bowen Wang", "Haoyuan Wu", "Yitong Li", "Dunjie Lu", "Haikong Lu", "Qi Zhen", "Xinyuan Wang", "Jiaqi Deng", "Yuhao Yang", "Cheng Chen", "Boyuan Zheng", "Alex Su", "Xiao Yu", "Hao Zou", "Saaket Agashe", "Xing Han Lu", "Manpreet Kaur", "Zhengyang Qi", "Vincent Sunn Chen", "Frederic Sala", "Dayiheng Liu", "Junyang Lin", "Zhou Yu", "Yu Su", "Siva Reddy", "Xin Eric Wang", "Peng Qi", "Tianbao Xie", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "03035fcbc6cd3e6bf40f244870fb35694ba985ff7f809da66e2a376cbf99d66b", "submitted_at": "2026-06-28T17:59:17Z", "updated_at": "2026-07-13T10:30:17Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.29537v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.29537", "identifiers": {"arxiv": "2606.29537"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.29537"}], "name": "OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces extended computer-use workflows with realistic artifacts, changing information, hidden state, and partial-progress and safety assessments.", "topics": ["evaluation", "long-horizon", "workflow"], "updated_at": "2026-07-13", "url": "https://arxiv.org/abs/2606.29537"}
{"aliases": [], "authors": ["Chen Liu", "Ling Chen", "Hanzhang Zhou", "Liangyu Chen", "Chenglin Cai", "Xin Yu", "Steven Hoi", "Yue Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "50e44e08750d4152f9b5408885f8a6a016c5f789af9712e447ee147fb25adb8d", "submitted_at": "2026-06-29T10:20:39Z", "updated_at": "2026-06-29T10:20:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.30084v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.30084", "identifiers": {"arxiv": "2606.30084"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.30084"}], "name": "One Forward Beats Two: InnerZoom for Accurate and Efficient GUI Grounding", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "InnerZoom preserves target-region cues from intermediate decoder layers and reinjects a compact evidence state during coordinate decoding. Its single-forward GUI grounder is compared with crop-and-rerun zooming and matched supervised/RL baselines for accuracy and inference cost.", "topics": ["grounding", "efficiency", "screen-understanding", "evaluation"], "updated_at": "2026-06-29", "url": "https://arxiv.org/abs/2606.30084"}
{"aliases": [], "authors": ["Iliana Fayolle", "Sihem Bouhenniche", "Samuel Pélissier", "Pierre Laperdrix", "Clémentine Maurice", "Walter Rudametkin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "770c901e3490b2a43b02296136825e8e6bd035fb7494ea74b3a9af74be1afed1", "submitted_at": "2026-06-29T10:51:13Z", "updated_at": "2026-06-29T10:51:13Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.30119v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.30119", "identifiers": {"arxiv": "2606.30119"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.30119"}], "name": "On the Internet, Nobody Knows You're an LLM Bot: Unmasking Web Agents with Multi-Layer Fingerprinting", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates web agents against anti-bot mechanisms on instrumented honeysites and analyzes network, HTTP and browser fingerprints. The study compares resistance to blocking with the ability to distinguish agents, humans and agent implementations.", "topics": ["safety-security", "observability", "evaluation"], "updated_at": "2026-06-29", "url": "https://arxiv.org/abs/2606.30119"}
{"aliases": [], "authors": ["Xueqiao Sun", "Xiaohan Wang", "Ludwig Schmidt", "Serena Yeung-Levy", "Yuhui Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8911abdc0c9c481d82abb213d368c16fba9fee909c354e0bc6272e14c59abfb7", "submitted_at": "2026-06-30T07:44:37Z", "updated_at": "2026-06-30T07:44:37Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.31270v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.31270", "identifiers": {"arxiv": "2606.31270"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.31270"}], "name": "Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses failed computer-use trajectories to diagnose failure modes and generate inference-time agent code patches, with light human verification. An OSWorld study evaluates this failure-driven improvement loop without additional model training.", "topics": ["self-reflection", "robustness", "training-free", "evaluation"], "updated_at": "2026-06-30", "url": "https://arxiv.org/abs/2606.31270"}
{"aliases": [], "authors": ["Meng Chen", "Anya Ji", "Tsung-Han Wu", "Tobias Maringgele", "David M. Chan", "Alane Suhr", "Amy Pavel"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a0a3e641e0a71051f5fbd20f57af3db06f2b65c61025e78adcdd122f773d034b", "submitted_at": "2026-06-30T17:18:47Z", "updated_at": "2026-09-01T04:17:49Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.31980v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.31980", "identifiers": {"arxiv": "2606.31980"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.31980"}], "name": "DigitalCoach: Communication and Grounding Gaps in Human and Agentic Computer Use Coaching", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-30", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "DigitalCoach records expert–novice software coaching conversations aligned with screen and input events. Automated and interactive evaluations examine how model coaches differ from humans in explanation, error diagnosis and visual grounding when helping a learner operate software.", "topics": ["human-in-the-loop", "video", "evaluation", "grounding"], "updated_at": "2026-09-01", "url": "https://arxiv.org/abs/2606.31980"}
{"aliases": [], "authors": ["Guijia Zhang", "Yuxun Chen", "Yuheng Qi", "Harry Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e7fdae5d6e5ea17639fe0e2b8e2d655df7201a438d84acdeb3e0d02a78051ae2", "submitted_at": "2026-07-05T14:33:29Z", "updated_at": "2026-07-18T16:56:38Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.04334v2"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.04334", "identifiers": {"arxiv": "2607.04334"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.04334"}], "name": "Do GUI Agents Believe Their Eyes? Diagnosing State-Belief Reliance on Pixels versus Structure", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Defines visual state reliance and probes whether GUI agents resolve screenshot/structure conflicts using pixels, serialized interface structure or prior beliefs. Paired interventions and live multi-step tests connect misattributed state beliefs to failures and compare consistency-based mitigations.", "topics": ["screen-understanding", "state-tracking", "evaluation", "robustness", "safety-security"], "updated_at": "2026-07-18", "url": "https://arxiv.org/abs/2607.04334"}
{"aliases": [], "authors": ["Woohyuk Choi", "Juhee Kim", "Taehyun Kang", "Jihyeon Jeong", "Luyi Xing", "Byoungyoung Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c5fb30b8be2c6e452471e6398720291ec59a2c5ef96e64baca0735fb01b7a12e", "submitted_at": "2026-07-06T14:07:49Z", "updated_at": "2026-07-06T14:07:49Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.05120v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}], "id": "arxiv:2607.05120", "identifiers": {"arxiv": "2607.05120"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.05120"}], "name": "Agent Data Injection Attacks are Realistic Threats to AI Agents", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-07-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study examines agent data injection, where untrusted content impersonates trusted metadata or tool-context data. It reports arbitrary-click vulnerabilities in web agents alongside separate coding-agent attacks, highlighting trust-boundary failures beyond instruction injection.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-07-06", "url": "https://arxiv.org/abs/2607.05120"}
{"aliases": [], "authors": ["Wei Dong", "Tianyu Fu", "Zhe Yu", "Hanning Wang", "Anyang Su", "Zhizhou Fang", "Yuyang Chen", "Shuo Wang", "Minghui Wu", "Ping Jiang", "Zhen Lei", "Chenxu Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cc6d846f43b9992388e94f3bcfdfc9a25d09b317fea62336aa22951c8fce9f8b", "submitted_at": "2026-07-07T10:27:31Z", "updated_at": "2026-07-07T10:27:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.06118v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.06118", "identifiers": {"arxiv": "2607.06118"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.06118"}], "name": "WebRetriever: A Large-Scale Comprehensive Benchmark for Efficient Web Agent Evaluation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebRetriever evaluates navigation, knowledge-assisted interaction and end-to-end information-extraction workflows under separate protocols. Its NavEval judge uses richer interaction context to diagnose cases where navigation success does not translate into useful task completion.", "topics": ["evaluation", "web-navigation", "observability"], "updated_at": "2026-07-07", "url": "https://arxiv.org/abs/2607.06118"}
{"aliases": [], "authors": ["Chengguang Gan", "Hanjun Wei", "Yunhao Liang", "Zhixi Cai", "Qinghao Zhang", "Shiwen Ni"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "37cfa436d182ed8fbc8823213b9173fbfcf9fb421e8d2ae89a093a5393c374ba", "submitted_at": "2026-07-11T02:08:46Z", "updated_at": "2026-07-16T03:28:43Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2607.10079v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.10079", "identifiers": {"arxiv": "2607.10079"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.10079"}], "name": "MAG: A Web-Agent Benchmark and Harness for Multimodal Action and Guide Generation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAG jointly evaluates browser task actions and user-facing guides grounded in screenshots through element marks or pixel coordinates. Its harness covers annotation, live evaluation and joint metrics, with an expert-trajectory-augmented GRPO training study.", "topics": ["evaluation", "grounding", "reinforcement-learning", "action-prediction"], "updated_at": "2026-07-16", "url": "https://arxiv.org/abs/2607.10079"}
{"aliases": [], "authors": ["Yixiong Chen", "Xinyi Bai", "Alan Yuille"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0473ced7881ed2007aa49c2767860c33719d5bf1c798dfe4771a4061b13d655e", "submitted_at": "2026-07-12T07:05:31Z", "updated_at": "2026-07-12T07:05:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.10608v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}], "id": "arxiv:2607.10608", "identifiers": {"arxiv": "2607.10608"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.10608"}], "name": "The Compliance Trap: Diagnosing How AI Agents Consume Conflicting Memory", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-07-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "The Entry–Propagation–Recovery framework traces where conflicting memory first changes an agent action, how the effect propagates and whether recovery occurs. WebArena and the controlled MemTrapBench separate memory-consumption failures across these stages.", "topics": ["memory", "safety-security", "robustness", "evaluation", "long-horizon"], "updated_at": "2026-07-12", "url": "https://arxiv.org/abs/2607.10608"}
{"aliases": [], "authors": ["Chengguang Gan", "Zhixi Cai", "Yunhao Liang", "Hanjun Wei", "Shiwen Ni", "Qinghao Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e96ea0cf71ff705c064db1bcd1d2cbf8b8cd712bb4e8030931cafbb31ff29b10", "submitted_at": "2026-07-14T11:17:35Z", "updated_at": "2026-07-14T11:17:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.12640v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.12640", "identifiers": {"arxiv": "2607.12640"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.12640"}], "name": "A Learning-Rate-Gated Failure of GRPO in a Small Language and Vision-Language Model Web Agent: A Controlled Null and Its Mechanism", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies a controlled setting where GRPO fails to improve a strong small web agent and can reduce text-track performance. Matched reachable-reward tasks and component grafting distinguish limited optimization headroom from learning-rate-dependent failure mechanisms.", "topics": ["reinforcement-learning", "evaluation", "robustness"], "updated_at": "2026-07-14", "url": "https://arxiv.org/abs/2607.12640"}
{"aliases": [], "authors": ["Yong Liu", "Zhenyi Zhong", "Zhanpeng Shi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a779b2c1ff9ba3b531d399204b0986e3affcea81c67fdd7951e3b7da343b0e17", "submitted_at": "2026-07-16T00:35:25Z", "updated_at": "2026-07-16T00:35:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.14443v1"}}, "environments": ["desktop", "macos"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.14443", "identifiers": {"arxiv": "2607.14443"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.14443"}], "name": "Tactile: Giving Computer-Using Agents Hands and Feet", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-07-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Tactile turns accessibility, OCR and visual evidence into explicit desktop targets with executable actions and verification cues. Its execution loop chooses semantic actions or coordinate fallbacks while retaining provenance for replay and diagnosis.", "topics": ["grounding", "automation", "observability", "tool-integration", "robustness", "evaluation"], "updated_at": "2026-07-16", "url": "https://arxiv.org/abs/2607.14443"}
{"aliases": [], "authors": ["Yaohan Yang", "Minglei Shi", "Borui Zhang", "Jie Zhou", "Jiwen Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6488798626abbec89ca739a0c5d188ae440484eb59b3be11f2b319ce742b22ec", "submitted_at": "2026-07-19T03:29:08Z", "updated_at": "2026-07-19T03:29:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.17050v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}], "id": "arxiv:2607.17050", "identifiers": {"arxiv": "2607.17050"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:09Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.17050"}], "name": "EvoGUI: An Evolution-Aware Benchmark for GUI State-Transition Understanding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces EvoGUI, which derives temporal ordering, inverse action/value prediction, and successor-discrimination probes from recorded GUI trajectories. Its web-derived benchmark isolates transition understanding from the other abilities involved in end-to-end task success.", "topics": ["evaluation", "world-models", "action-prediction"], "updated_at": "2026-07-19", "url": "https://arxiv.org/abs/2607.17050"}
{"aliases": [], "authors": ["Behzad Ousat", "Nikita Turkmen", "Lalchandra Rampersaud", "Dillan Bailey", "Amin Kharraz"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e382b15a9d25e99aa187a72c7d2a33bbeca82b2b347ba580ca0a1d73a8e01d19", "submitted_at": "2026-07-21T03:02:57Z", "updated_at": "2026-07-21T03:02:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.18659v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2607.18659", "identifiers": {"arxiv": "2607.18659"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.18659"}], "name": "Broken Gates: Re-evaluating Web Bot Defenses in the Age of LLM Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Measures how autonomous LLM browser agents interact with commercial bot-management and anti-automation defenses on the web.", "topics": ["safety-security", "web-navigation", "evaluation"], "updated_at": "2026-07-21", "url": "https://arxiv.org/abs/2607.18659"}
{"aliases": [], "authors": ["Zedong Yu", "Qianxing Li", "Zhi Gao", "Liuyu Xiang", "Chenrui Shi", "Yang Liu", "Huiming Wu", "Yujie Wei", "Yuhao Fei", "Yubo Fu", "Zhaofeng He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "078ff9a56b769923d15a98581685a3e2f47aac7d898ea731999cabaec3918fa7", "submitted_at": "2026-07-17T12:48:03Z", "updated_at": "2026-07-17T12:48:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.22689v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.22689", "identifiers": {"arxiv": "2607.22689"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.22689"}], "name": "Beyond Sequential Interaction: Benchmarking Parallel Execution and Coordination for GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces ParaGUIBench, with tasks and efficiency measurements for coordinated GUI agents operating separate desktop instances over a shared file system. Its ParaGUI planner-worker system decomposes tasks and dispatches concurrent subtasks for comparison with serial execution.", "topics": ["evaluation", "planning", "efficiency", "long-horizon"], "updated_at": "2026-07-17", "url": "https://arxiv.org/abs/2607.22689"}
{"aliases": [], "authors": ["Yang Wan", "Zhenhao Zhang", "Jierui Wang", "Xiangxin Zhou", "Linchao Zhu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2de35aa3c7f4457d43d485eb136d8b380ea61870cc7253e807880ae731c51316", "submitted_at": "2026-07-25T16:00:45Z", "updated_at": "2026-09-26T12:15:44Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.23263v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.23263", "identifiers": {"arxiv": "2607.23263"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.23263"}], "name": "SeekJudge: A Practical Reward Framework for Reinforcement Learning in Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SeekJudge, which iteratively searches and analyzes GUI trajectories to assess instruction fulfillment. The work combines dense judgment-data synthesis, a step-label benchmark, and reward-server scheduling to study accurate and practical reward feedback for agent training.", "topics": ["reward-models", "evaluation", "data-generation"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2607.23263"}
{"aliases": [], "authors": ["Chenrui Shi", "Yuwei Wu", "Yang Liu", "Ruining Feng", "Zirui Shang", "Zhi Gao", "Lifeng Fan", "Che Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d1688a528d35a8213a8027385b1b1327536567f6ec307aa6562eea543ee71d7f", "submitted_at": "2026-07-28T16:01:38Z", "updated_at": "2026-07-29T05:13:57Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.25904v2"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}], "id": "arxiv:2607.25904", "identifiers": {"arxiv": "2607.25904"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:09Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.25904"}], "name": "Interactive Reward Agent: GUI Task Evaluation via Environment-State Verification", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces an Interactive Reward Agent that proposes completion conditions and verifies them using system, application, and GUI tools after execution. GUI-RewardBench provides annotated desktop trajectories for evaluating these evidence-based judgments and their use in reinforcement learning.", "topics": ["reward-models", "evaluation", "tool-integration"], "updated_at": "2026-07-29", "url": "https://arxiv.org/abs/2607.25904"}
{"aliases": [], "authors": ["Abhishek Pillai", "Samir Kumar Nayak", "Yuan Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "352c892a52692301e14125e126a52508feb39ef7cd034d7cb0c48a27fb54b8ac", "submitted_at": "2026-07-28T17:49:51Z", "updated_at": "2026-07-29T19:07:50Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.26041v2"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.26041", "identifiers": {"arxiv": "2607.26041"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.26041"}], "name": "Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Desktop-Delta Bench, an offline diagnostic of GUI transition understanding using temporal-ordering tasks and before/after action inference. It includes decoy observations to test whether models track the source and task relevance of screen changes.", "topics": ["evaluation", "screen-understanding", "state-tracking"], "updated_at": "2026-07-29", "url": "https://arxiv.org/abs/2607.26041"}
{"aliases": [], "authors": ["Zihan Dong", "Zhiyuan Ma", "Zekun Wang", "Yunqing Li", "Zirou Liu", "Ruixuan Deng", "Qishi Zhan", "Rui Qian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d75f6c942d6854a259556b7ef1060fa0f142755162da329580c6177cd1adef15", "submitted_at": "2026-07-30T15:29:42Z", "updated_at": "2026-07-30T15:29:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.28367v1"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.28367", "identifiers": {"arxiv": "2607.28367"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.28367"}], "name": "How Benchmarks Mis-Score Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Audits failure-scored computer-use trajectories to distinguish agent errors from broken tasks and evaluator false negatives. It organizes reliability issues across task construction, trajectory evidence, scoring, and reporting, then derives corresponding evaluation design practices.", "topics": ["evaluation", "observability", "robustness"], "updated_at": "2026-07-30", "url": "https://arxiv.org/abs/2607.28367"}
{"aliases": [], "authors": ["Woongkyu Lee", "Jungwook Choi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bacd1c5df5bf3163c83fae546ae4ced36996f33115568dfbc9e44b17fe507a31", "submitted_at": "2026-07-30T17:36:36Z", "updated_at": "2026-07-30T17:36:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.28573v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.28573", "identifiers": {"arxiv": "2607.28573"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.28573"}], "name": "Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Examines inference-time scaling for local computer-use models by varying context, step budgets, decomposition, and parallel execution. OSWorld experiments track both task performance and changing failure modes, exposing cases where extra computation extends errors or adds coordination overhead.", "topics": ["efficiency", "evaluation", "robustness"], "updated_at": "2026-07-30", "url": "https://arxiv.org/abs/2607.28573"}
{"aliases": [], "authors": ["Qiushi Sun", "Kanzhi Cheng", "Yian Wang", "Bowen Yang", "Hang Yan", "Liheng Chen", "Fangzhi Xu", "Zichen Ding", "Nuo Chen", "Jialin Cao", "Xingdong Gong", "Zehao Li", "Kaiming Jin", "Xinfeng Yuan", "Zhoumianze Liu", "Jingyang Gong", "Zhangyue Yin", "Jiahui Gao", "Zhiyong Wu", "Tianbao Xie", "Jianbing Zhang", "Ben Kao", "Lingpeng Kong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e3e82df7abffbd0c8f7e969ed8edb77423263a2b1d826a3e72529b0a51903d58", "submitted_at": "2026-07-30T17:57:41Z", "updated_at": "2026-08-06T17:55:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.28609v2"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.28609", "identifiers": {"arxiv": "2607.28609"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.28609"}], "name": "OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OSReward and its harder and multidimensional variants for assessing judgments of computer-use trajectories. It also releases a reasoning-annotated judgment corpus and trains reward models, examining reliability, leniency bias, and evaluation cost.", "topics": ["reward-models", "evaluation", "data-generation"], "updated_at": "2026-08-06", "url": "https://arxiv.org/abs/2607.28609"}
{"aliases": [], "authors": ["Dunjie Lu", "Shuai Bai", "Tianyi Bai", "Sicheng Fan", "Chang Gao", "Jian Guan", "Feng Hu", "Mianqiu Huang", "Xingyang Huang", "Yizhen Jiang", "Yuheng Jing", "Dehui Kong", "Ning Li", "Dayiheng Liu", "Shixuan Liu", "Zheng Liu", "Que Shen", "Bowen Wang", "Junli Wang", "Chencan Wu", "Rui Xie", "Tianbao Xie", "Zhihui Xie", "Haiyang Xu", "An Yang", "Tao Yu", "Wenzhen Yuan", "Xi Zhang", "Zhenru Zhang", "Mingkang Zhu", "Zhaoqing Zhu", "Yizhong Cao", "Kai Dang", "Binyuan Hui", "Kaixin Li", "Junyang Lin", "Haiquan Wang", "Zekun Wang", "Yiheng Xu", "Fan Yan", "Mengqi Yuan", "Danyang Zhang", "Jiajun Zhang", "Zhipeng Zhang", "Fan Zhou", "Fan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "26a61ccfdddb5cdc2da2b14d29ae8d4098544fea4e9458126f2a49c28418c4f0", "submitted_at": "2026-08-03T15:04:20Z", "updated_at": "2026-08-03T15:04:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.02352v1"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.02352", "identifiers": {"arxiv": "2608.02352"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.02352"}], "name": "Qwen-CUA: Native Computer Use for (almost) Everything", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Qwen-CUA operates through screenshots and keyboard/mouse events while compressing older visual history into reusable context blocks. Its report combines verifiable task generation, trajectory-based training and separate browser and Bash-augmented studies.", "topics": ["reinforcement-learning", "imitation-learning", "data-generation", "interaction-history", "long-horizon", "evaluation"], "updated_at": "2026-08-03", "url": "https://arxiv.org/abs/2608.02352"}
{"aliases": [], "authors": ["Weijia Zhang", "Kunlun Zhu", "Zeyi Liu", "Yinting Chen", "Tianyi Ma", "Jiateng Liu", "Jiaxun Zhang", "Bingxuan Li", "Xiangru Tang", "Heng Ji", "Jiaxuan You"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a11e065bef4cf860c36ca2f682801ca21b552fbd40bd1b6f5d9ae39b56e7d77a", "submitted_at": "2026-07-31T13:59:45Z", "updated_at": "2026-09-06T04:43:23Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.02643v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.02643", "identifiers": {"arxiv": "2608.02643"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.02643"}], "name": "CUADebug: Diagnosing and Repairing Computer-Use Agent Failures", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "CUADebug combines a failure taxonomy and human-annotated OSWorld failure trajectories with an agent that locates causal steps and proposes corrections from screenshots and action traces. It evaluates root-cause diagnosis separately from controlled task re-execution and failure recovery.", "topics": ["evaluation", "robustness", "self-reflection", "state-tracking"], "updated_at": "2026-09-06", "url": "https://arxiv.org/abs/2608.02643"}
{"aliases": [], "authors": ["Zichuan Fu", "Shirong Wang", "Wenlin Zhang", "Guojing Li", "Yimin Deng", "Jingtong Gao", "Junjia Qi", "Hanyu Yan", "Yefeng Zheng", "Xiaopeng Li", "Wanyu Wang", "Xian Wu", "Xiangyu Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fb64bb5faa37e68b35a2398ad896fc02181a4d2749a3067391e6b3bf023a0f48", "submitted_at": "2026-08-04T07:47:37Z", "updated_at": "2026-08-04T07:47:37Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.03270v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.03270", "identifiers": {"arxiv": "2608.03270"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.03270"}], "name": "GUI-Lens: Coarse-to-Fine Cropping for GUI Grounding with General-Purpose VLMs", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-Lens uses OCR and detected interface components as coordinate references while a general-purpose VLM selects progressively enlarged screenshot crops. Instruction-conditioned checks constrain the crop and click proposals before mapping the final location back to the original screen.", "topics": ["grounding", "screen-understanding", "evaluation"], "updated_at": "2026-08-04", "url": "https://arxiv.org/abs/2608.03270"}
{"aliases": [], "authors": ["Siqi Fan", "Minghao Li", "Xiaoqian Ma", "Wenhui Tan", "Xiusheng Huang", "Juntong Wu", "Liujie Zhang", "Shuo Shang", "Weihang Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2c8692712194d0ff822091d13324e6bb540339103a824f1030f7a971699d0a0b", "submitted_at": "2026-08-04T08:35:51Z", "updated_at": "2026-08-06T07:26:24Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.03327v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.03327", "identifiers": {"arxiv": "2608.03327"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.03327"}], "name": "Screenshots or Tools? Eliciting Tool Use and Managing Multimodal Context in Hybrid GUI-MCP Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study separates tool-adoption behavior from task competence in hybrid GUI–MCP agents. It tests tool-use incentives and screenshot-history compression, showing why changed action preferences alone need not improve held-out accuracy.", "topics": ["tool-integration", "reinforcement-learning", "efficiency", "interaction-history", "evaluation"], "updated_at": "2026-08-06", "url": "https://arxiv.org/abs/2608.03327"}
{"aliases": [], "authors": ["Yu Ran", "Wentao Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "152e21cd4f1f290aafadfdcfaa10e15a03378404540a46b5f1a5464a2d432d23", "submitted_at": "2026-08-04T14:34:41Z", "updated_at": "2026-09-28T13:07:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.03740v2"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.03740", "identifiers": {"arxiv": "2608.03740"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.03740"}], "name": "MissClick: Execution-Aware Adversarial Attacks on Coordinate Generation in GUI Grounding Models", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "MissClick designs white-box attacks around the numerical effect of coordinate digits on executable GUI clicks. Separate untargeted displacement and targeted place-weighted objectives are evaluated across desktop, web and mobile grounding settings.", "topics": ["safety-security", "grounding", "robustness", "evaluation"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2608.03740"}
{"aliases": [], "authors": ["Longtao Guo", "Zelin Zhang", "Kaifeng Huang", "Yang Shi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "30e5ce3d86881041491ae507839b2f3d302691aea1d2fdfc92cfc00f35efefe7", "submitted_at": "2026-08-05T12:08:16Z", "updated_at": "2026-08-05T12:08:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.04741v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.04741", "identifiers": {"arxiv": "2608.04741"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.04741"}], "name": "LoginTrap: Uncovering Task-Agnostic Phishing-Style Indirect Prompt Injection Attacks against LLM-based Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces LoginTrap, a black-box indirect-injection evaluation in which attacker-controlled page context and login flows induce unauthorized authentication. It measures end-to-end private-data leakage across web-agent backbones, architectures and defenses.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": "2026-08-05", "url": "https://arxiv.org/abs/2608.04741"}
{"aliases": [], "authors": ["Dongsheng Chen", "Yuxuan Li", "Guanhua Chen", "Jiaxin Zhang", "Xiangyu Zhao", "Lei Ma", "Xin Yao", "Xuetao Wei"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5f8b84135f72558eea1fae428df7cb30f62232631de49e690e73010aeca22d4c", "submitted_at": "2026-08-05T12:23:16Z", "updated_at": "2026-08-05T12:23:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.04755v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.04755", "identifiers": {"arxiv": "2608.04755"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.04755"}], "name": "\"Allow\" to Achieve, Over-Privileged Inadvertently: The Unintended Cost of Task-Completion-Driven Pop-up Decisions in Mobile GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "This permission-literacy study tests whether mobile GUI agents grant only task-necessary permissions. Controlled changes to task context and requester identity expose authorization biases and the trade-off between suppressing unnecessary and legitimate grants.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-08-05", "url": "https://arxiv.org/abs/2608.04755"}
{"aliases": [], "authors": ["Linqiang Guo", "Wei Liu", "Li Gu", "Yang Wang", "Tse-Hsun", "Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dc7934927c82a43e6ae6a246ade1d50c4551cc49d263ca67cfb32906cbec9ea4", "submitted_at": "2026-08-06T04:17:27Z", "updated_at": "2026-08-06T04:17:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.05587v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.05587", "identifiers": {"arxiv": "2608.05587"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.05587"}], "name": "StepReflect: Structured UI Transition Reflection for Mobile GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "StepReflect predicts structured per-action reflections from transition specifications and paired visual evidence. The paper distinguishes offline transition-label accuracy from online task success when the reflection model is integrated into mobile agents.", "topics": ["self-reflection", "state-tracking", "imitation-learning", "reward-models", "evaluation", "efficiency"], "updated_at": "2026-08-06", "url": "https://arxiv.org/abs/2608.05587"}
{"aliases": [], "authors": ["Jiaming Wei", "Zekun Wu", "Adriano Koshiyama", "Maria Perez-Ortiz"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9055ec9c70777fc94179983d82bf841bac12332ebed7f49342883c8d842d0fbc", "submitted_at": "2026-08-06T15:37:04Z", "updated_at": "2026-08-06T15:37:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.06171v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.06171", "identifiers": {"arxiv": "2608.06171"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.06171"}], "name": "Routing Is Least Learnable Where It Is Most Valuable: Bounds on Representation Routing for Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Measures the complementarity, cost and rerun variability of six browser observation modes on WebArena and VisualWebArena. Its routing experiments distinguish an oracle cost opportunity from policies that fail to robustly outperform a fixed observation mode.", "topics": ["screen-understanding", "efficiency", "evaluation", "robustness"], "updated_at": "2026-08-06", "url": "https://arxiv.org/abs/2608.06171"}
{"aliases": [], "authors": ["Xiaoou Liu", "Longchao Da", "Hanyang Chen", "Yuan Ling", "Hua Wei"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4d99685a75d2023db5c9ad943e103ec779eebd037fe7a113848d8a31e08fd68b", "submitted_at": "2026-08-07T21:47:43Z", "updated_at": "2026-08-07T21:47:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.07775v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:00:40.283922+00:00", "role": "primary", "sha256": "ddd5bd966c7430bd9d4fd0269e7d3d46d8839fa5bfa57d601b97b4dc735cfc9a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:00:40.283922+00:00", "role": "primary", "sha256": "ddd5bd966c7430bd9d4fd0269e7d3d46d8839fa5bfa57d601b97b4dc735cfc9a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:10:44.898646+00:00", "role": "primary", "sha256": "d4afe27efcce2a3b9401f9677304db3cdedfa66f537293bb742bd1abfb16ee59", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/pdf/2608.07775v1"}], "id": "arxiv:2608.07775", "identifiers": {"arxiv": "2608.07775"}, "kind": "paper", "last_verified_at": "2026-10-18T10:00:40.283922+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.07775"}], "name": "AndroidReality: How Far Are Mobile Agents from the Real World?", "notes": "The reviewed protocol normally activates one perturbation type at a time while preserving task semantics and reward. Results do not establish robustness to arbitrary combined deployment disturbances; experiments were not independently reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (3.1 Perturbation formulation; 3.2 Injection; 3.3 Evaluation protocol; 4.3 Error analysis); not an entire-paper reading or reproduction claim.", "published_at": "2026-08-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Extends AndroidWorld with controlled state, transition, and action perturbations to test mobile-agent robustness, and studies a training-free introspective recovery mechanism.", "topics": ["evaluation", "robustness", "self-reflection", "mobile-navigation"], "updated_at": "2026-08-07", "url": "https://arxiv.org/abs/2608.07775"}
{"aliases": [], "authors": ["Zejun Xu", "Taiyi Chen", "Jin Li", "Yongtong Gu", "Qi Cheng", "Aixuan Lv", "Shuai Zhu", "Pengfei Zhu", "Kaichen Yang", "Boyu Sun", "Yixian Yang", "Mulong Xie", "Xin Liu", "Dagang Li", "Xiaoteng Ma", "Hongru Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d0e469cd7d84080e11df46d826460bea79eae71b8fa5786cacd96aac02b8b0a8", "submitted_at": "2026-08-09T01:08:19Z", "updated_at": "2026-08-09T01:08:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.08392v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.08392", "identifiers": {"arxiv": "2608.08392"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.08392"}], "name": "CAP: A Scalable Benchmark for Evaluating Cross-Site Browser Agents with Complex Actions and Perception", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "CAP builds cross-site browser workflows by decomposing websites into operation and perception requirements and recomposing them into tasks. Its judge-based evaluation diagnoses complex interactions and visual-perception failures across real websites.", "topics": ["evaluation", "web-navigation", "cross-app", "screen-understanding"], "updated_at": "2026-08-09", "url": "https://arxiv.org/abs/2608.08392"}
{"aliases": [], "authors": ["Yuke Li", "Xuehan Hou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f05f128cccb514987b7defb52c8f43b45d917e0e0f9570db4e75c4fe23b38533", "submitted_at": "2026-08-10T14:29:07Z", "updated_at": "2026-08-10T14:29:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.09654v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.09654", "identifiers": {"arxiv": "2608.09654"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.09654"}], "name": "Hallucination-Free GUI Grounding via Regression-Free Layout-Aware Matching", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "This framework separates frozen-model instruction expansion from a trained layout-aware matcher that localizes GUI elements without coordinate regression. Binary text/icon supervision supports the matcher, with evaluations on GUI grounding and Mind2Web element-selection tasks.", "topics": ["grounding", "screen-understanding", "evaluation"], "updated_at": "2026-08-10", "url": "https://arxiv.org/abs/2608.09654"}
{"aliases": [], "authors": ["Santosh Patapati"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d5ccb0d41798a1a3236e6f9b4361814cccdd40facf9b43bf0bf52932337484dc", "submitted_at": "2026-07-01T16:24:07Z", "updated_at": "2026-07-01T16:24:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.09944v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:00:45.064380+00:00", "role": "primary", "sha256": "f26fba818a44c6a7f541ab3b0341ca8f4798d837a425a2ac9fb0fc47a553b7bc", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:00:45.064380+00:00", "role": "primary", "sha256": "f26fba818a44c6a7f541ab3b0341ca8f4798d837a425a2ac9fb0fc47a553b7bc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:05:26.923383+00:00", "role": "primary", "sha256": "777197864eeaa34d180bf6b91f32408b7b3d8cc5c9faecc95eca276f5bd426de", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/html/2608.09944v1"}], "id": "arxiv:2608.09944", "identifiers": {"arxiv": "2608.09944"}, "kind": "paper", "last_verified_at": "2026-10-18T10:00:45.064380+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.09944"}], "name": "Navigation Alone Is Not Enough: Evaluating Explanatory Assistive UI Agents", "notes": "Packaged tasks use recorded linear transition graphs. Reported model experiments use a text-input validation subset, so their results are not evidence of screenshot-based navigation performance. No results were independently reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (3.3 Interface states and task annotations; 3.4 Evaluation; 4.1 Models and experimental setup; 5 Discussion); not an entire-paper reading or reproduction claim.", "published_at": "2026-07-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces NeXUI, an assistive UI-agent benchmark that evaluates navigation together with interface-grounded explanations and user oversight in recorded interface states.", "topics": ["evaluation", "human-in-the-loop", "web-navigation"], "updated_at": "2026-07-01", "url": "https://arxiv.org/abs/2608.09944"}
{"aliases": [], "authors": ["Mizanur Rahman", "Mohammed Saidul Islam", "Ridwan Mahbub", "Md Tahmid Rahman Laskar", "Shafiq Joty", "Enamul Hoque Prince"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e1089e57e0f0146f29a4a411c5c6f72da3c9f45be223fbbc5c9630315be6a680", "submitted_at": "2026-08-11T01:45:56Z", "updated_at": "2026-08-11T01:45:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.10366v1"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "kind", "topics"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:06:26Z", "role": "primary", "sha256": "2158cabc9a6211e0b75e8e3307f80ebc0d42c42fccc03d0143dcf574c2f663db", "supports": ["summary", "environments", "scope"], "url": "https://github.com/vis-nlp/DSAgentBench"}], "id": "arxiv:2608.10366", "identifiers": {"arxiv": "2608.10366"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:09Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.10366"}], "name": "DSAgentBench: Can Agents Automate End-to-End Data-Science Workflows in Real Computer Environments?", "notes": "The paper abstract and official project README were reviewed. The README documents an Ubuntu desktop and GUI action interface and includes LLM-assisted evaluators; this annotation does not assert that every evaluator is deterministic. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "DSAgentBench evaluates end-to-end data-science workflows spanning analysis, modeling and visualization inside a real desktop environment. Task evaluators check produced analytical and visual artifacts, while the released interface supports screenshot or accessibility observations and GUI actions.", "topics": ["evaluation", "workflow", "science", "long-horizon"], "updated_at": "2026-08-11", "url": "https://arxiv.org/abs/2608.10366"}
{"aliases": [], "authors": ["Ziqiang Wang", "Li Gu", "Zhixiang Chi", "Zhi Liu", "Seyed Mehdi Ayyoubzadeh", "Yuanhao Yu", "Yang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a2606ae96ca3ea99b5c7278485601f9bde3b64eefdd3bfecc2bbfc678c39b4fc", "submitted_at": "2026-08-11T21:00:46Z", "updated_at": "2026-08-11T21:00:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.11434v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}], "id": "arxiv:2608.11434", "identifiers": {"arxiv": "2608.11434"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.11434"}], "name": "Benchmarking LLM Judges for Mobile Agent Evaluation", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-08-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileJudgeBench compares completion judges against human labels on mobile-agent trajectories. It studies screenshot-sampling baselines, model-dependent error profiles and how judge quality relates to agent rankings and reinforcement-learning rewards.", "topics": ["evaluation", "reward-models", "robustness"], "updated_at": "2026-08-11", "url": "https://arxiv.org/abs/2608.11434"}
{"aliases": [], "authors": ["Thilo Reintjes", "Sivajeet Chand", "Derui Zhu", "Sushant Kumar Pandey", "Alexander Pretschner"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f8a6bd69d421bdfb9a21c22e715a59d0afb0b0cab75510f28ff569903ecffae5", "submitted_at": "2026-08-14T09:36:33Z", "updated_at": "2026-08-14T09:36:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.14131v1"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.14131", "identifiers": {"arxiv": "2608.14131"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.14131"}], "name": "LegacyWorld: Atomicity-Aware Evaluation of GUI Agents for Legacy Workflows", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "LegacyWorld evaluates Windows GUI workflows under an atomicity criterion: finish correctly or fail without persistent unintended changes. It compares expert prompts with prompts derived from recorded golden paths and separates useful completion, safe failure and non-atomic side effects.", "topics": ["evaluation", "safety-security", "workflow", "recording"], "updated_at": "2026-08-14", "url": "https://arxiv.org/abs/2608.14131"}
{"aliases": [], "authors": ["Xiaokai Yan", "Jingtao Ding", "Yong Li", "Zhiwen Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e2660e01921a78bed7280b34f35a1b587e62339c2e280d6294bbb6d6e666786e", "submitted_at": "2026-08-14T09:36:59Z", "updated_at": "2026-08-14T09:36:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.14132v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:00:40.283922+00:00", "role": "primary", "sha256": "ddd5bd966c7430bd9d4fd0269e7d3d46d8839fa5bfa57d601b97b4dc735cfc9a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:00:40.283922+00:00", "role": "primary", "sha256": "ddd5bd966c7430bd9d4fd0269e7d3d46d8839fa5bfa57d601b97b4dc735cfc9a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:05:22.211464+00:00", "role": "primary", "sha256": "8d6aa8484ff4a7cbaac589c2c220407ebf6ecd1e216015a248125abdb0f2ed2d", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/html/2608.14132v1"}], "id": "arxiv:2608.14132", "identifiers": {"arxiv": "2608.14132"}, "kind": "paper", "last_verified_at": "2026-10-18T10:00:40.283922+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.14132"}], "name": "Act2Intention: A Benchmark For Developing Active Mobile Agents Through Inferring User Intention from GUI Actions", "notes": "The benchmark combines real and generated trajectory material. Opportune intervention timing and real-user acceptance are stated limitations. Section 4.2.2 specifies Android emulator instances for generated action trajectories. No results were independently reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (3 Task definition; 4.1 Raw data collection; 4.2 Trajectory construction; 7 Limitations and future directions); not an entire-paper reading or reproduction claim.", "published_at": "2026-08-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces intention understanding, prediction, and confirmed execution tasks over mobile GUI trajectories, with a proactive agent combining personalized intention models and experience-guided actions.", "topics": ["human-in-the-loop", "planning", "mobile-navigation", "evaluation"], "updated_at": "2026-08-14", "url": "https://arxiv.org/abs/2608.14132"}
{"aliases": [], "authors": ["Sujin Chen", "Lijun Li", "Tianyi Du", "Jing Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c49a67b38314152b83e793b9ade7ce4102a44aa27c331504a8758b94df906c67", "submitted_at": "2026-08-18T11:33:22Z", "updated_at": "2026-08-18T11:33:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.17659v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.17659", "identifiers": {"arxiv": "2608.17659"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.17659"}], "name": "MobileWorldSafety: Benchmarking GUI Agent Safety Against Environmental Injection Attacks in Android Apps", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates environmental injection risks in Android tasks using programmatic risk checks and model-based adjudication of ambiguous cases.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-08-18", "url": "https://arxiv.org/abs/2608.17659"}
{"aliases": [], "authors": ["Tianchen Guan", "Xinlei Lin", "Royce Cheng-Yue", "Xiangjun Wang", "Shuyan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "de4889c021f35c169092e17801fec414cce21b37c74a4ff72af6a6c67a1a0a25", "submitted_at": "2026-08-18T20:38:26Z", "updated_at": "2026-08-18T20:38:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.18307v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.18307", "identifiers": {"arxiv": "2608.18307"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.18307"}], "name": "ComponentBench: Diagnosing Component-Level Failures in Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "ComponentBench evaluates short interactions with canonical web UI components using verified tasks and matched human reference trajectories. Its diagnostic pipeline compares observation/action representations and analyzes structural difficulty, failure patterns and interaction efficiency.", "topics": ["evaluation", "screen-understanding", "efficiency", "robustness"], "updated_at": "2026-08-18", "url": "https://arxiv.org/abs/2608.18307"}
{"aliases": [], "authors": ["Qijia Chen", "Giulio Jacucci"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cabd094270d01b20df7b029be29c5d460d10beb84e43e3e783f4f11997b74114", "submitted_at": "2026-08-22T06:30:03Z", "updated_at": "2026-08-22T06:30:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.21794v1"}}, "environments": ["web", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2608.21794", "identifiers": {"arxiv": "2608.21794"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.21794"}], "name": "Lexical Coupling in GUI Element Grounding: Sentence Embeddings Track Labels across Mobile and Web", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies whether text-embedding GUI grounding scores reflect semantic matching or recovery of visible labels. Same-screen ranking comparisons with lexical baselines expose differences between deployable fusion, oracle selection, and performance on label-poor targets.", "topics": ["grounding", "evaluation", "screen-understanding"], "updated_at": "2026-08-22", "url": "https://arxiv.org/abs/2608.21794"}
{"aliases": [], "authors": ["Md Abrar Jahin", "Md Rizwan Parvez"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "970441704314fabffc46cb4a340dd6e0bde0e84ab9b7fcd2db2c9a8be9dc61d8", "submitted_at": "2026-08-22T08:04:59Z", "updated_at": "2026-08-27T07:03:01Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.21832v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.21832", "identifiers": {"arxiv": "2608.21832"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.21832"}], "name": "GUI-Primitives: Diagnosing Spatial Reasoning Failures in Vision-Language GUI Grounding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUI-Primitives, a paired-instruction benchmark that changes a spatial relation while holding the screenshot and anchor fixed. Its diagnostics separate failure to localize either candidate from failure to choose the candidate satisfying the relation.", "topics": ["grounding", "evaluation", "screen-understanding"], "updated_at": "2026-08-27", "url": "https://arxiv.org/abs/2608.21832"}
{"aliases": [], "authors": ["Guo Gan", "Yilun Zhao", "Cong Chen", "Jinbiao Wei", "Tingyu Song", "Zheyuan Yang", "Lin Fu", "Hong Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "51901009de5898c52a1b9d4958b698808a8f8b850d19afa32dd7a3f32d42a851", "submitted_at": "2026-08-25T05:57:25Z", "updated_at": "2026-08-25T05:57:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.24099v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.24099", "identifiers": {"arxiv": "2608.24099"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.24099"}], "name": "Are Android GUI Agents Robust Against Runtime Anomalies? AnTrap: Evaluating Agents in Dynamic Adversarial Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AnTrap, which injects task-solvable runtime anomalies into Android GUI-agent trajectories using a layered failure taxonomy. It compares model robustness and adversarial reinforcement learning to distinguish recoverable interface traps from persistent contextual failures.", "topics": ["evaluation", "robustness", "reinforcement-learning"], "updated_at": "2026-08-25", "url": "https://arxiv.org/abs/2608.24099"}
{"aliases": [], "authors": ["Tao Xiong", "Xavier Hu", "Wenkai Wang", "Qinzhuo Wu", "Changqiao Wu", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "76cd09f70bfec0d2a49cddae8adcf32ca37fb39bcade7c1d36d5f3dc3fa6bfa7", "submitted_at": "2026-08-25T07:40:12Z", "updated_at": "2026-08-25T07:40:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.24174v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2608.24174", "identifiers": {"arxiv": "2608.24174"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.24174"}], "name": "Task-Adaptive Rubrics for GUI Reward Modeling", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AdaptRubric, which retrieves reusable criteria for a GUI task family and then generates instruction-specific checks for concrete values and constraints. It evaluates these adaptive rubrics in offline reward assessment and online agent optimization.", "topics": ["reward-models", "evaluation", "reinforcement-learning"], "updated_at": "2026-08-25", "url": "https://arxiv.org/abs/2608.24174"}
{"aliases": [], "authors": ["Joy Chen", "Alejandro Castillejo Munoz", "Pierluca D'Oro", "Yuxuan Sun", "Chloe Evans", "Joseph Tighe"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "50c26a7ab741ecb3cb855305ee4f7d5b5729296248d650c469600aeb73101781", "submitted_at": "2026-08-25T23:21:57Z", "updated_at": "2026-09-15T04:43:00Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.26204v2"}}, "environments": ["desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.26204", "identifiers": {"arxiv": "2608.26204"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.26204"}], "name": "ADeptS-Bench: Measuring the Trustworthiness of Computer Use Agents Across Devices", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates interface-embedded threats and whether computer-use agents seek clarification when mobile or desktop instructions are ambiguous.", "topics": ["safety-security", "evaluation", "human-in-the-loop"], "updated_at": "2026-09-15", "url": "https://arxiv.org/abs/2608.26204"}
{"aliases": [], "authors": ["Rui Xie", "Lu Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e1626d5190f145c668c9b7f363b7f84904c12daae485e63b858e440ded5ea6df", "submitted_at": "2026-08-27T11:38:19Z", "updated_at": "2026-08-27T11:38:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.26991v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:35Z", "role": "primary", "sha256": "6499b4f4917eec0765d16d6e26a08d7edc5146a30753b19cf8218517656bb070", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.16211%2C2608.16391%2C2608.17067%2C2608.17975%2C2608.19741%2C2608.21500%2C2608.22301%2C2608.22510%2C2608.23035%2C2608.23320%2C2608.26238%2C2608.26546%2C2608.26991%2C2608.27260%2C2608.29814%2C2608.31076%2C2609.02292%2C2609.04173%2C2609.04298%2C2609.05776%2C2609.06059%2C2609.07529%2C2609.08149%2C2609.08418%2C2609.09187%2C2609.11561%2C2609.12533%2C2609.15096%2C2609.15544%2C2609.20732%2C2609.21293%2C2609.23986%2C2609.24971%2C2609.24997%2C2609.27455%2C2609.29647%2C2609.29964%2C2609.30798%2C2609.32835%2C2609.34428%2C2609.35002%2C2609.35232%2C2609.35394%2C2609.35581%2C2609.36071%2C2609.36352%2C2609.36777%2C2609.37017%2C2609.38923%2C2609.39222&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:35Z", "role": "primary", "sha256": "6499b4f4917eec0765d16d6e26a08d7edc5146a30753b19cf8218517656bb070", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2608.16211%2C2608.16391%2C2608.17067%2C2608.17975%2C2608.19741%2C2608.21500%2C2608.22301%2C2608.22510%2C2608.23035%2C2608.23320%2C2608.26238%2C2608.26546%2C2608.26991%2C2608.27260%2C2608.29814%2C2608.31076%2C2609.02292%2C2609.04173%2C2609.04298%2C2609.05776%2C2609.06059%2C2609.07529%2C2609.08149%2C2609.08418%2C2609.09187%2C2609.11561%2C2609.12533%2C2609.15096%2C2609.15544%2C2609.20732%2C2609.21293%2C2609.23986%2C2609.24971%2C2609.24997%2C2609.27455%2C2609.29647%2C2609.29964%2C2609.30798%2C2609.32835%2C2609.34428%2C2609.35002%2C2609.35232%2C2609.35394%2C2609.35581%2C2609.36071%2C2609.36352%2C2609.36777%2C2609.37017%2C2609.38923%2C2609.39222&max_results=50"}], "id": "arxiv:2608.26991", "identifiers": {"arxiv": "2608.26991"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.26991"}], "name": "ASIL: Replacing Screenshot-and-Click with Structured State and Semantic Actions", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "ASIL exposes application state as structured JSON and provides code-executable semantic actions. Its application benchmark compares this interface with screenshot control and application-native APIs, and studies supervised and reinforcement learning on the structured interface.", "topics": ["automation", "tool-integration", "evaluation", "reinforcement-learning", "imitation-learning"], "updated_at": "2026-08-27", "url": "https://arxiv.org/abs/2608.26991"}
{"aliases": [], "authors": ["Divake Kumar", "Sina Tayebati", "Devashri Naik", "Amanda Sofie Rios", "Nilesh Ahuja", "Omesh Tickoo", "Ranganath Krishnan", "Amit Ranjan Trivedi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "14864aae8f977d474ff3e7dd3e38951374862abfb8077395f2dc5a9e1cc5a489", "submitted_at": "2026-08-28T00:50:47Z", "updated_at": "2026-08-28T00:50:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.27808v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.27808", "identifiers": {"arxiv": "2608.27808"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.27808"}], "name": "CURA: Certified Runtime Alarms for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CURA, a telemetry-based runtime monitor that turns computer-use trajectories into sequential failure alarms and can trigger mid-task oversight. Its certificate controls false alarms, while the study separately reports detection performance and limits of retrospective monitoring.", "topics": ["observability", "safety-security", "evaluation"], "updated_at": "2026-08-28", "url": "https://arxiv.org/abs/2608.27808"}
{"aliases": [], "authors": ["Mingshuo Wang", "Hanqing Guo", "Huining Li", "Yuliang Fu", "Jing Xu", "Chenhan Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7b17f858c9d6992f6a8d8f21cc79d468b0f1c81fc44084045791d9c3acacd6d0", "submitted_at": "2026-08-30T20:55:21Z", "updated_at": "2026-08-30T20:55:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.30038v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2608.30038", "identifiers": {"arxiv": "2608.30038"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.30038"}], "name": "ActReal: System-Level Mobile Agents Challenge Mobile Automation Detection", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces ActReal, which translates semantic mobile-agent actions into coordinated touch and sensor signals to study automation-detection weaknesses. It evaluates detectors that observe touch events alone or together with inertial measurements under a privileged executor threat model.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2026-08-30", "url": "https://arxiv.org/abs/2608.30038"}
{"aliases": [], "authors": ["Chen Xiong", "Zhiyuan He", "Pin-Yu Chen", "Stjepan Picek", "Tsung-Yi Ho"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "87c3f58eebe2db437b1166ff8e83b336db4cde37684a7cd86f61c4b34fca7237", "submitted_at": "2026-08-31T03:39:43Z", "updated_at": "2026-10-03T10:41:41Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.30207v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.30207", "identifiers": {"arxiv": "2608.30207"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.30207"}], "name": "SIR: Self-improving Red-teaming for Compute Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SIR, which iteratively refines a reusable library of prompt-injection strategies from failed computer-use attack trajectories. Its evaluation uses deterministic system-state checks and counts attacks only when the adversarial objective and legitimate task both complete.", "topics": ["safety-security", "evaluation", "data-generation"], "updated_at": "2026-10-03", "url": "https://arxiv.org/abs/2608.30207"}
{"aliases": [], "authors": ["Lin Fu", "Zheyuan Yang", "Tianhui Zhang", "Jinbiao Wei", "Guo Gan", "Boxu Liu", "Yilun Zhao", "Yu Rong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dba8eb3f30a7e88ad973080d07fbf272f63400f4f8fa858d5f2657735ce093c0", "submitted_at": "2026-08-30T05:19:43Z", "updated_at": "2026-08-30T05:19:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.00048v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2609.00048", "identifiers": {"arxiv": "2609.00048"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.00048"}], "name": "GUI-CC: Benchmarking Contextual Consistency of GUI World Models as Agent Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUI-CC, a benchmark for assessing GUI world models as multi-step agent environments. Separate reference-action and agent-loop tracks test whether generated screens preserve context, plausible transitions, and task progress throughout interaction.", "topics": ["world-models", "evaluation", "long-horizon"], "updated_at": "2026-08-30", "url": "https://arxiv.org/abs/2609.00048"}
{"aliases": [], "authors": ["Satwik Ram Kodandaram", "Monalika Padma Reddy", "Xiaojun Bi", "Jiawei Zhou", "I. V. Ramakrishnan", "Vikas Ashok"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "deb130d93bbbd6dfdc04c72d232a9dcfa1e92b0586831e5864070e7fe0de14dd", "submitted_at": "2026-09-01T00:44:19Z", "updated_at": "2026-09-01T00:44:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.00524v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.00524", "identifiers": {"arxiv": "2609.00524"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.00524"}], "name": "Are We There Yet? Assessing Computer-Use Agents for Blind Users' Accessible Interaction with Desktop Applications", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "A diary study deploys the screen-reader-accessible OLLA computer-use prototype with blind users across desktop applications. Recorded commands and interaction traces support model comparisons and analysis of grounding, planning, constraint tracking and termination failures.", "topics": ["evaluation", "human-in-the-loop", "grounding"], "updated_at": "2026-09-01", "url": "https://arxiv.org/abs/2609.00524"}
{"aliases": [], "authors": ["Sitong Pan", "Yipeng Shen", "Yilin Lu", "Caiwen Ding", "Lu Cheng", "Qianwen Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "967f1cdda69349da3280827c0b90301e4942c44398902ab6cfb868f1d3e94a84", "submitted_at": "2026-09-02T03:34:47Z", "updated_at": "2026-09-02T03:34:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.02057v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.02057", "identifiers": {"arxiv": "2609.02057"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.02057"}], "name": "Monitoring Web Agents Without Internal Signals: Observable Trajectories and Key-Step Supervision", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Predicts web-agent failure risk from observable trajectory prefixes using cross-step behavior and black-box consistency queries. Key-step labels preserve valid early prefixes of failed runs, and evaluation tests early intervention and transfer across website categories.", "topics": ["observability", "evaluation", "robustness", "interaction-history"], "updated_at": "2026-09-02", "url": "https://arxiv.org/abs/2609.02057"}
{"aliases": [], "authors": ["Yimeng Liu", "Hua Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b462dddd45e6b04a3bf6fa0bc73576a23a5f9617f61aeb257a845c0f456f731d", "submitted_at": "2026-09-02T06:06:43Z", "updated_at": "2026-09-02T06:06:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.02154v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2609.02154", "identifiers": {"arxiv": "2609.02154"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.02154"}], "name": "WiP: Characterizing and Defending Against Mobile-Agent-Driven MFA Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies how mobile automation can combine authentication factors within one agent environment and evaluates motion sensing as a user-presence signal.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-09-02", "url": "https://arxiv.org/abs/2609.02154"}
{"aliases": [], "authors": ["Kelvin Li", "Dhruv Pendharkar", "Anish Pahilajani", "Chuyi Shang", "Leon Oks", "Leonid Karlinsky", "Rogerio Feris", "Trevor Darrell", "Roei Herzig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8703474cb4f577b307731cf1a986f76f800470fde625b6ca4b3569d30e74f903", "submitted_at": "2026-09-02T17:59:40Z", "updated_at": "2026-09-02T17:59:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.02885v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.02885", "identifiers": {"arxiv": "2609.02885"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.02885"}], "name": "Discriminative World Models for Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains discriminative web world models to distinguish the true next state from outcomes of alternative actions. Branching web trajectories support training, followed by tests of predicted-state matching, process-reward action ranking and end-to-end WebArena-Lite selection.", "topics": ["world-models", "reward-models", "planning", "data-generation", "evaluation"], "updated_at": "2026-09-02", "url": "https://arxiv.org/abs/2609.02885"}
{"aliases": [], "authors": ["Zhaoyuan Huang", "Tianjie Ju", "Pengzhou Cheng", "Zheng Wu", "Yansi Li", "Chuanbiao Song", "Jun Lan", "Huijia Zhu", "Weiqiang Wang", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a59f7d54845de6efebf6365a188e61e41fdf31e9ebe894b1d01ce7b86d6be4ec", "submitted_at": "2026-09-03T06:48:41Z", "updated_at": "2026-09-03T06:48:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.03438v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.03438", "identifiers": {"arxiv": "2609.03438"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.03438"}], "name": "Do GUI Agents Know When Not to Act? Enabling Conflict-Aware Termination for Multimodal GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "CONFLICTGUI evaluates termination under instruction conflicts and mismatches between instructions and interface context. CONFLICTGUARD couples feasibility checks with conditional action modulation to reduce unnecessary execution while tracking normal-task behavior.", "topics": ["safety-security", "evaluation", "robustness", "planning"], "updated_at": "2026-09-03", "url": "https://arxiv.org/abs/2609.03438"}
{"aliases": [], "authors": ["Weide Zhan", "Qumu Shaqu", "Yuanqing Liu", "Peng Zhang", "Jiahao Liu", "Kam Him Lam", "Ning Gu", "Zhan Hu", "Tun Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "61587798aefd8e9592ddaef756a13e02f0e4a987e4068ef582d0aaf66e8eb680", "submitted_at": "2026-09-04T08:09:47Z", "updated_at": "2026-09-04T08:09:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.04850v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2609.04850", "identifiers": {"arxiv": "2609.04850"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.04850"}], "name": "ElderBench: Benchmarking Autonomous Mobile Agents for Older Adults", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "ElderBench evaluates mobile GUI agents on smartphone requests elicited from older adults, including indirect and underspecified language. Online and offline tests, instruction normalization and linguistic analyses examine how these request patterns contribute to failures.", "topics": ["evaluation", "human-in-the-loop", "mobile-navigation", "robustness"], "updated_at": "2026-09-04", "url": "https://arxiv.org/abs/2609.04850"}
{"aliases": [], "authors": ["Longtao Hu", "Xiao Liang", "Linchao Zhu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "78db974038e0dde9a1347132382573fa9e2aafb1b0e1a574da3d769574329be9", "submitted_at": "2026-09-04T08:29:45Z", "updated_at": "2026-09-04T08:29:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.04869v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.04869", "identifiers": {"arxiv": "2609.04869"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.04869"}], "name": "From Interaction Traces to Persistent Skills: Online Evolution for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "An online skill-evolution framework turns GUI interaction traces and evaluator feedback into a versioned library of reusable procedures without updating model parameters. Controlled OSWorld comparisons examine both the benefits of shared procedural memory and its domain-dependent instability under repeated revision.", "topics": ["memory", "workflow", "evaluation", "training-free"], "updated_at": "2026-09-04", "url": "https://arxiv.org/abs/2609.04869"}
{"aliases": [], "authors": ["Jintian Feng", "Long Chen", "Xiao Yu", "Jiayi Dai", "Chenglong Liu", "Haoru Wang", "Zizhen Xue", "Yuxuan Shi", "Ziyang Wang", "Yichen Gong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b4090a2c7e7730ea415564b6b418d85aad77a74a222cef35435b0aa93beafd7f", "submitted_at": "2026-09-07T16:25:13Z", "updated_at": "2026-09-07T16:25:13Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.07712v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.07712", "identifiers": {"arxiv": "2609.07712"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.07712"}], "name": "APPSim-Bench: Bridging Real-world Apps and Reproducible Evaluation for Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces controlled simulated mobile apps with deterministic backend data and outcome checks for reproducible evaluation of realistic interactions.", "topics": ["evaluation", "sandbox", "mobile-navigation"], "updated_at": "2026-09-07", "url": "https://arxiv.org/abs/2609.07712"}
{"aliases": [], "authors": ["Zhihao Liu", "Hongyu Sun", "Zhiyuan Fu", "Xiaonan Duan", "Jice Wang", "Shangru Zhao", "Weizhi Meng", "Wuxin Yang", "Yangfan Zhou", "Yuqing Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5a3a7a81ffe43885d9c8362df4c8aff08000304fe60477f484a732caf35124b2", "submitted_at": "2026-09-06T03:49:45Z", "updated_at": "2026-09-06T03:49:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.09212v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.09212", "identifiers": {"arxiv": "2609.09212"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.09212"}], "name": "AgentHijack: Visual Patch Attacks on Multimodal Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "AgentHijack evaluates whether visual patch effects propagate from screenshots through model outputs and action parsing into real environment changes. Its controlled web-page experiments distinguish intermediate attack signals from verified end-to-end consequences.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": "2026-09-06", "url": "https://arxiv.org/abs/2609.09212"}
{"aliases": [], "authors": ["Xiaolei Li", "Jialun Cao", "Zhijian Hou", "Yuzhi Zhao", "Yepang Liu", "Shing-Chi Cheung"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3307d90ef09c16afaee160eacc9d2994d142be9b9a26ce9c4a57f61345e2e8ec", "submitted_at": "2026-09-09T11:04:53Z", "updated_at": "2026-09-09T11:04:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.10031v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2609.10031", "identifiers": {"arxiv": "2609.10031"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.10031"}], "name": "GraphDroid: Asynchronous LLM-Based Mobile App GUI Testing via History-Aware Exploration and Hybrid Intent Fulfillment", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "GraphDroid combines clustered memory of visited GUI states with asynchronous test-intent generation. A hybrid executor reserves LLM calls for complex intents and uses heuristics for simpler ones, evaluated through Android app coverage, bug discovery and crash detection.", "topics": ["evaluation", "mobile-navigation", "memory", "interaction-history", "efficiency"], "updated_at": "2026-09-09", "url": "https://arxiv.org/abs/2609.10031"}
{"aliases": [], "authors": ["Zixiang Chen", "Yuheng Lu", "Zihao Cheng", "Zeming Liu", "Jizeng Bai", "Ziye Huang", "Zhiyin Lin", "Zihan Li", "Yuhang Guo", "Yunhong Wang", "Haifeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5ec17516bce87a8e349983acac14415a6ceeb12c167fce57aa37d6157cd3cd83", "submitted_at": "2026-09-09T16:59:59Z", "updated_at": "2026-09-09T16:59:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.10451v1"}}, "environments": ["desktop", "android", "windows", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.10451", "identifiers": {"arxiv": "2609.10451"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.10451"}], "name": "JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "JarvisGUI composes typed input-output task transformations into dynamic workflows spanning Android, Windows and Ubuntu devices. Its virtual-environment benchmark tests transfer of intermediate results, shared state and cross-device dependencies.", "topics": ["evaluation", "cross-app", "planning", "long-horizon", "state-tracking"], "updated_at": "2026-09-09", "url": "https://arxiv.org/abs/2609.10451"}
{"aliases": [], "authors": ["Qihang Cen", "Tianshuo Cong", "Da Song", "Xinlei He", "Jiaxing Song", "Ke Xu", "Qi Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1eda33b358d4c4caf8c1fb0b1bd92698c3b37cee995d820f40ecffe82010edb4", "submitted_at": "2026-09-12T10:55:54Z", "updated_at": "2026-09-12T10:55:54Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.13873v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.13873", "identifiers": {"arxiv": "2609.13873"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.13873"}], "name": "PriMobiBench: Characterizing Visual Privacy Leakage in VLM-Driven Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "PriMobiBench evaluates direct privacy leakage and user profiling from mobile GUI screenshot trajectories, supported by the MobiLeak trace dataset. It also tests masking sensitive, task-irrelevant interface elements before cloud processing and measures the resulting privacy and utility trade-off.", "topics": ["evaluation", "safety-security", "screen-understanding", "mobile-navigation"], "updated_at": "2026-09-12", "url": "https://arxiv.org/abs/2609.13873"}
{"aliases": [], "authors": ["Yizhou Liu", "Fei Tang", "Yuchen Yan", "Zhengxi Lu", "Songqin Nong", "Tao Jiang", "Wenhao Xu", "Wenqi Zhang", "Weiming Lu", "Jun Xiao", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3e2081a382ae815b48eb3166d18ee7c6e53afb4cd6d75538f3ff3e29f2ed5954", "submitted_at": "2026-09-14T09:56:10Z", "updated_at": "2026-09-14T09:56:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.15307v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.15307", "identifiers": {"arxiv": "2609.15307"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.15307"}], "name": "Learning from Reliable Negatives: Confidence-Anchored Test-Time Adaptation for GUI Grounding", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Confidence-Anchored Learning uses coordinate-token confidence to filter pseudo-supervision for GUI grounding. Its negative-only variant performs label-free test-time updates to reduce reliance on potentially incorrect positive examples.", "topics": ["grounding", "robustness", "evaluation"], "updated_at": "2026-09-14", "url": "https://arxiv.org/abs/2609.15307"}
{"aliases": [], "authors": ["Zihan Dong", "Yuanzhe Liu", "Zhiyuan Ma", "Qishi Zhan", "Dehan Kong", "Guohao Li", "Kaixin Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0f5fd0a3024311f9aba44a2971c9c752ee2c38444139823746b90a7c9a6d5ed4", "submitted_at": "2026-09-14T19:19:05Z", "updated_at": "2026-09-21T14:30:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.16251v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.16251", "identifiers": {"arxiv": "2609.16251"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.16251"}], "name": "CADWorld: Computer-Use Benchmark for Long-Horizon Computer-Aided Design", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "CADWorld evaluates screenshot-and-GUI agents on long FreeCAD workflows, from modeling and assembly to manufacturing and simulation. Executable checks inspect saved engineering artifacts for geometry, constraints, construction structure and downstream state.", "topics": ["evaluation", "long-horizon", "workflow", "science"], "updated_at": "2026-09-21", "url": "https://arxiv.org/abs/2609.16251"}
{"aliases": [], "authors": ["Heng Li", "Fulin Zhao", "Zhe Geng", "Zhiyuan Yao", "Wei Yuan", "Xiapu Luo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b86516b7695d23ffc9179c0829209dc28b96c3c819525eab62a293a68ab5daf0", "submitted_at": "2026-09-15T07:02:10Z", "updated_at": "2026-09-15T07:02:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.16732v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}], "id": "arxiv:2609.16732", "identifiers": {"arxiv": "2609.16732"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:17Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.16732"}], "name": "When Agents See Differently: Exposing UI Desynchronization Threats in Mobile Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Investigates mismatches between what humans see on a mobile display and what agents receive through screenshots or accessibility metadata. An automated evaluation framework embeds instruction-agnostic perturbations into functional APK clones and compares static and dynamic agent deviations with human noticeability.", "topics": ["safety-security", "screen-understanding", "evaluation", "robustness"], "updated_at": "2026-09-15", "url": "https://arxiv.org/abs/2609.16732"}
{"aliases": [], "authors": ["Kratika Bhagtani", "Kusha Sridhar", "Maziyar Baran Pouyan", "Yuying Zhao", "Eugene Siow"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ab0dbea71d0b8bfe035d3b0a71d6ed99fbd261202976cdc322a0769074827d76", "submitted_at": "2026-09-15T22:23:17Z", "updated_at": "2026-09-23T01:31:01Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.17885v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.17885", "identifiers": {"arxiv": "2609.17885"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.17885"}], "name": "ERPBench: A State-Grounded Evaluation Paradigm for Computer-Use Agents in Enterprise Software", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "ERPBench evaluates screenshot-only agents in enterprise software using database values to check the business records produced by their actions. It pairs these state-based checks with a harness for human approval and analyzes cases where visible completion masks incorrect persisted data.", "topics": ["evaluation", "workflow", "human-in-the-loop", "robustness"], "updated_at": "2026-09-23", "url": "https://arxiv.org/abs/2609.17885"}
{"aliases": [], "authors": ["Yinzhu Quan", "Zefang Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5e7fb5704932a99147ff7c64a82a6a4041565f3edf6506a6252f75f832834d96", "submitted_at": "2026-09-17T00:29:46Z", "updated_at": "2026-10-08T04:53:11Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.19523v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.19523", "identifiers": {"arxiv": "2609.19523"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.19523"}], "name": "EconSkills: Studying Skill Transfer and Retrieval for Web Agents on Live Economic Data", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "EconSkills converts verified economic-data browsing traces into parameterized procedures with navigation, verification and recovery guidance. Its evaluation separates known-skill transfer from library retrieval, exposing the limits of approximate matches to new tasks.", "topics": ["memory", "web-navigation", "evaluation", "workflow"], "updated_at": "2026-10-08", "url": "https://arxiv.org/abs/2609.19523"}
{"aliases": [], "authors": ["Haya Halimeh", "Sascha Kaltenpoth", "Kevin Bösch", "Oliver Müller"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "70a7588abcf345418ed759fb2bdb96fc60c6e1444a3e0dc98766510c755f8478", "submitted_at": "2026-09-17T07:54:15Z", "updated_at": "2026-09-17T07:54:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.19843v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.19843", "identifiers": {"arxiv": "2609.19843"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.19843"}], "name": "A Dual-Process Perspective on Nudge Susceptibility in LLM-Based GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies how automatic defaults and reflective social-influence cues affect GUI-agent shopping decisions under different reasoning configurations. A randomized simulated-shopping experiment finds that reasoning changes the type of susceptibility rather than uniformly removing interface influence.", "topics": ["safety-security", "evaluation", "ui-design", "robustness"], "updated_at": "2026-09-17", "url": "https://arxiv.org/abs/2609.19843"}
{"aliases": [], "authors": ["Shuai Bai", "Jiayong Deng", "Sicheng Fan", "Yikun Fu", "Chang Gao", "Xuhao Hu", "Mianqiu Huang", "Yizhen Jiang", "Yuheng Jing", "Dehui Kong", "Keliang Li", "Ning Li", "Wanli Li", "Dayiheng Liu", "Dunjie Lu", "Changwei Luo", "Que Shen", "Zheyuan Wang", "Zijian Wang", "Jie Wu", "Gao Wu", "Zhihui Xie", "Rui Xie", "Haiyang Xu", "An Yang", "Jiakang Yuan", "Yanming Zhang", "Jiajun Zhang", "Xi Zhang", "Zhenru Zhang", "Zhuo Zhen", "Mingkang Zhu", "Bowen Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3b312c5a4d638ec9f9a7408fd9ce2d8cbf74fa462d48c8aa37f59e4ad3bd8edc", "submitted_at": "2026-09-18T17:00:56Z", "updated_at": "2026-09-21T07:09:10Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.22000v2"}}, "environments": ["desktop", "linux", "macos", "windows", "android", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.22000", "identifiers": {"arxiv": "2609.22000"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.22000"}], "name": "RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "RecreationWorld gives hybrid agents running reference applications to explore and recreate using native GUI controls and coding tools. Its multi-platform benchmark checks interaction-dependent behavior with programmatic and visual assertions, while trajectory-training transfer is evaluated separately.", "topics": ["evaluation", "data-generation", "tool-integration", "workflow"], "updated_at": "2026-09-21", "url": "https://arxiv.org/abs/2609.22000"}
{"aliases": [], "authors": ["Yunxiang Li", "Xixin Wu", "Helen Meng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "64936af65d0ca351b19da3f63325f2bc2a5c9e0951e5a7806cd4dd82b18784a7", "submitted_at": "2026-09-21T08:42:08Z", "updated_at": "2026-09-21T08:42:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.24277v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.24277", "identifiers": {"arxiv": "2609.24277"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.24277"}], "name": "How Many Pixels Is a Digit Worth? Place-Aware Coordinate Entropy for GUI Agent Confidence Estimation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "PACE weights coordinate-digit entropy by place value to estimate confidence in GUI clicks from one forward pass. It evaluates correctness ranking and selective accuracy against repeated-sampling alternatives on ScreenSpot-Pro and ScreenSpot-v2.", "topics": ["grounding", "efficiency", "evaluation", "robustness"], "updated_at": "2026-09-21", "url": "https://arxiv.org/abs/2609.24277"}
{"aliases": [], "authors": ["Zhilin Wang", "Shaokun Zhang", "Yifan Zhang", "Hao Zhang", "Jin Xu", "Binfeng Xu", "Jian Hu", "Yunheng Zou", "Karan Sapra", "Andrew Tao", "Jan Kautz", "Yi Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0d3cb920a677108910530788eb68c263127dba9510274f182910aed9c2f848c9", "submitted_at": "2026-09-21T16:55:24Z", "updated_at": "2026-09-21T16:55:24Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.24890v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.24890", "identifiers": {"arxiv": "2609.24890"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.24890"}], "name": "OSWorld-Pro: Process-based Evaluation for Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Adds annotated intermediate subgoals to computer-use tasks and uses process-oriented evaluation to identify action and progress failures.", "topics": ["evaluation", "observability"], "updated_at": "2026-09-21", "url": "https://arxiv.org/abs/2609.24890"}
{"aliases": [], "authors": ["Yuxuan Li", "Will Epperson", "Wesley Deng", "Zezhou Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8d241c1ea9731bc815eef5df9c409d410608adb052616406d955d7d33cecd9b9", "submitted_at": "2026-09-23T03:00:00Z", "updated_at": "2026-09-27T20:49:01Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.27273v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.27273", "identifiers": {"arxiv": "2609.27273"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.27273"}], "name": "CAVEAT: Towards Robust Computer-Use Agents in Incentive-Misaligned Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CAVEAT, a controlled marketplace benchmark for testing whether delegated agents preserve user objectives when interfaces favor other outcomes. It diagnoses decision failures and studies harness interventions and targeted training to improve resistance to marketplace steering.", "topics": ["evaluation", "robustness", "safety-security"], "updated_at": "2026-09-27", "url": "https://arxiv.org/abs/2609.27273"}
{"aliases": [], "authors": ["Chengguang Gan", "Yunhao Liang", "QingHao Zhang", "Shiwen Ni"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2b2bda1c50fd47fabbe77c40a2938979a6a16b407b40f11737fc6f1fc795d090", "submitted_at": "2026-09-23T04:49:07Z", "updated_at": "2026-09-23T04:49:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.27353v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.27353", "identifiers": {"arxiv": "2609.27353"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.27353"}], "name": "Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebMRE pairs guidance sentences with grounded actions in an offline benchmark derived from successful WebArena traces. Controlled decoding-prefix interventions test how guide content affects action selection, while separating offline prediction metrics from live browser task success.", "topics": ["evaluation", "action-prediction", "imitation-learning", "grounding"], "updated_at": "2026-09-23", "url": "https://arxiv.org/abs/2609.27353"}
{"aliases": [], "authors": ["Zeyu Zhang", "Dingyi Rong", "Zijian Chen", "Zicheng Zhang", "Xiongkuo Min", "Guangtao Zhai"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3a89489bb1ff3001a1d806c3eefd8a6aa19503d1dfbe0265b916766119f96ec2", "submitted_at": "2026-09-23T07:27:23Z", "updated_at": "2026-09-23T07:27:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.27461v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.27461", "identifiers": {"arxiv": "2609.27461"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.27461"}], "name": "Invisible in Space, Visible in Time: Motion Vision CAPTCHA against GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "MVCAP-Bench tests live browser CAPTCHAs whose foreground meaning emerges through motion against dynamic backgrounds. A foreground-only control separates temporal camouflage difficulty from answer format and browser interaction.", "topics": ["evaluation", "screen-understanding", "video", "safety-security"], "updated_at": "2026-09-23", "url": "https://arxiv.org/abs/2609.27461"}
{"aliases": [], "authors": ["Zhitong Guo", "Jing Yu Koh", "Ruiyu Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "80f6fb973e9f8d62e515d02105d08be774fc32ed5b20a7ecee92b45d112415aa", "submitted_at": "2026-08-16T19:54:53Z", "updated_at": "2026-08-16T19:54:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.27770v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.27770", "identifiers": {"arxiv": "2609.27770"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.27770"}], "name": "Improving LLM-based Autonomous Web Agents with Filtering", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains HTML-element relevance filters on Mind2Web trajectories and transfers them to WebArena to reduce irrelevant browser context. It also evaluates a zero-shot ColBERT retriever and documents baseline web-agent failure modes.", "topics": ["web-navigation", "screen-understanding", "efficiency", "evaluation"], "updated_at": "2026-08-16", "url": "https://arxiv.org/abs/2609.27770"}
{"aliases": [], "authors": ["Alexander Gill", "Md Farhan Ishmam", "Xuyen Nguyen", "Neha Bhat", "Parker Henry DeYoung", "Fateme Hashemi Chaleshtori", "Nathan Stringham", "Kenneth Marino", "Ana Marasović"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d6ed206d8f8fec53c03956fffcafec6b55d94a47fc705bbc8b77c832b76419e6", "submitted_at": "2026-09-24T22:36:10Z", "updated_at": "2026-09-24T22:36:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.30604v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.30604", "identifiers": {"arxiv": "2609.30604"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.30604"}], "name": "The Hard Part Comes After Search: Benchmarking Web Agents on Synthesizing, Organizing, and Displaying Knowledge", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "KNOWS benchmarks browser workflows that retrieve and synthesize information into documents, presentations and spreadsheets. Task-specific evaluators combine deterministic checks with model judgments, distinguishing partial progress from usable end-to-end artifact completion.", "topics": ["evaluation", "web-navigation", "workflow", "long-horizon"], "updated_at": "2026-09-24", "url": "https://arxiv.org/abs/2609.30604"}
{"aliases": [], "authors": ["Kiana Hooshanfar", "Amirhossein Kazerouni", "Alireza Hosseini", "Michael Brudno", "Babak Taati"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9ce76f34abe391289643321027a33ac3cac8b5bd64eaa901f9a5fbf7c5561362", "submitted_at": "2026-09-25T14:59:14Z", "updated_at": "2026-09-25T14:59:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.31364v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2609.31364v1"], "reason": "Document declaration only. The paper does not establish code, model-weight or underlying dataset permissions through this license."}}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T15:06:37Z", "role": "primary", "sha256": "a3d753815e82190ee2de6f322cd816c32f2dd391f862845305fc7e57300303a6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22desktop+agent%22+OR+abs%3A%22computer+interaction%22+OR+abs%3A%22GUI+automation%22+OR+abs%3A%22web+navigation%22+OR+abs%3A%22browser+automation%22+OR+abs%3A%22visual+grounding%22+OR+abs%3A%22screen+understanding%22+OR+abs%3A%22mobile+agents%22%29+AND+submittedDate%3A%5B202609200000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T19:01:29.156954Z", "role": "primary", "sha256": "cd4ba28778b725c63f4a7329f683b2a287db2e46ed228a0d47006fcd95bbce80", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2609.31364v1"}], "id": "arxiv:2609.31364", "identifiers": {"arxiv": "2609.31364"}, "kind": "paper", "last_verified_at": "2026-10-20T19:01:29.156954Z", "licenses": {"code": null, "data": null, "document": {"basis": "The exact-v1 arXiv HTML infobox (#license-tr) declares License: CC BY 4.0 for this paper document.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T19:01:29.156954Z", "url": "https://arxiv.org/html/2609.31364v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.31364"}], "name": "OpenVAM: Open-World Visual Attention Modeling with VLMs", "notes": "Static human-attention modeling on U-EYE webpage screenshots, with FIWI held out from training; no live-browser, agent-action or task-success evaluation. Gemini-generated reference rationales are expert-checked, and JScore compares generated text with references; human validation covers a separate 100-image subset. Reference text and heatmap alignment do not establish causal explanations of gaze or temporal scanpaths. Several baseline scores come from prior papers. Code/data/weight releases and licenses were not verified; no independent reproduction.", "published_at": "2026-09-25", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Combines dense saliency prediction with explanations of attended elements, including webpage layouts; evaluates human-attention agreement and textual rationales rather than agent actions.", "topics": ["screen-understanding", "ui-design", "evaluation", "data-generation"], "updated_at": "2026-09-25", "url": "https://arxiv.org/abs/2609.31364"}
{"aliases": [], "authors": ["Chenyue Li", "Xiaoxiao Sun", "Yubo Deng", "Qinlin Zhao", "Serena Yeung-Levy", "Yuhui Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "12976e58eb48d9c4709fb03508ac7ed7eeafffd3df675d3dccfd09bdc354f3db", "submitted_at": "2026-09-25T22:07:29Z", "updated_at": "2026-09-25T22:07:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32036v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.32036", "identifiers": {"arxiv": "2609.32036"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32036"}], "name": "ScreenHaystack: Finding Blind Zones in GUI Grounding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Moves controlled target icons across high-resolution interfaces to diagnose spatial blind zones in grounding models and study targeted augmentation.", "topics": ["grounding", "evaluation", "robustness"], "updated_at": "2026-09-25", "url": "https://arxiv.org/abs/2609.32036"}
{"aliases": [], "authors": ["Chuhan Zhang", "Qi Xie", "Ziyue Wang", "Jianing Yin", "Yunfan Zhou", "Dazhen Deng", "Yingcai Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f091cab38c932e08d21afa936dfe241c93c0b2139aba1d1ce7a6d1513ad5da1a", "submitted_at": "2026-09-26T08:58:50Z", "updated_at": "2026-09-26T08:58:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32385v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.32385", "identifiers": {"arxiv": "2609.32385"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32385"}], "name": "DashAct: A Progressive Diagnostic Benchmark for GUI Agents in Interactive Dashboard Analysis", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Diagnoses dashboard-agent failures by progressively restoring interaction context, next-action support, and visual target information within the same task.", "topics": ["evaluation", "grounding", "planning"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32385"}
{"aliases": [], "authors": ["Jinming Hu", "Haodong Zhao", "Qi Jia", "Die Chen", "Tianhang Zhao", "Sufeng Duan", "Gongshen Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ff1c4eff95094ea8ecf37b7bc621eebbd5e4a677b0e11240e802a72b9b229c7f", "submitted_at": "2026-09-26T11:59:39Z", "updated_at": "2026-09-26T11:59:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32511v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T15:06:37Z", "role": "primary", "sha256": "a3d753815e82190ee2de6f322cd816c32f2dd391f862845305fc7e57300303a6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22desktop+agent%22+OR+abs%3A%22computer+interaction%22+OR+abs%3A%22GUI+automation%22+OR+abs%3A%22web+navigation%22+OR+abs%3A%22browser+automation%22+OR+abs%3A%22visual+grounding%22+OR+abs%3A%22screen+understanding%22+OR+abs%3A%22mobile+agents%22%29+AND+submittedDate%3A%5B202609200000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:08:56Z", "role": "primary", "sha256": "ce7210dccecb2896202639e6f7be41876fee20de1135b0d23278cead2d695339", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2609.32511v1"}], "id": "arxiv:2609.32511", "identifiers": {"arxiv": "2609.32511"}, "kind": "paper", "last_verified_at": "2026-10-20T15:08:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32511"}], "name": "Learning from Others, Acting for You: Cross-User Memory Sharing for LLM Agents", "notes": "Main Mind2Web evaluation uses 252 recorded tasks assigned to 20 simulated users, frozen demonstration-derived memory and task-macro step metrics; it is not a live-browser completion test. The main split withholds same-site local experience, and sharing gains shrink when local coverage rises. Mind2Web has no preference channel; preference claims come from separate service and coding settings. User-bound retrieval is not a demonstrated privacy or security guarantee. Selected sections were source-read; no independent reproduction.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "ShareMem consolidates reusable experience across users while keeping preference retrieval user-bound; its web evaluation predicts target elements and actions on recorded Mind2Web states.", "topics": ["memory", "web-navigation", "action-prediction", "evaluation"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32511"}
{"aliases": [], "authors": ["Prince Zizhuang Wang", "Chenhao Liang", "Zelong Xu", "Aojie Yuan", "Xiaolin Zhou", "Haiyue Zhang", "Yue Zhao", "Xiyang Hu", "Shuli Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d53f2512b99828849f047f63412264437a59b730c117c801de252e6dc80a33fc", "submitted_at": "2026-09-26T13:18:00Z", "updated_at": "2026-09-26T13:18:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32600v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.32600", "identifiers": {"arxiv": "2609.32600"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32600"}], "name": "CUA-SWE: When Computer-Use Agents Meet Visual Software Engineering", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CUA-SWE, a benchmark and execution environment that combines code changes with graphical interaction and visual inspection of running software. Task-specific tests verify repairs and changes whose required information may be available only through the application interface.", "topics": ["evaluation", "workflow", "tool-integration"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32600"}
{"aliases": [], "authors": ["Dongsheng Liu", "Chao Jin", "Wenkui Yang", "Hejin Wang", "Junwei Yang", "Zeren Zhang", "Ziwei Chen", "Huaibo Huang", "Jie Cao", "Ran He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d641819d168eedc07270607da33f3197c15f2f90cf99f538c3839eed506f3308", "submitted_at": "2026-09-26T14:35:45Z", "updated_at": "2026-09-26T14:35:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32679v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}], "id": "arxiv:2609.32679", "identifiers": {"arxiv": "2609.32679"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:17Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32679"}], "name": "The GUI Is Not the State: Diagnosing State Aliasing in GUI World Models", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces StateAliasBench to isolate GUI world-model ambiguities in which matching visible screens and actions can lead to different futures. It augments frozen world models with structured state recovered from interaction history and evaluates the effect on prediction and agent execution.", "topics": ["world-models", "state-tracking", "evaluation"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32679"}
{"aliases": [], "authors": ["Asif Shahriar", "Md Nafiu Rahman", "Sadif Ahmed", "Farig Sadeque", "Md Rizwan Parvez"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "802342ce2c486c30c84c33c897becb0348831034d4d11f8d3d3861ec0380b65c", "submitted_at": "2026-09-26T20:09:16Z", "updated_at": "2026-09-26T20:09:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32915v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}], "id": "arxiv:2609.32915", "identifiers": {"arxiv": "2609.32915"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:17Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32915"}], "name": "AgentTell: Behavioural Side-Channel Leakage in Browser-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AgentTell, a benchmark for privacy leakage through browser actions after an agent learns a private fact on another website. It examines whether agents choose revealing options despite nondisclosure instructions and whether their final reports accurately describe that behavior.", "topics": ["safety-security", "evaluation", "memory"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32915"}
{"aliases": [], "authors": ["Ben Hagag", "William L. Anderson", "Srija Chakraborty", "Christian Schroeder de Witt"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3c380079220050fbe354509238a5a79c7df2917a245490e4ec0cff512f7565ab", "submitted_at": "2026-09-27T02:20:08Z", "updated_at": "2026-09-27T02:20:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.33102v1"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T12:58:20.217517Z", "role": "primary", "sha256": "f275c47a8b6d349a70e39a9e55eea5a09d99d5f35cc896100f3bcfd5b0923e0b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22computer-use+agents%22+OR+abs%3A%22GUI+agents%22+OR+abs%3A%22browser+agents%22+OR+abs%3A%22mobile+GUI%22%29+AND+submittedDate%3A%5B202609200000+TO+202610201255%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:28.833839+00:00", "role": "primary", "sha256": "3ea667f4187ba601011e5ff00c4d63676b6d9903d12b80a0d7d5dd9920f00c65", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2609.33102v1"}], "id": "arxiv:2609.33102", "identifiers": {"arxiv": "2609.33102"}, "kind": "paper", "last_verified_at": "2026-10-20T13:12:28.833839+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.33102"}], "name": "ORBIT: A Framework for Multi-Agent Safety and Security Evaluations", "notes": "BrowserART uses an accessibility-tree action interface, while the desktop adapter uses mouse, keyboard and screenshots. The adapted tasks are not claimed to preserve the constructs of their source benchmarks. Experiments cover a representative configuration slice with fixed attack templates; the OS-Harm results are imported from prior work using a different model and stack. Browser defense tables also differ in baseline versus defended judges and in judged benign denominators. Human judge validation used one grader. These results do not establish a universally effective defense. The linked repository and its code/data licenses were not independently inspected, and no experiment was run.", "published_at": "2026-09-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Inspect-based framework that varies multi-agent roles, communication, attacks and defenses, with BrowserGym and virtual-desktop GUI adapters alongside coding and other tool scenarios.", "topics": ["evaluation", "safety-security", "web-navigation"], "updated_at": "2026-09-27", "url": "https://arxiv.org/abs/2609.33102"}
{"aliases": [], "authors": ["Shaoqing Zhang", "Kehai Chen", "Xuefeng Bai", "Zhuosheng Zhang", "Pengfei Zhang", "Yang Xiang", "Min Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2f4116455785579ece9a2ff1bae304b62f1cbe13a91db8d146bd281a27d19eb0", "submitted_at": "2026-09-28T01:48:48Z", "updated_at": "2026-09-29T09:05:42Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.34113v2"}}, "environments": ["android", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.34113", "identifiers": {"arxiv": "2609.34113"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.34113"}], "name": "GUITAR: Structured Failure Diagnosis of GUI Agents via State Transitions", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUITAR, which groups visually different screens into functional states and analyzes agent failures over a state-transition graph. It uses the resulting bottlenecks to diagnose recurring mobile and web errors and evaluate targeted guidance.", "topics": ["evaluation", "observability", "robustness"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.34113"}
{"aliases": [], "authors": ["Tien Tran", "Namho Koh", "Daiki E. Matsunaga", "Ayush Jain", "Kee Eung Kim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0ee7ddc170afcd2fa09d1c8d12a07bbd6aac692bc12ffc27bd39f1de0c912dfa", "submitted_at": "2026-09-28T02:21:13Z", "updated_at": "2026-09-28T02:21:13Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.34139v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.34139", "identifiers": {"arxiv": "2609.34139"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.34139"}], "name": "Same Tasks, Different Apps: Why Mobile GUI Agents Fail to Generalize?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AnyAppBench, a live Android benchmark that holds user goals fixed while changing the application used to achieve them. It combines category-controlled task pairs with failure annotations to distinguish task understanding from application-specific execution skill.", "topics": ["evaluation", "robustness", "mobile-navigation"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.34139"}
{"aliases": [], "authors": ["Ziqiang Wang", "Li Gu", "Zhixiang Chi", "Linlian Jiang", "Zihuan Jiang", "Linqiang Guo", "Siobhan Reid", "Zhi Liu", "Yang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7fbaeb0b4f891d64cc12a3d5663b6804df6c7db82e6a7ccfd54aa10e837010de", "submitted_at": "2026-09-28T05:00:45Z", "updated_at": "2026-09-28T05:00:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.34321v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.34321", "identifiers": {"arxiv": "2609.34321"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.34321"}], "name": "One Rollout Is All You Get: Fully Test-Time Adaptation for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Defines a test-time adaptation setting with one counted attempt per arriving GUI task and no ground-truth feedback. Its SOLO method filters or relabels completed episodes, then updates a small adapter from a short window of the agent’s own experience.", "topics": ["memory", "evaluation", "robustness"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.34321"}
{"aliases": [], "authors": ["Bingo Zhang", "Haochuan Lu", "Zongjie Li", "Genjian Li", "Ari Yu Zhang", "Chaozheng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f81fd07262dccf6783729e4336fbd77835f2fcdea030eef8dc2272e7ef82b8fc", "submitted_at": "2026-09-28T09:49:25Z", "updated_at": "2026-09-28T09:49:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.34769v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.34769", "identifiers": {"arxiv": "2609.34769"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.34769"}], "name": "LongPuzzleBench: Evaluating GUI Agents on Long-Horizon Visual Puzzles", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced. Agents execute real GUI moves in six puzzle games; the tasks diagnose computer-control planning but remain game-based rather than general productivity workflows.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "LongPuzzleBench evaluates coherent long-horizon planning on persistent visual puzzle boards through native GUI actions. It compares GUI-only and code-assisted agents and uses controlled hints and memory diagnostics to examine decisions that preserve or eliminate future solutions.", "topics": ["evaluation", "planning", "long-horizon", "state-tracking"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.34769"}
{"aliases": [], "authors": ["Anton Emelyanov", "Maria Tikhonova", "Zaven Martirosian", "Sergei Averkiev", "Alena Fenogenova"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d5090c6b7879e38102b1180ee1e4dfd1552af777172e2f94bc0f9a684fb59dc7", "submitted_at": "2026-09-28T12:24:23Z", "updated_at": "2026-09-28T12:24:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.35026v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.35026", "identifiers": {"arxiv": "2609.35026"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.35026"}], "name": "WebPageBench: Event-Level Verification and Controlled UI-Variant Generation for Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Verifies web tasks through typed interface events and generates controlled UI variations while preserving instructions and success conditions.", "topics": ["evaluation", "robustness"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.35026"}
{"aliases": [], "authors": ["Xunjian Yin", "Tianchen Guan", "Jinao Wang", "Weili Cao", "Daisy Xinlei Lin", "Royce Cheng-Yue", "Keagan Long", "Kyle Wong", "Bhuwan Dhingra", "Xiangjun Wang", "Shuyan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "197cfb8424584955bc85602d6559037ad31ee815f89b45dc8f9f18c8636552ad", "submitted_at": "2026-09-20T16:16:23Z", "updated_at": "2026-09-20T16:16:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.35814v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}], "id": "arxiv:2609.35814", "identifiers": {"arxiv": "2609.35814"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.35814"}], "name": "Constructing Challenging Browser-Use Tasks by Controlled Environment Interventions", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "BreakingWeb pairs solvable browser tasks with controlled environment interventions that preserve the goal, target and backend success criterion. The paired evaluations isolate robustness failures, including agents declaring completion without the required state change.", "topics": ["evaluation", "robustness", "data-generation", "state-tracking"], "updated_at": "2026-09-20", "url": "https://arxiv.org/abs/2609.35814"}
{"aliases": [], "authors": ["Jianxing Chen", "Xiao Yu", "Shipra Agrawal", "Zhou Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e92c63d4cf58375aa7068570f8a98c145e24f8d393c5807c0aff3ed67f18b9f0", "submitted_at": "2026-09-28T20:01:35Z", "updated_at": "2026-09-28T20:01:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.36201v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.36201", "identifiers": {"arxiv": "2609.36201"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.36201"}], "name": "SCOUT: Synergizing Reasoning and Tool-Use for Computer-Use Safety", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SCOUT, a safety verifier that first derives task-specific completion and safety rubrics from a computer-use trajectory, then actively probes the resulting environment for evidence. It evaluates whether these two stages detect harmful consequences that screenshots alone can miss.", "topics": ["safety-security", "evaluation", "tool-integration"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.36201"}
{"aliases": [], "authors": ["Dongchan Shin", "Xing Han Lù", "Jiaqi Deng", "Jay Gala", "Tomás Vergara Browne", "Jaewon Moon", "Fengyuan Liu", "Alexandre Drouin", "Siva Reddy", "Alexandre Lacoste"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2fe39365133228daa770445f19002f6abe34ca1151cf4952a225ba0001de5250", "submitted_at": "2026-09-29T01:46:48Z", "updated_at": "2026-09-29T01:46:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.36488v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.36488", "identifiers": {"arxiv": "2609.36488"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.36488"}], "name": "AdaptArena: Evaluating Test-Time Personalization of Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AdaptArena to test whether web agents can infer unstated user preferences from relevant past trajectories and carry those preferences into new tasks. Its AdaptiveAgent baseline separates preference-retrieval challenges from failures in downstream web execution.", "topics": ["evaluation", "memory", "web-navigation"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.36488"}
{"aliases": [], "authors": ["Shengtian Yang", "Ziyu Xiong", "Kaibing Yang", "Guangfeng Cai", "Yewen Li", "Peng Jiang", "Gai Kun", "Qingpeng Cai", "Lei Feng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d45cee8117e302aa190d38c8298bf479936a5a27aa9b85716a76a2a90499069c", "submitted_at": "2026-09-29T07:51:57Z", "updated_at": "2026-09-29T07:51:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.36939v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.36939", "identifiers": {"arxiv": "2609.36939"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.36939"}], "name": "SCA: Spatial Credit Assignment for Reinforcement Learning of GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Spatial Credit Assignment, which uses click coordinates to refine group-relative reinforcement-learning updates for GUI grounding. It distinguishes failed clicks by spatial evidence while leaving the deployed action policy unchanged.", "topics": ["grounding", "reinforcement-learning", "evaluation"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.36939"}
{"aliases": [], "authors": ["Mei Wu", "Rui Xie", "Runyu Zhang", "Yuqiang Li", "Tianfan Fu", "Bo Chen", "Kai Yu", "Xin Chen", "Lu Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e032f67257a52dbc59310d921d802e4dc85af3af110d22bb6d005d202bbaa6b2", "submitted_at": "2026-09-29T08:58:51Z", "updated_at": "2026-09-29T08:58:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.37053v1"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}], "id": "arxiv:2609.37053", "identifiers": {"arxiv": "2609.37053"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.37053"}], "name": "MatToolBench: Benchmarking Multimodal Agents in Real-World Materials Science Workflows", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MatToolBench, a Windows-based materials-science benchmark combining GUI operation, scientific-software scripting, and database queries. Expert-defined subcriteria support partial-credit scoring, while the evaluation examines domain knowledge gaps and cross-tool handoff failures.", "topics": ["science", "evaluation", "cross-app"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.37053"}
{"aliases": [], "authors": ["Hongcheng Gao", "Hailong Qu", "Yu Lei", "Henghui Sun", "Haoyang Li", "Yipeng Wei", "Naihao Xue", "Xiaohan Yu", "Zhuo Tao", "Yihe Zang", "Yajiao Wang", "Jingyi Tang", "Yi Li", "Jingjing Zhou", "Jie Luo", "Bohan Zeng", "Chengyu Shen", "Hao Jiang", "Chong Chen", "Bowen Qu", "Olive Huang", "Zeqiang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aab7f586cb58a1f2d20267fd810ba34d6d4d166e3da715d2b3ff0b77e464473e", "submitted_at": "2026-09-29T14:32:35Z", "updated_at": "2026-10-09T08:15:21Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.37686v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}], "id": "arxiv:2609.37686", "identifiers": {"arxiv": "2609.37686"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.37686"}], "name": "EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces professional engineering tasks across software domains, with GUI and CLI access and artifact checks for geometric, physical, and design validity.", "topics": ["evaluation", "science", "cross-app"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2609.37686"}
{"aliases": [], "authors": ["Yongjiang Wu", "Junyuan Zhang", "Ada Chen", "Kuiyi Gao", "Wenxuan Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f14844abefab6b4eb464a1c7675f9de7f460849872ebb776230724ed5828feed", "submitted_at": "2026-08-07T07:51:13Z", "updated_at": "2026-08-07T07:51:13Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.38184v1"}}, "environments": ["web", "desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.38184", "identifiers": {"arxiv": "2609.38184"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.38184"}], "name": "RealGUINoise: An Interactive Cross-Platform Benchmark for GUI Agent Robustness under Real-World Interface Noise", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates interactive GUI agents under varied realistic interface noise, comparing task completion, safety, and behavior with clean-environment runs.", "topics": ["robustness", "evaluation", "safety-security"], "updated_at": "2026-08-07", "url": "https://arxiv.org/abs/2609.38184"}
{"aliases": [], "authors": ["Terumi Chiba", "Guangzhi Sun", "Zheqi Yuan", "Chao Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bded059a37101b0e6a91ba662e0d70e5e010ddb8c3557598dff2ba82afa45e34", "submitted_at": "2026-09-30T03:19:25Z", "updated_at": "2026-09-30T03:19:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.38867v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}], "id": "arxiv:2609.38867", "identifiers": {"arxiv": "2609.38867"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.38867"}], "name": "Talk2Agent: Benchmarking Voice Interfaces for Text Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Talk2Agent, a benchmark of spoken instructions for computer-use tasks that measures whether voice interfaces preserve task-critical information. It pairs human speech with an execution-free intention-based metric and checks how that metric relates to downstream task completion.", "topics": ["evaluation", "robustness"], "updated_at": "2026-09-30", "url": "https://arxiv.org/abs/2609.38867"}
{"aliases": [], "authors": ["Dingyuan Dai", "Heli Qi", "Lei Liu", "Yinxi Li", "Baiding Chen", "Zijun Dou", "Qingcheng Zeng", "Qi Kang", "Oliver Sun", "Eric Wang", "Bo Zhou", "Haixin Wang", "Yufan Du", "Shi Bo", "Ruihan Lin", "Mengqi Yuan", "Dunjie Lu", "Steven Dillmann", "Yiming Shi", "Tina Su", "Amy Xin", "Minghao Liu", "Xi Wang", "Xu Huang", "Ge Zhang", "Pengyu Nie", "Zhen Yang", "Jie Tang", "Juanzi Li", "Weihao Xuan", "Tianyu Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "690cd75934bae65311240cd4e6db7b8fd607056bd60799bf70cddd03efda9261", "submitted_at": "2026-09-30T15:03:05Z", "updated_at": "2026-09-30T15:03:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.39903v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.39903", "identifiers": {"arxiv": "2609.39903"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.39903"}], "name": "OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Benchmarks computer agents on scientific-software workflows with task-specific checks of application states and generated research artifacts.", "topics": ["evaluation", "science"], "updated_at": "2026-09-30", "url": "https://arxiv.org/abs/2609.39903"}
{"aliases": [], "authors": ["Yong Du", "Tongbo Chen", "Zhengxi Lu", "Yizhou Liu", "Bofan Chen", "Tao Jiang", "Wenhao Xu", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8f3745b6b497c8819dc990bdb5d74e7af463d23654cbc7cdd76ade0faf2b3e0c", "submitted_at": "2026-09-30T17:40:09Z", "updated_at": "2026-10-01T17:50:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.40253v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.40253", "identifiers": {"arxiv": "2609.40253"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.40253"}], "name": "ComputerSD: Online Self-Distillation from Real-Time Feedback for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces ComputerSD, which turns feedback from executed GUI transitions into online self-distillation guidance and step-level scores. A GUI analyzer supplies these signals, which are combined with trajectory-level reinforcement learning in asynchronous training.", "topics": ["reinforcement-learning", "evaluation"], "updated_at": "2026-10-01", "url": "https://arxiv.org/abs/2609.40253"}
{"aliases": [], "authors": ["Pranjal Aggarwal", "Lawrence Keunho Jang", "Sean Welleck", "Daniel Fried", "Ruslan Salakhutdinov", "Jing Yu Koh"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "502912cdc35afd61beeb0202d55fc8feefcd92cd3719cd0e85bc71561d6f5a3c", "submitted_at": "2026-09-30T17:48:06Z", "updated_at": "2026-09-30T17:48:06Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.40284v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.40284", "identifiers": {"arxiv": "2609.40284"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.40284"}], "name": "cua-speedrun: Standardized Benchmarking of the Speed of Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Standardizes virtual machines, interaction pipelines, and agent interfaces to compare computer-use speed, cost, and task performance.", "topics": ["evaluation", "efficiency"], "updated_at": "2026-09-30", "url": "https://arxiv.org/abs/2609.40284"}
{"aliases": [], "authors": ["Chengguang Gan", "Zimeng He", "Yoshihiro Tsujii", "Ken-ichiro Kobayashi", "Hiroki Itoh", "Kotaro Funakoshi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9a5b97264b3928c8dc74b955cc4baa7005cd34b0ce9e42cbc48a63bc6dd1f5a7", "submitted_at": "2026-10-01T11:31:37Z", "updated_at": "2026-10-01T11:31:37Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.01491v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.01491", "identifiers": {"arxiv": "2610.01491"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.01491"}], "name": "Auditing Web Agent Evaluation on WebArena-Lite: Human Review of Outcomes and Trajectories", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Audits WebArena-Lite outcomes and trajectories with human review to identify automatic-evaluator false negatives and the first consequential execution errors. It also studies explicit execution-state memory and procedural guidance, showing why final task scores can miss substantial differences in agent behavior.", "topics": ["evaluation", "observability", "memory"], "updated_at": "2026-10-01", "url": "https://arxiv.org/abs/2610.01491"}
{"aliases": [], "authors": ["Beining Wu", "Zihao Ding", "Jun Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7e85a601c31c6d06831ed1c590deef29d1c365e5d4751a308e42e8496ef11b19", "submitted_at": "2026-10-01T14:34:32Z", "updated_at": "2026-10-01T14:34:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.01787v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.01787", "identifiers": {"arxiv": "2610.01787"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.01787"}], "name": "Not All Experience Belongs in the Weights: Component Routing for Self-Improving GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies how to route different parts of GUI experience into model weights or runtime context rather than treating whole trajectories as a single unit. It separates locators, procedures, state facts, and lessons, then predicts their destination from recurrence and dependence on the current state.", "topics": ["memory", "evaluation", "efficiency"], "updated_at": "2026-10-01", "url": "https://arxiv.org/abs/2610.01787"}
{"aliases": [], "authors": ["Dhananjay Ashok", "Shantanu Agarwal", "Vivek Datla", "Jonathan May", "Alfy Samuel"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "35ded32e7bac9a875a0fa41efad43fed5fb0aab54fade5cfc2a0628b1e666587", "submitted_at": "2026-10-01T22:30:08Z", "updated_at": "2026-10-01T22:30:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.02542v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.02542v1"], "reason": "The retained version-specific arXiv HTML infobox explicitly labels this paper License: CC BY 4.0. This document declaration does not establish code, training-data or model-weight licenses."}}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T15:06:37Z", "role": "primary", "sha256": "a3d753815e82190ee2de6f322cd816c32f2dd391f862845305fc7e57300303a6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22desktop+agent%22+OR+abs%3A%22computer+interaction%22+OR+abs%3A%22GUI+automation%22+OR+abs%3A%22web+navigation%22+OR+abs%3A%22browser+automation%22+OR+abs%3A%22visual+grounding%22+OR+abs%3A%22screen+understanding%22+OR+abs%3A%22mobile+agents%22%29+AND+submittedDate%3A%5B202609200000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:08:56Z", "role": "primary", "sha256": "d8c22c6b3f07d4156c38af9dd7ca7c06a5d0a11889a7350d821a020e21604590", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.02542v1"}], "id": "arxiv:2610.02542", "identifiers": {"arxiv": "2610.02542"}, "kind": "paper", "last_verified_at": "2026-10-20T15:08:56Z", "licenses": {"code": null, "data": null, "document": {"basis": "The retained version-specific arXiv HTML infobox explicitly labels this paper License: CC BY 4.0. This document declaration does not establish code, training-data or model-weight licenses.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T15:08:56Z", "url": "https://arxiv.org/html/2610.02542v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.02542"}], "name": "How To Train Your World Model: Fine-tuning vs RAG for LM-based World Modeling", "notes": "Adjacent cross-domain study with WebShop as its web evaluation. WebShop uses text-rendered search/product pages and an explicit valid-action menu, not screenshot control or live purchases. A frozen policy proposes five actions; a world model predicts their next observations and the policy scores them before one is executed in the simulator. Utility is normalized environment reward, while fidelity uses a model judge; neither is a universal GUI task-success rate. Fine-tuning and retrieval use collected training experience; the hybrid masks same-episode retrieval, which does not establish general contamination freedom. The retained Atom and v1 HTML abstract wording differs; original bibliographic metadata and its abstract digest are preserved.", "published_at": "2026-10-01", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Compares fine-tuned, retrieval-based and hybrid language-model world models for next-observation prediction and action selection in text environments including WebShop.", "topics": ["world-models", "planning", "evaluation", "web-navigation"], "updated_at": "2026-10-01", "url": "https://arxiv.org/abs/2610.02542"}
{"aliases": [], "authors": ["Yun-Yun Tsai", "Yuning Mao", "Shiqi Wang", "Junfeng Yang", "Sinong Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7ed117876ece99f55e241e6365d0e785b907f9a7a24cceb01b1fd757191a35c9", "submitted_at": "2026-10-02T00:17:11Z", "updated_at": "2026-10-02T00:17:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.02617v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:00:31.520084+00:00", "role": "primary", "sha256": "f053c4c65f6443acbda26e25cf36f7b6e8ea45f2916535604229f7c4e62ed679", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:00:31.520084+00:00", "role": "primary", "sha256": "f053c4c65f6443acbda26e25cf36f7b6e8ea45f2916535604229f7c4e62ed679", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.02617", "identifiers": {"arxiv": "2610.02617"}, "kind": "paper", "last_verified_at": "2026-10-18T10:00:31.520084+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.02617"}], "name": "WebUIProof: Benchmarking WebUI Code Generators with UI-Agent Execution Harness", "notes": "Adjacent evaluation infrastructure: the target is generated WebUI functionality, not a general score for computer-use agents. This annotation is based on the complete retained abstract; released artifacts and experimental results were not independently inspected or reproduced.", "published_at": "2026-10-02", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces an execution-based benchmark for generated web interfaces, with a headless-browser UI-agent harness that checks DOM interactions and functional assertions.", "topics": ["evaluation", "web-navigation", "ui-design"], "updated_at": "2026-10-02", "url": "https://arxiv.org/abs/2610.02617"}
{"aliases": [], "authors": ["Giulio Zingrillo", "Hanna Foerster", "Ilia Shumailov", "Yiren Zhao", "Robert Mullins"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7469484bb1186dd633b26e41dddc890008e893c303f93d2352f93c9d47fd78c2", "submitted_at": "2026-10-02T10:08:39Z", "updated_at": "2026-10-02T10:08:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.03089v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.03089", "identifiers": {"arxiv": "2610.03089"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.03089"}], "name": "Securing Computer-Use Agents Against Branch Steering Attacks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces STEER-Bench to study branch-steering attacks that manipulate untrusted web data toward hazardous paths already permitted by an agent’s plan. Its COBRA architecture combines trusted branching plans with advance constraints on the parameters and destinations each branch can use.", "topics": ["safety-security", "planning", "evaluation"], "updated_at": "2026-10-02", "url": "https://arxiv.org/abs/2610.03089"}
{"aliases": [], "authors": ["Maria Sanchez del Rio", "Mikhail Terekhov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f9dd28bdfb8879cbdb68f539c286365a06f4e9d04eba2420bb46f3575fdbfa05", "submitted_at": "2026-10-02T00:31:34Z", "updated_at": "2026-10-12T23:17:28Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2610.03818v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.03818", "identifiers": {"arxiv": "2610.03818"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.03818"}], "name": "Control OSWorld: An AI Control Environment for GUI Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Pairs OSWorld tasks with harmful side objectives to evaluate monitors that detect malicious behavior in computer-use agent trajectories.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-10-12", "url": "https://arxiv.org/abs/2610.03818"}
{"aliases": [], "authors": ["Ziyang Long", "Xinqi Li", "Lujing Xing", "Hsin-Jung Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "baacd8b4fad7b146bb9152a54d60a617fdbed5d0142c0c0768dacea2e7f18e60", "submitted_at": "2026-10-03T22:56:09Z", "updated_at": "2026-10-03T22:56:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.04800v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.04800", "identifiers": {"arxiv": "2610.04800"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.04800"}], "name": "MedImageOSWorld: Benchmarking GUI Agents for Medical Image Consoles", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates GUI agents in simulated medical-imaging consoles, separately checking workflow completion and acquisition outcomes against hidden anatomical targets.", "topics": ["evaluation", "science", "planning"], "updated_at": "2026-10-03", "url": "https://arxiv.org/abs/2610.04800"}
{"aliases": [], "authors": ["Yifan Zhang", "Yutong Dai", "Viraj Prabhu", "Zhiyuan Hu", "Ran Xu", "Zeyuan Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2f3e9860574951cd6b749d47da748e4aec36d4f5afe226f4a3d0d5459fd07935", "submitted_at": "2026-10-05T17:57:51Z", "updated_at": "2026-10-05T17:57:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.06829v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.06829", "identifiers": {"arxiv": "2610.06829"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.06829"}], "name": "CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CLIFT, which certifies agent-generated verification signals against training-time judges and reuses them for per-step rewards and trajectory selection. At inference, a frozen question bank helps compare candidate web rollouts without querying an external judge.", "topics": ["reinforcement-learning", "evaluation", "web-navigation"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2610.06829"}
{"aliases": [], "authors": ["Li Hu", "Kanghua Mo", "Yingbin Jin", "Qingqing Ye", "Haibo Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5afea9414462386ee921b9ead9381c0204963c710c9f616a96f4bc47079f059e", "submitted_at": "2026-09-29T03:03:00Z", "updated_at": "2026-09-29T03:03:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.06898v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:00:31.520084+00:00", "role": "primary", "sha256": "f053c4c65f6443acbda26e25cf36f7b6e8ea45f2916535604229f7c4e62ed679", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:00:31.520084+00:00", "role": "primary", "sha256": "f053c4c65f6443acbda26e25cf36f7b6e8ea45f2916535604229f7c4e62ed679", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:10:35.098754+00:00", "role": "primary", "sha256": "73f83c27e58e46086dbd5f32f108df2c5faa9ab2bd968d48062f39d94b2cc14b", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/pdf/2610.06898v1"}], "id": "arxiv:2610.06898", "identifiers": {"arxiv": "2610.06898"}, "kind": "paper", "last_verified_at": "2026-10-18T10:00:31.520084+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.06898"}], "name": "DIBench: Benchmarking Decision Integrity of GUI-based Mobile Agents Under Deceptive Injections", "notes": "The commercial-app protocol uses fixed screenshots and context for selection decisions, rather than live end-to-end transactions. Android is supported by the paper's emulator setup. Completion and integrity are distinct measures; no reported scores were reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (3.1 Benchmark components; 3.2 Instance generation; 3.3 Paired protocol/metrics; 5 Limitations/Ethics; Appendix B (Android emulator setup)); not an entire-paper reading or reproduction claim.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates whether deceptive UI content steers mobile agents away from instruction-consistent choices, using paired benign and injected decision states and decision-integrity metrics.", "topics": ["evaluation", "safety-security", "mobile-navigation"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2610.06898"}
{"aliases": [], "authors": ["Aadi Chauhan", "Arthur Ilyasov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "81342daf25b0fb26f6481fcf836c3d958e26c2358be96eadb51e6914ea87e43c", "submitted_at": "2026-10-05T21:50:50Z", "updated_at": "2026-10-05T21:50:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.07444v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.07444", "identifiers": {"arxiv": "2610.07444"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.07444"}], "name": "Decoupling What from Where: How Should a Small GUI Grounding Model Receive the Action Type?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares ways to condition a small GUI grounding model on action type under matched training conditions on Android in the Wild. The study traces apparent gains to a coordinate-serialization choice and reports limited evidence that the gains persist with predicted action types.", "topics": ["grounding", "evaluation", "robustness"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2610.07444"}
{"aliases": [], "authors": ["Yeji Park", "Jaeyun Shim", "Taesik Gong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b792cc0355a0947c61014c471c482aef173f761f7396a939c14816ac644abe66", "submitted_at": "2026-10-06T08:38:30Z", "updated_at": "2026-10-06T08:38:30Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.07972v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.07972", "identifiers": {"arxiv": "2610.07972"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.07972"}], "name": "Can Agents Work for Everyone? Cross-User Reliability for Mobile GUI Agents in Personalized User Interfaces", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces PAIR for controlled evaluation of mobile GUI agents across user-conditioned application states and RePAIR for training from cross-user subgoal outcomes. The study examines failures when personalized interfaces change which items an agent must identify.", "topics": ["evaluation", "robustness", "mobile-navigation", "reinforcement-learning"], "updated_at": "2026-10-06", "url": "https://arxiv.org/abs/2610.07972"}
{"aliases": [], "authors": ["Wanjing Han", "Levi Taiji Li", "Mu Zhang", "Yue Jiang", "Guanhong Tao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3b6ef7e785ad509ebd13fdb3da97fc583ee4c28dd42b4eab842c6cd88d7ef51f", "submitted_at": "2026-10-07T00:05:53Z", "updated_at": "2026-10-07T00:05:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.09240v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.09240", "identifiers": {"arxiv": "2610.09240"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.09240"}], "name": "Adversarial Images Hijack Web Agents from Visual Grounding to Browser Execution", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebMirage, a visual red-teaming framework that models webpage element competition and action post-processing to test whether manipulated grounding decisions lead to executed browser actions.", "topics": ["safety-security", "grounding", "evaluation"], "updated_at": "2026-10-07", "url": "https://arxiv.org/abs/2610.09240"}
{"aliases": [], "authors": ["Michael Ofengenden", "Maksym Andriushchenko"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "03aea09f3ed1b75a8a6584efdfa51d45d1cfa9dc53fb3a25efd820c5b72e094a", "submitted_at": "2026-10-07T12:26:24Z", "updated_at": "2026-10-08T09:06:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2610.09944v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T13:48:10Z", "role": "primary", "sha256": "6d8c0463b6d50320246021f0394d0e39c9fd839f1ef1f742c8cdfa231872b059", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22+OR+ti%3AGUI+OR+ti%3A%22Android+agent%22+OR+abs%3A%22computer-use+agents%22+OR+abs%3A%22GUI+agents%22+OR+abs%3A%22browser+agents%22+OR+abs%3A%22mobile+GUI%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:08:56.488063+00:00", "role": "primary", "sha256": "5424a5f9278ef276c84799a08d4e2e171ee93947a38c1d6670c36fa76dc84fe3", "supports": ["name", "kind", "summary", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.09944v2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:15:05.368243+00:00", "role": "primary", "sha256": "9bef76f95b41a7ca39c6fa3276d6122e4b27e1e0bd56a777edcfd7e11e7d7342", "supports": ["environments", "notes"], "url": "https://github.com/michaelofengenden/agenttimebench"}], "id": "arxiv:2610.09944", "identifiers": {"arxiv": "2610.09944"}, "kind": "paper", "last_verified_at": "2026-10-20T15:08:56.488063+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.09944"}], "name": "AgentTime: Can Agents Estimate and Control Their Own Runtime?", "notes": "This cross-domain suite includes 16 OSWorld 2.0 tasks among 222 tasks from 18 sources. A single run per task and requested duration, provider effects and incomplete native grades limit comparisons; elapsed time is not evidence of productive work. The OSWorld task 092 video judge was skipped without an API key. The inspected repository README identifies desktop-VM timing and states that the current 222-task forecasting runs will be added; its existing forecasting code covers an earlier study, and transcript work/wait labels are not public. No code or benchmark was executed and no license was verified.", "published_at": "2026-10-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates duration following, runtime forecasts and elapsed-time estimates in native agent harnesses, including OSWorld desktop tasks, with external timers, native graders and work-versus-wait analysis.", "topics": ["evaluation", "observability"], "updated_at": "2026-10-08", "url": "https://arxiv.org/abs/2610.09944"}
{"aliases": [], "authors": ["Xing Han Lù", "Dheeraj Vattikonda", "Sina Hajimiri", "Fatemeh Pesaran Zadeh", "Parishad BehnamGhader", "Ghazwa Darwiche", "Amirhossein Kazemnejad", "Christopher Pal", "Alexandre Drouin", "Siva Reddy"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "37413f11bf8263f1516a8edca8c594227fcab104c0c767a7daa8eead9645ec34", "submitted_at": "2026-10-08T01:03:48Z", "updated_at": "2026-10-08T01:03:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.11050v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.11050", "identifiers": {"arxiv": "2610.11050"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.11050"}], "name": "AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AgentHorizon, a benchmark for judging long computer-use trajectories using closely related instructions paired with recorded actions and screenshots. It tests whether judges can detect constraint violations and distinguish successful tasks from similar but incompatible requests.", "topics": ["evaluation", "long-horizon", "cross-app"], "updated_at": "2026-10-08", "url": "https://arxiv.org/abs/2610.11050"}
{"aliases": [], "authors": ["Jiaming Zhang", "Xuan Wang", "Fuyao Zhang", "Yang Cao", "Lingjuan Lyu", "Wei Yang Bryan Lim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "53120f960c7b98e87210c93ac661a2e397b858cc9f211786b0f8233fa61c6dd3", "submitted_at": "2026-10-08T13:33:23Z", "updated_at": "2026-10-08T13:33:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.11942v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.11942", "identifiers": {"arxiv": "2610.11942"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.11942"}], "name": "Right Screen, Wrong Transition: World Models as Verifiers for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents LGWM, a decoder-free GUI world model that predicts the next screen representation from an action and compares it with the observed state to verify transitions. Its paired diagnostic separates legitimate and hijacked transitions that lead to the same credential screen.", "topics": ["world-models", "safety-security", "evaluation"], "updated_at": "2026-10-08", "url": "https://arxiv.org/abs/2610.11942"}
{"aliases": [], "authors": ["Utku Boran Torun", "Veli Karakaya", "Eray Tüzün"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0e77dcf101ba95684aac51084e43b914c0aca86fc25d061612224118004bdb21", "submitted_at": "2026-10-08T13:44:53Z", "updated_at": "2026-10-08T13:44:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.11963v1"}}, "environments": ["web", "desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T12:58:20.217517Z", "role": "primary", "sha256": "f275c47a8b6d349a70e39a9e55eea5a09d99d5f35cc896100f3bcfd5b0923e0b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22computer-use+agents%22+OR+abs%3A%22GUI+agents%22+OR+abs%3A%22browser+agents%22+OR+abs%3A%22mobile+GUI%22%29+AND+submittedDate%3A%5B202609200000+TO+202610201255%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:25.581656+00:00", "role": "primary", "sha256": "9fad44164a6f00446d2c91a31e3e77a4aab8145ac3747e9d1ce0165c1aebda95", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.11963v1"}], "id": "arxiv:2610.11963", "identifiers": {"arxiv": "2610.11963"}, "kind": "paper", "last_verified_at": "2026-10-20T13:12:25.581656+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.11963"}], "name": "Can LLMs Fix It Without Code? Toward Automated Verification of No-Code Bug Fixes", "notes": "The evaluation reuses 322 fixes from 12 generation configurations and is restricted to nine of 51 candidate issues after environment setup and sanity checks. Resolution counts any success in three runs with a 20-step episode limit; executor harnesses and decoding settings differ. Opening the update page triggers a harness-controlled switch to the fixed browser build. Checker behavior, reconstructed starting states and the two-author human reference limit interpretation. The replication package was linked by the paper but was not inspected; no software, agent, model or result was run or reproduced.", "published_at": "2026-10-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Tests natural-language no-code fixes through GUI execution in Linux Brave. Scripted fixed/buggy-state gates and issue-specific checkers distinguish fix outcomes from the executor's completion claims.", "topics": ["evaluation", "web-navigation"], "updated_at": "2026-10-08", "url": "https://arxiv.org/abs/2610.11963"}
{"aliases": [], "authors": ["Tian Jin", "Wenhao Wang", "Jing Kang", "Haoting Shi", "Rui Ye", "Yijun Lu", "Siheng Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "06b25b25e804683e11305e21dd04feb9131aad27f901998eda70ac91e527c6b9", "submitted_at": "2026-10-08T18:31:16Z", "updated_at": "2026-10-14T08:40:10Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2610.12611v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.12611", "identifiers": {"arxiv": "2610.12611"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.12611"}], "name": "SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SoftwareGym, a pipeline that synthesizes runnable GUI applications and iteratively repairs them using execution checks. It also provides a task benchmark and studies whether training on synthesized application trajectories transfers to other computer-use environments.", "topics": ["data-generation", "evaluation", "sandbox"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.12611"}
{"aliases": [], "authors": ["Yisheng Li", "Dongming Han", "Hongkun Pan", "Bingcheng Mao", "Zijian Wang", "Yunlong Mao", "Wei Tong", "Sheng Zhong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e6f1521fdfdf6a11a2a195feb05fd915afd6e9bb2332486b870da6375214ac1a", "submitted_at": "2026-10-09T00:03:13Z", "updated_at": "2026-10-09T00:03:13Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.12823v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.12823", "identifiers": {"arxiv": "2610.12823"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.12823"}], "name": "When Charts Deceive Web Agents: A Paired Benchmark for Misleading Visualization-Grounded Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MisVisAgentBench, a paired executable benchmark that changes chart evidence while holding web tasks fixed, to trace how misleading visualizations affect agent decisions and downstream actions.", "topics": ["evaluation", "robustness", "screen-understanding"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2610.12823"}
{"aliases": [], "authors": ["Shangquan Sun", "Ruoyu Chen", "Zhirui Liu", "Xiang Liu", "Chenghao Qian", "Wenqi Ren", "Xiaochun Cao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7f7effcf5e101cfcc7e229fe4aab40d07401b1dd98cf5ce3b566fef863a42190", "submitted_at": "2026-10-09T01:50:26Z", "updated_at": "2026-10-09T01:50:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.12866v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}], "id": "arxiv:2610.12866", "identifiers": {"arxiv": "2610.12866"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.12866"}], "name": "CaptchaAgent: Verifiable Visual Reasoning for CAPTCHA Solving", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents CaptchaAgent, which builds an explicit visual representation of objects and relations, refines ambiguous regions, and checks counterfactual consistency before acting on interactive CAPTCHAs.", "topics": ["grounding", "screen-understanding", "evaluation"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2610.12866"}
{"aliases": [], "authors": ["Yuchen Ling", "Shengcheng Yu", "Zhenyu Chen", "Chunrong Fang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fb5f0220266a43b74e7a5f547abbe78a54a2ef993fee5d27f67827cb9715f5dc", "submitted_at": "2026-10-09T06:45:23Z", "updated_at": "2026-10-09T06:45:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.13092v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T15:06:37Z", "role": "primary", "sha256": "a3d753815e82190ee2de6f322cd816c32f2dd391f862845305fc7e57300303a6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22desktop+agent%22+OR+abs%3A%22computer+interaction%22+OR+abs%3A%22GUI+automation%22+OR+abs%3A%22web+navigation%22+OR+abs%3A%22browser+automation%22+OR+abs%3A%22visual+grounding%22+OR+abs%3A%22screen+understanding%22+OR+abs%3A%22mobile+agents%22%29+AND+submittedDate%3A%5B202609200000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:08:54Z", "role": "primary", "sha256": "3db7474953cbe5a3052fb22793038f2383bdd0e7c775af06494a5e8501646d18", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.13092v1"}], "id": "arxiv:2610.13092", "identifiers": {"arxiv": "2610.13092"}, "kind": "paper", "last_verified_at": "2026-10-20T15:08:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.13092"}], "name": "When Faster Decisions Do Not Make Better Agents: An Empirical Study of Jev in Agent Workflows", "notes": "The mobile evaluation uses text UI observations and one fixed instance for each of 116 AndroidWorld task types; three repeats share seed 0. Five agent families use different budgets, with candidate-interface and tool-call granularity changes in some variants. Reported costs use fixed experimental prices, and outcome-selected trace labels are not population failure rates. Faster selection does not imply faster or more reliable task completion. Selected sections were source-read; no independent reproduction.", "published_at": "2026-10-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates Jev as a bounded action selector inside three AndroidWorld agents and two text-based ALFWorld agents, separating decision savings from task success, recovery and resources per delivered outcome.", "topics": ["efficiency", "action-prediction", "evaluation", "mobile-navigation", "self-reflection"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2610.13092"}
{"aliases": [], "authors": ["Vagner Figueredo de Santana", "Diogo Cortiz"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "447a4f283746069b2ea51d49c149379ad48a6e0b71f35dd15dedcb19417ab0cd", "submitted_at": "2026-10-09T14:59:24Z", "updated_at": "2026-10-09T14:59:24Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.13703v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}], "id": "arxiv:2610.13703", "identifiers": {"arxiv": "2610.13703"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.13703"}], "name": "A Baseline Assessment for the Agentic Web Technology Stack", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents MASTIF, a framework for comparing combinations of agent models, protocols, and implementation frameworks on web tasks, with an initial assessment using 300 Mind2Web tasks.", "topics": ["evaluation", "web-navigation"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2610.13703"}
{"aliases": [], "authors": ["Md Imran Hossen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "63e19265f1df2c336e28c4a13e11331021bd60866269c0b921d92a3c0fec080d", "submitted_at": "2026-10-10T19:30:31Z", "updated_at": "2026-10-10T19:30:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.15138v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.15138", "identifiers": {"arxiv": "2610.15138"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.15138"}], "name": "The Agent Clicks Approve: Injected Page Text Makes Browser Agents Authorize Access the Task Never Named", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies prompt-injected account-access approvals in browser agents and proposes a service-side authority envelope that checks authorization requests against the user task before allowing them to complete.", "topics": ["safety-security", "evaluation"], "updated_at": "2026-10-10", "url": "https://arxiv.org/abs/2610.15138"}
{"aliases": [], "authors": ["Jianlin Chen", "Wenhui Chen", "Chi Man Vong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e6c60bb4d3b7bb2e2cba92201924d8da4f5e18ce5553c81fc9d5c6fb62b84751", "submitted_at": "2026-10-11T16:18:11Z", "updated_at": "2026-10-11T16:18:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.15738v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.15738v1"], "reason": "Publisher document declaration only; no code, data or weights license inferred."}}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T13:48:10Z", "role": "primary", "sha256": "6d8c0463b6d50320246021f0394d0e39c9fd839f1ef1f742c8cdfa231872b059", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22+OR+ti%3AGUI+OR+ti%3A%22Android+agent%22+OR+abs%3A%22computer-use+agents%22+OR+abs%3A%22GUI+agents%22+OR+abs%3A%22browser+agents%22+OR+abs%3A%22mobile+GUI%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:31:34.620763Z", "role": "primary", "sha256": "7ec947ceb462b5fd23a4a67df64e931e427224ef721b681d3fc92c1d4379449f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.15738v1"}], "id": "arxiv:2610.15738", "identifiers": {"arxiv": "2610.15738"}, "kind": "paper", "last_verified_at": "2026-10-20T15:31:34.620763Z", "licenses": {"code": null, "data": null, "document": {"basis": "The retained version-specific arXiv HTML infobox explicitly declares License: CC BY 4.0 for this document.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T15:31:34.620763Z", "url": "https://arxiv.org/html/2610.15738v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.15738"}], "name": "Visible but Not Avoided: State-Dependent Re-Proposal of Rejected Candidates in Catalog-Based LLM Optimization Loops", "notes": "The downstream test uses one screenshot-only agent, nine synthetic apps and 60 cached development items, with TAP-tag or text-answer outputs; this is not demonstrated live-app completion. Two catalogs reword the same 40 edits. Re-proposal effects do not establish internal causal mechanisms or task-success gains. Main serving providers were unpinned; rationale labels came from LLM annotators. No independent reproduction.", "published_at": "2026-10-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies repeated rejected edits while optimizing a screenshot-only GUI agent, contrasting static probes with reached loop states and testing how rejection wording changes subsequent edit proposals.", "topics": ["evaluation", "grounding", "training-free"], "updated_at": "2026-10-11", "url": "https://arxiv.org/abs/2610.15738"}
{"aliases": [], "authors": ["Jiahao Lu", "Yize Shen", "Chuanyuan Tan", "Mianqiu Huang", "Shicheng Fang", "Xinchi Chen", "Yining Zheng", "Yuxin Wang", "Xipeng Qiu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73073d2d533307cd71b05afef2441a8a7b8f198f804f11f83536d4297db2491a", "submitted_at": "2026-10-12T12:52:21Z", "updated_at": "2026-10-12T12:52:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.16720v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.16720", "identifiers": {"arxiv": "2610.16720"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.16720"}], "name": "Learn from Move: the Next Step for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces LatentGUIWorld to test learning interface semantics and dynamics through interaction, and studies finer movement actions with intermediate visual feedback and exploration demonstrations.", "topics": ["evaluation", "grounding", "reinforcement-learning"], "updated_at": "2026-10-12", "url": "https://arxiv.org/abs/2610.16720"}
{"aliases": [], "authors": ["Zian Yuan", "Wei Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c5a2743a18bae1deeed13fbd1dd7379c3f21c67488627057ccb466ffe51d09dc", "submitted_at": "2026-10-12T13:34:52Z", "updated_at": "2026-10-12T13:34:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.16778v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.16778", "identifiers": {"arxiv": "2610.16778"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.16778"}], "name": "PrivacyGuard: Source-Linked Sanitization and Locally Resolvable PII References for Cloud GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces PrivacyGuard, which sanitizes source-linked personal information and maintains locally resolvable references across mobile-agent turns to reduce exposure to remote planners.", "topics": ["safety-security", "mobile-navigation", "evaluation"], "updated_at": "2026-10-12", "url": "https://arxiv.org/abs/2610.16778"}
{"aliases": [], "authors": ["Yuetian Chen", "Kaiyuan Zhang", "Yuntao Du", "Thomas Li", "Mark Tenenholtz", "Kyle Polley", "Jerry Ma", "Denis Yarats", "Ninghui Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "247cd4c9634dea78780a2230138c728adf53871e8b65614697f76ba53dea858c", "submitted_at": "2026-10-12T22:17:51Z", "updated_at": "2026-10-12T22:17:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.17457v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}], "id": "arxiv:2610.17457", "identifiers": {"arxiv": "2610.17457"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.17457"}], "name": "BrowseSafe++: Benchmarking Prompt Injection Detection Across Web Contexts, Languages, and Browser Views", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Extends BrowseSafe with paired attacked and clean webpages in several representations, and studies text and screenshot detectors plus feedback-adapted prompt injections.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": "2026-10-12", "url": "https://arxiv.org/abs/2610.17457"}
{"aliases": [], "authors": ["Wayne Chi", "I-Hung Hsu", "Li Chen", "Wei-Lin Chiang", "Ion Stoica", "Anastasios Nikolas Angelopoulos", "Ameet Talwalkar", "Chris Donahue", "Cho-Jui Hsieh"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "14d2128155d1540955e54c3bdf9010724f27cf2e101079650315602856c98efa", "submitted_at": "2026-10-13T07:49:50Z", "updated_at": "2026-10-14T06:37:47Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2610.17895v2"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}], "id": "arxiv:2610.17895", "identifiers": {"arxiv": "2610.17895"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.17895"}], "name": "Interaction-as-a-Judge", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses computer-use agents to interact with generated applications, studying whether interaction improves preference judging, requirement checks, and feedback for application generation.", "topics": ["evaluation", "data-generation", "automation"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.17895"}
{"aliases": [], "authors": ["Hang He", "Linting Wang", "Guanming Liu", "Bo Mao", "Jinxiu Liu", "Yiran Suo", "Maosen Zhou", "Zhiheng Xi", "Guanting Dong", "Zhou Shao", "Rui Zheng", "Jie Zhou", "Chengcheng Wan", "Tao Gui", "Qi Zhang", "Xipeng Qiu", "Xuanjing Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6add363f25f87ea1814c7b79ebf6b51f7431b3745a7d2d910960b78e2d97d13f", "submitted_at": "2026-10-13T09:52:02Z", "updated_at": "2026-10-13T09:52:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.18067v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.18067v1"], "reason": "Document declaration only; code, data and weights permissions are not inferred."}}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:39Z", "role": "primary", "sha256": "c52e4e283e0dae88ef8b3ce67b7424b1d4b48e8f41015d872f4da69f98782eaa", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.39924%2C2610.00346%2C2610.00583%2C2610.01939%2C2610.05099%2C2610.09127%2C2610.09484%2C2610.09633%2C2610.11251%2C2610.11922%2C2610.13098%2C2610.13166%2C2610.13361%2C2610.13623%2C2610.16117%2C2610.17953%2C2610.18067%2C2610.19366%2C2610.20424&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T19:02:20Z", "role": "primary", "sha256": "f1fbdd2ecb2a54b7723372118e2b0ac818420f3a3dd280c2486c8e064cf486d6", "supports": ["name", "kind", "summary", "scope", "environments", "topics", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.18067v1"}], "id": "arxiv:2610.18067", "identifiers": {"arxiv": "2610.18067"}, "kind": "paper", "last_verified_at": "2026-10-20T19:02:20Z", "licenses": {"code": null, "data": null, "document": {"basis": "The exact-v1 arXiv HTML infobox explicitly declares CC BY 4.0 for this paper.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T19:02:20Z", "url": "https://arxiv.org/html/2610.18067v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.18067"}], "name": "WorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment Evolution", "notes": "App WorkEnvs expose UI observations and interaction interfaces alongside file and tool workflows; no specific operating system or pixel-control success is established. Verifier shaping depends on deterministic-check and dependency coverage. External reference scores use differing harnesses or subsets; GDPval Elo is not comparable to percentage metrics. A successful-trace analysis subset is separate from online RL outcomes. No implementation release or independent reproduction was verified.", "published_at": "2026-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Synthesizes resettable workspaces with files, application observations and skills; repairs environment incompatibilities and trains long-task agents using verifier-linked turn credit.", "topics": ["data-generation", "long-horizon", "evaluation", "reward-models", "reinforcement-learning", "workflow"], "updated_at": "2026-10-13", "url": "https://arxiv.org/abs/2610.18067"}
{"aliases": [], "authors": ["Yubo Li", "Ramayya Krishnan", "Rema Padman"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "070f02f1ea798369817f471df5fe1d002b68d544135a1ac8e3a2aed2830e8c01", "submitted_at": "2026-10-13T15:43:30Z", "updated_at": "2026-10-13T15:43:30Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.18579v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.18579v1"], "reason": "The retained version-specific arXiv HTML infobox explicitly labels this paper License: CC BY 4.0. This document declaration does not establish code, training-data or model-weight licenses."}}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T15:06:37Z", "role": "primary", "sha256": "a3d753815e82190ee2de6f322cd816c32f2dd391f862845305fc7e57300303a6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22desktop+agent%22+OR+abs%3A%22computer+interaction%22+OR+abs%3A%22GUI+automation%22+OR+abs%3A%22web+navigation%22+OR+abs%3A%22browser+automation%22+OR+abs%3A%22visual+grounding%22+OR+abs%3A%22screen+understanding%22+OR+abs%3A%22mobile+agents%22%29+AND+submittedDate%3A%5B202609200000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:08:55Z", "role": "primary", "sha256": "bcbc640d1e732ba0415a9755bce120a811a13a2ba7fba3e58a167b3a9111c2a6", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.18579v1"}], "id": "arxiv:2610.18579", "identifiers": {"arxiv": "2610.18579"}, "kind": "paper", "last_verified_at": "2026-10-20T15:08:55Z", "licenses": {"code": null, "data": null, "document": {"basis": "The retained version-specific arXiv HTML infobox explicitly labels this paper License: CC BY 4.0. This document declaration does not establish code, training-data or model-weight licenses.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T15:08:55Z", "url": "https://arxiv.org/html/2610.18579v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.18579"}], "name": "Reading Right, Acting Wrong: Diagnosing Decision Failures of System One Models in Agent Loops", "notes": "Adjacent diagnostic research across simulators, policies, tools, coding and an offline Mind2Web subset. Models choose among supplied admissible actions; they do not generate actions or execute a closed-loop browser policy in the reported evaluation. The paired probes measure selected visible facts, not complete understanding or internal causation. Pooled results are not a web-only leaderboard, and the paper reports coverage limits and label-sensitivity analyses. No benchmark release, code, data or weights were independently inspected.", "published_at": "2026-10-13", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Pairs factual probes with next-action choices on frozen agent traces, including Mind2Web, to diagnose reading-decision gaps without closed-loop execution.", "topics": ["evaluation", "web-navigation", "interaction-history"], "updated_at": "2026-10-13", "url": "https://arxiv.org/abs/2610.18579"}
{"aliases": [], "authors": ["Yuxiang Lai", "Jike Zhong", "Chengzhi Liu", "Ming Li", "Haonan Wang", "Xiaofeng Yang", "Liang Zhao", "Youbao Tang", "Ruixi Fan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5f72958f9a70e7714f17b0eb51c5453e08ea8ad4e8e7d4caa5e3170231088251", "submitted_at": "2026-10-13T20:01:18Z", "updated_at": "2026-10-13T20:01:18Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.19128v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.19128", "identifiers": {"arxiv": "2610.19128"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.19128"}], "name": "CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation Pipeline", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CLASHarness, generating verifiable computer-use tasks with conflicting agent reports, shared-state constraints, and coordination requirements to evaluate conflict detection and resolution.", "topics": ["evaluation", "robustness", "workflow"], "updated_at": "2026-10-13", "url": "https://arxiv.org/abs/2610.19128"}
{"aliases": [], "authors": ["Minghao Guo", "Qingyue Jiao", "Xiujin Liu", "Jianhong Pang", "Ruimeng Yang", "Naihao Xu", "Liwei Che", "Haochen Xue", "Ziqi Wang", "Xinying Cai", "Jiaxuan Luo", "Jiaxin Deng", "Xuesong Ye", "Jintao Qu", "Wenyao Gao", "Jingwen Gu", "Chenyang Zhao", "Shichun Liu", "Guibin Zhang", "Zefang Liu", "Siyin Wang", "Wujiang Xu", "Danrui Li", "Zeru Shi", "Woongyeong Yeo", "Shuang Liu", "Wei-Chieh Huang", "Weizhi Zhang", "Tianyu Yang", "Kangsan Kim", "Yebowen Hu", "Bowen Wen", "Bowen Jiang", "Haohao Su", "Ming Yin", "Sheldon Yu", "Mian Zhang", "Yuyan Chen", "Hanrong Zhang", "Kai Mei", "Zihao Zhang", "Mingyu Jin", "Xi Zhu", "Hengyi Wang", "Tianxin Wei", "Linyong Nan", "Zhenwen Liang", "Zhendong Chu", "Ziqi Huang", "Yongye Su", "Sixun Dong", "Shujian Liu", "Kaiqu Liang", "Jiaojiao Han", "Zhaoye Fei", "Kun Wang", "Shihan Dou", "Zhichun Guo", "Xingjian Diao", "Zeqi Gu", "Heng Wang", "Tao Feng", "Cheng Qian", "Hamid Palangi", "Silei Xu", "Xi Yun", "Yi Zhu", "Shilong Liu", "Qingsong Wen", "Sung Ju Hwang", "Zuxuan Wu", "Yu-Gang Jiang", "Yiyu Shi", "Dimitris N. Metaxas", "Jiaxuan You", "Julian McAuley", "Mengdi Wang", "Jiawei Han", "Philip S. Yu", "Ziyi Ye"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7f1a48aa4b89fb97e8e1a186ba7345826a8b9efc9aa7b82e0dd1f5ce17d42b73", "submitted_at": "2026-10-14T01:11:57Z", "updated_at": "2026-10-14T01:11:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.19366v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.19366v1"], "reason": "Paper-specific document declaration only; no code, data or weights permissions are inferred."}}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:39Z", "role": "primary", "sha256": "c52e4e283e0dae88ef8b3ce67b7424b1d4b48e8f41015d872f4da69f98782eaa", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.39924%2C2610.00346%2C2610.00583%2C2610.01939%2C2610.05099%2C2610.09127%2C2610.09484%2C2610.09633%2C2610.11251%2C2610.11922%2C2610.13098%2C2610.13166%2C2610.13361%2C2610.13623%2C2610.16117%2C2610.17953%2C2610.18067%2C2610.19366%2C2610.20424&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T19:02:22.864948+00:00", "role": "primary", "sha256": "308796868bf9b1cb042bc6008e1e6f4ba6e776a2d4e13d40ad5684b477861563", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.19366v1"}], "id": "arxiv:2610.19366", "identifiers": {"arxiv": "2610.19366"}, "kind": "survey", "last_verified_at": "2026-10-20T19:02:22.864948+00:00", "licenses": {"code": null, "data": null, "document": {"basis": "The retained exact-v1 arXiv HTML infobox explicitly declares License: CC BY 4.0 for this document.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T19:02:22.864948+00:00", "url": "https://arxiv.org/html/2610.19366v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.19366"}], "name": "Seeing, Maintaining, and Learning: The Evolution of Multimodal Agent Memory", "notes": "A broad literature synthesis spanning images, video, audio, GUI and embodied settings. Its GUI section compares interface-state retention, action outcomes, cross-session reuse and visual evidence, but the paper does not provide a new GUI-agent experiment. Taxonomy corpus counts and benchmark classifications are authors' annotations, not independently audited field coverage. Updating stored content alone does not establish the survey's self-evolving-policy tier; endpoint task success does not isolate memory use. Lifecycle, resource-budget and provenance recommendations are proposed evaluation directions rather than demonstrated guarantees. Platform facets stay unspecified; cited systems and releases were not independently verified.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Surveys multimodal memory using representation and management axes, distinguishing fixed-policy updates from policy evolution and comparing GUI memory evaluation with other modalities.", "topics": ["survey", "memory", "evaluation", "interaction-history"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.19366"}
{"aliases": [], "authors": ["Yuke Wang", "Benlei Cui", "Shen Pang", "Xuemei Dong", "Longtao Huang", "Hui Xue", "Yuwen Zhai", "Junjie Li", "Jingqun Tang", "Haiwen Hong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "da334d943d98b4bf9915c6ca28aab734d736f376e67cbab950aa1af65f540ebf", "submitted_at": "2026-10-14T01:30:03Z", "updated_at": "2026-10-14T01:30:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.19384v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.19384v1"], "reason": "Paper-specific document declaration only; no code, data or weights permissions are inferred."}}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:01Z", "role": "primary", "sha256": "63e37cd26bfe69455a3fa835f3d6ba4cfcc5bb7dc91f90505920d37b1e08f0b6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.19384%2C2610.22672%2C2610.22697%2C2610.23303%2C2610.23307%2C2610.24729&max_results=6"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T19:02:16.329923+00:00", "role": "primary", "sha256": "ca995ec330f62a43d3d4fbbf28dfe83c5fa8b05f5c1e446929493409760268d1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.19384v1"}], "id": "arxiv:2610.19384", "identifiers": {"arxiv": "2610.19384"}, "kind": "paper", "last_verified_at": "2026-10-20T19:02:16.329923+00:00", "licenses": {"code": null, "data": null, "document": {"basis": "The retained exact-v1 arXiv HTML infobox explicitly declares License: CC BY 4.0 for this document.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T19:02:16.329923+00:00", "url": "https://arxiv.org/html/2610.19384v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.19384"}], "name": "MMA-SafetyBench: A Benchmark for Multimodal Agent Safety Evaluation", "notes": "ASR measures predicted coordinates inside an annotated target or compromised generated content; actual clicks, environment changes and downstream harm are not verified. SCR additionally counts adopted adversarial intent or task/schema disruption. Clean-task and attacked-trajectory denominators differ, including 42 clean desktop tasks versus 209 adversarial desktop trajectories. Payloads and injection points are fixed, and neutral prompts change both compliance cues and instruction-priority guidance. GPT-5 judging has a 300-trajectory human-label check, with borderline disagreements. Android-style mobile dialogs do not establish an Android runtime. Code/data links were identified but not inspected; no independent reproduction.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds trajectory-guided visual-injection tests for web, desktop and mobile observations, separating target-matching action predictions from broader semantic compromise.", "topics": ["evaluation", "safety-security", "grounding"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.19384"}
{"aliases": [], "authors": ["Ridong Zhang", "Shuxian Liang", "Chaoqun Wan", "Zhen Huang", "Xian-Sheng Hua"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "50d5cb78e6715f0749c5973e900745ecdb3672febc2c8e090da94cb708127352", "submitted_at": "2026-10-14T09:58:40Z", "updated_at": "2026-10-14T09:58:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.19900v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:46:06Z", "role": "primary", "sha256": "652e8369ef11334d76c0eb67afeaa5c9f7841fd31ca02243b0dfb1493316849a", "supports": ["environments", "notes"], "url": "https://iaie-physicalai.github.io/GraceBench"}], "id": "arxiv:2610.19900", "identifiers": {"arxiv": "2610.19900"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.19900"}], "name": "How GUI Agents Fail: GraceBench for Execution Stability Diagnosis", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced. Android platform tagging is additionally supported by the primary project documentation.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GraceBench, using application transition graphs and recorded interaction evidence to diagnose deviations, loops, and recovery that final task-success scores can hide.", "topics": ["evaluation", "mobile-navigation", "robustness"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.19900"}
{"aliases": [], "authors": ["Mingyu Ouyang", "Yiran Wang", "Siyuan Hu", "Xingyilang Yin", "Huiqiang Sun", "Guangzhi Wang", "Junfu Pu", "Cheng Cheng", "Honghao Chen", "Ying Shan", "Mike Zheng Shou", "Hwee Tou Ng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "67b539468bd2bbf0d3b49a5a7eed4791a8ddca4b56ad9515d23a29204d1f069f", "submitted_at": "2026-10-14T17:57:12Z", "updated_at": "2026-10-14T17:57:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.20584v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}], "id": "arxiv:2610.20584", "identifiers": {"arxiv": "2610.20584"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.20584"}], "name": "Player-1: Learning to Play from Synthetic Experience", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Extends computer-use policies to game control through synthetic training environments and richer keyboard and pointer actions. Its evaluation concerns games rather than ordinary productivity workflows.", "topics": ["data-generation", "reinforcement-learning", "evaluation"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.20584"}
{"aliases": [], "authors": ["Panrong Tong", "Quyu Kong", "Pengxiang Li", "Tianyu Xia", "Hanzhang Zhou", "Lei Wang", "Chenglin Cai", "Long Li", "Long Chen", "Jianan Zhang", "Zhengwen Zeng", "Xu Zhang", "Gongjie Zhang", "Yue Wang", "Steven Hoi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a5f4eda178718aac211eddc7281d572dc1d189dfd0cd130d4eba9f30f5a05ca2", "submitted_at": "2026-10-14T21:06:01Z", "updated_at": "2026-10-14T21:06:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.20939v1"}}, "environments": ["cross-platform", "desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.20939", "identifiers": {"arxiv": "2610.20939"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.20939"}], "name": "MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Unifies GUI evaluation, trajectory collection, environment synthesis, and online reinforcement learning across mobile, computer, and web tasks.", "topics": ["sandbox", "evaluation", "reinforcement-learning"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.20939"}
{"aliases": [], "authors": ["Tengda Guo", "Jiawang Hu", "Yifei Tao", "Yin Wang", "Haotian Hu", "Fanyi Wang", "Liujuan Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "025c7018c86ae4de79a659242070b2f91c5913333e9ecd4bd7fd357d13d27a34", "submitted_at": "2026-10-14T14:39:41Z", "updated_at": "2026-10-14T14:39:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22408v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T12:58:20.217517Z", "role": "primary", "sha256": "f275c47a8b6d349a70e39a9e55eea5a09d99d5f35cc896100f3bcfd5b0923e0b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22computer-use+agents%22+OR+abs%3A%22GUI+agents%22+OR+abs%3A%22browser+agents%22+OR+abs%3A%22mobile+GUI%22%29+AND+submittedDate%3A%5B202609200000+TO+202610201255%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:54.814500Z", "role": "primary", "sha256": "0389d96a8f6c4975129f5c526e8df9ca4690f31364a2f4e66aaea4a75f0eabbc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.22408v1"}], "id": "arxiv:2610.22408", "identifiers": {"arxiv": "2610.22408"}, "kind": "paper", "last_verified_at": "2026-10-20T13:12:54.814500Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.22408"}], "name": "ECHOES: Seeing Beyond the Current Image via Recursive History Compression and Privileged On-Policy Distillation", "notes": "Privileged screenshot history is used during training, while deployment retains self-generated textual state. GUI results use action/step metrics and model-specific history settings; online Sokoban transfer is separate. GRPO rewards action correctness, valid output format and compression length. Diagnostic memory replay is distinct from the validation gap used to schedule training. No independent reproduction.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "ECHOES learns recursive textual task memory through privileged screenshot-history distillation and GRPO, then predicts GUI actions from the current screenshot and memory.", "topics": ["memory", "interaction-history", "state-tracking", "long-horizon", "imitation-learning", "reinforcement-learning", "efficiency", "evaluation"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.22408"}
{"aliases": [], "authors": ["Adamo Orsini", "Seyed Mehdi Ayyoubzadeh", "Sadman Rakib Pinon", "Hannuo Zhang", "Koosha Gholipour Baradari", "Yuanhao Yu", "Xinxin Zuo", "Yang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "caaf24ef4d8b86c8aaebc852235d0fe947a5ed859dc1ba8329ce07a2f7be9aa1", "submitted_at": "2026-10-15T21:40:34Z", "updated_at": "2026-10-15T21:40:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22672v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:01Z", "role": "primary", "sha256": "63e37cd26bfe69455a3fa835f3d6ba4cfcc5bb7dc91f90505920d37b1e08f0b6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.19384%2C2610.22672%2C2610.22697%2C2610.23303%2C2610.23307%2C2610.24729&max_results=6"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T12:56:56.464343+00:00", "role": "primary", "sha256": "5fd9d7926e99d068fa8fb7238a9e7811e3f8c55d00e6b89d5672754cb3e619bc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.22672v1"}], "id": "arxiv:2610.22672", "identifiers": {"arxiv": "2610.22672"}, "kind": "paper", "last_verified_at": "2026-10-20T12:56:56.464343+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.22672"}], "name": "Not All Graphs Are Maps: A Task-Agnostic Exploration Framework for Mobile GUI Navigation", "notes": "The controlled cohort has 84 tasks across six apps; the broader inventory spans 20 apps. Main results use human destination labels. DroidTask and SPA-Bench are converted to navigation-only protocols. Some comparison cells received another graph condition’s memory, with sensitivity analyses reported. Cross-family runs differ in app versions, state and denominators. A small stale-graph probe does not establish robustness. Models, code, artifacts and performance were not independently reproduced.", "published_at": "2026-10-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "TAG-Nav supplies Android GUI agents with destination and route memory distilled from offline app exploration. DestBench measures arrival at a reference pre-action screen, separately from downstream task execution.", "topics": ["mobile-navigation", "memory", "evaluation"], "updated_at": "2026-10-15", "url": "https://arxiv.org/abs/2610.22672"}
{"aliases": [], "authors": ["Haoming Li", "Defu Cao", "Xu Hu", "Yingtao Luo", "Yiyang Wang", "Xianghe Xu", "Jiashu He", "Wanting Jiang", "Yifan Zhang", "Yunhan Zhao", "Siyang Wu", "Chen Yang", "Haixin Wang", "Zheyuan Deng", "Taoran Li", "Jingwen Zhang", "Zhejian Peng", "Shushu Liang", "Yue Zhao", "Bingzhe Li", "Feng Chen", "Yifan Wang", "Jintao Huang", "Yijun Wang", "Yuexing Hao", "Xiaomin Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f613ef7ada260b494a411058bc733c7a5e0b074cdb01906c38132acf7275b9b1", "submitted_at": "2026-10-16T06:14:55Z", "updated_at": "2026-10-16T06:14:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22990v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-19T04:42:05Z", "role": "primary", "sha256": "ffbb3f28a3a0fc8888f30ea8b2cdaf441239685aaacc2d80f311581f4b5a1d5c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610050000+TO+202610190000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T10:42:06.870916+00:00", "role": "primary", "sha256": "bf9066e8b3472100f120e21534f92655f4d0814f2521b9d23d778e5239c7a2a4", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.22990v1"}], "id": "arxiv:2610.22990", "identifiers": {"arxiv": "2610.22990"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.22990"}], "name": "PersonaCUABench: Benchmarking Computer-Use Agents Serving Diverse Personas", "notes": "Reviewed the complete retained abstract and selected exact-v1 construction, evaluation and limitations sections. Main interactions and satisfaction ratings use LLM-simulated personas; human annotators validate selected outputs. The harness observes page text, with screenshots only for auditing. Results are not evidence about real demographic populations or desktop agents. No independent reproduction.", "published_at": "2026-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "PersonaCUABench varies simulated user requests and clarification dialogue over fixed web tasks, comparing task completion with requirement handling and persona-agent satisfaction ratings.", "topics": ["evaluation", "web-navigation", "human-in-the-loop"], "updated_at": "2026-10-16", "url": "https://arxiv.org/abs/2610.22990"}
{"aliases": [], "authors": ["Songping Wang", "Borui Li", "Qinglong Liu", "Yueming Lyu", "Yi Huang", "Fengting Li", "Chao Feng", "Caifeng Shan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ea753c1cb0666950d89f2781511f26872e3d5b1c2fd2434896248214160fa22c", "submitted_at": "2026-10-16T11:06:37Z", "updated_at": "2026-10-16T11:06:37Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.23303v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:01Z", "role": "primary", "sha256": "63e37cd26bfe69455a3fa835f3d6ba4cfcc5bb7dc91f90505920d37b1e08f0b6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.19384%2C2610.22672%2C2610.22697%2C2610.23303%2C2610.23307%2C2610.24729&max_results=6"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T10:42:09.967883+00:00", "role": "primary", "sha256": "ec9d647818be346ddbf7807be87b4789cd53e50c5d3d7a7904ad4f49bc71c38e", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.23303v1"}], "id": "arxiv:2610.23303", "identifiers": {"arxiv": "2610.23303"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.23303"}], "name": "MobileJailBench: Red Queen Jailbreak Benchmarking for Mobile Agents", "notes": "Reviewed the complete retained abstract and selected exact-v1 threat-model, construction and evaluation sections. Execution uses a browser-based Android-like simulator, not real Android isolation or live services. The attacker controls instructions and designated carriers, not privileged simulator state/APIs or system/model/harness internals. Rates depend on release, budget and adjudication coverage. SAF labels track observable expressions and actions, not latent reasoning; unresolved episodes are excluded from rate denominators. No independent reproduction.", "published_at": "2026-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileJailBench evaluates mobile-agent jailbreaks in a browser-based Android-like simulator, combining feedback-driven attack evolution, separately admitted task expansions and evidence-grounded outcome adjudication.", "topics": ["safety-security", "mobile-navigation", "evaluation", "observability", "cross-app"], "updated_at": "2026-10-16", "url": "https://arxiv.org/abs/2610.23303"}
{"aliases": [], "authors": ["Songping Wang", "Peiyang Li", "Yi Huang", "Fengting Li", "Xuhao Ran", "Yueming Lyu", "Chao Feng", "Caifeng Shan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3fc22fbeda19ac2947b00af73801f0674c550e0d7c78919f811cae754b06128c", "submitted_at": "2026-10-16T11:10:30Z", "updated_at": "2026-10-16T11:10:30Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.23307v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:01Z", "role": "primary", "sha256": "63e37cd26bfe69455a3fa835f3d6ba4cfcc5bb7dc91f90505920d37b1e08f0b6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.19384%2C2610.22672%2C2610.22697%2C2610.23303%2C2610.23307%2C2610.24729&max_results=6"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T09:17:36.223826+00:00", "role": "primary", "sha256": "3803984a185b6258276b87b87ac1991583878fb41dc22618ff743f6f69c7bc7e", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.23307v1"}], "id": "arxiv:2610.23307", "identifiers": {"arxiv": "2610.23307"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.23307"}], "name": "MobileInject: Evaluating Reasoning-to-Action Security of MLLM-Based Mobile Agents under Cross-Modal Consistency Injection", "notes": "Reviewed the complete abstract and selected v1 construction, threat-model and evaluation sections. The attacker modifies observation surfaces, not system instructions, model parameters or the OS. State oracles are distinct from emitted-rationale diagnoses, whose labels include outcome-dependent constraints. Those diagnostics are not independent evidence of hidden-reasoning causality. No reproduction or release verification.", "published_at": "2026-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileInject evaluates visual/textual observation injection in stateful Android workflows using resettable emulators, benign/adversarial state oracles and trace-based diagnosis of emitted rationales and actions.", "topics": ["safety-security", "mobile-navigation", "evaluation", "observability", "cross-app"], "updated_at": "2026-10-16", "url": "https://arxiv.org/abs/2610.23307"}
{"aliases": [], "authors": ["Xinyu Fang", "Yuchi Wang", "Weihao Yu", "Rongyao Fang", "Zhaohai Li", "Zhixun Li", "Tiancheng Gu", "Xilin Wei", "Yuhang Zang", "Hongsheng Li", "Dahua Lin", "Guofeng Zhang", "Shuai Bai", "Zhibo Yang", "Dayiheng Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6dfa0c59ec1262cebcf2a4e61ab1e0b58c647513d8e6eedb2ce7e6a1a675558c", "submitted_at": "2026-10-14T12:46:54Z", "updated_at": "2026-10-14T12:46:54Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.23986v1"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T13:48:11Z", "role": "primary", "sha256": "89e110b6fa882639fb73936cb27e5373d11b632938c31968f563152534bf5670", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.14966%2C2610.23986&max_results=2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:09:58.148629Z", "role": "primary", "sha256": "bef4399f087e484b62ff5b8dc2c283f92e0357b94c93b464b86fd97fe7b069cb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.23986v1"}], "id": "arxiv:2610.23986", "identifiers": {"arxiv": "2610.23986"}, "kind": "paper", "last_verified_at": "2026-10-20T15:09:58.148629Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.23986"}], "name": "BlenderWorld: Benchmarking Visual Agents for 3D Creation from Real-World References", "notes": "The evaluated agents use Linux shell/Python/MCP tools and can inspect viewport screenshots; graphical CUA control and Windows/macOS are future work. Each agent is run once per task, with different Codex/Claude Code harnesses across systems. Scores combine weighted rubric satisfaction and model-rated visual quality. Tool counts include unsuccessful calls and do not establish matched compute or successful actions. No independent reproduction or artifact-release verification.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Benchmarks agents creating editable Blender projects through shell, Python and MCP tools, judging delivered scenes with multi-view renders, task-specific rubrics and visual-quality ratings.", "topics": ["evaluation", "tool-integration", "sandbox"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.23986"}
{"aliases": [], "authors": ["Egor Zverev", "Juhee Kim", "Leon Trampert", "Boris Köpf", "Aashish Kolluri", "Shruti Tople", "Santiago Zanella-Béguelin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e675a3082f571da547fb5b6cb89e4f68a034bfa57236886e1218cb2bb2cdf1c3", "submitted_at": "2026-10-16T22:12:45Z", "updated_at": "2026-10-16T22:12:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.24277v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T09:17:33.132448+00:00", "role": "primary", "sha256": "daceefecb2e6b910a6c7f2ffe35e89c7070f6282f4f35ee89e885a2e77aa94c4", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.24277v1"}], "id": "arxiv:2610.24277", "identifiers": {"arxiv": "2610.24277"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.24277"}], "name": "SPIDES: Information-Flow Control for Browser Agents", "notes": "Reviewed the complete abstract and selected v1 threat-model, policy, evaluation and limitations sections. Guarantees assume correct labels/trusted origins and a trusted browser/harness mediating actions; evaluation uses idealized approvals. Origin-level reader sets can miss recipient distinctions within one origin. No unrestricted prompt-injection immunity or independent reproduction is claimed.", "published_at": "2026-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "SPIDES propagates HTML integrity and confidentiality labels into agent context and mediates browser actions with information-flow policies and approval gates, evaluated in the labeled MicroNet ecosystem.", "topics": ["safety-security", "web-navigation", "human-in-the-loop", "evaluation"], "updated_at": "2026-10-16", "url": "https://arxiv.org/abs/2610.24277"}
{"aliases": [], "authors": ["Sizhe Tang", "Hanhan Zhou", "Jeffrey Svajlenko", "Rashmi Gangadharaiah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9d6808cf26a609b50b7835ae3ecece0a13580aafcd589823f0b1eeff54732f69", "submitted_at": "2026-10-17T02:03:20Z", "updated_at": "2026-10-17T02:03:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.24372v1"}}, "environments": ["android", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T12:58:15.922053Z", "role": "primary", "sha256": "b379e14eefa272f62641224fe1d839b7dfbc414d957124b8abeb103444a843ef", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3AGUI+OR+ti%3A%22computer-using%22+OR+ti%3A%22agentic+browsing%22+OR+ti%3A%22visual+grounding%22+OR+ti%3A%22UI+automation%22%29+AND+submittedDate%3A%5B202608210000+TO+202610201255%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:51.579373Z", "role": "primary", "sha256": "9dff18935a1faa7ec16366bac6a6e9e41f0381d561568c3aba39fe783efcfbd3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.24372v1"}], "id": "arxiv:2610.24372", "identifiers": {"arxiv": "2610.24372"}, "kind": "paper", "last_verified_at": "2026-10-20T13:12:51.579373Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.24372"}], "name": "Beyond Next-Screen Prediction: The Decision Value of GUI World Models", "notes": "The matched augmentation experiment measures offline ranking; AndroidWorld task-success comparisons use differently trained evaluators and do not isolate prediction value. A separate web evaluator is trained on Mind2Web, and web results do not uniformly favor direct scoring. The main study concerns one-step selection. No independent reproduction.", "published_at": "2026-10-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Separates next-screen prediction from action evaluation in GUI lookahead and trains Direct Action Ranking on reference actions, comparing prediction-augmented scoring under matched offline supervision.", "topics": ["evaluation", "world-models", "action-prediction", "imitation-learning"], "updated_at": "2026-10-17", "url": "https://arxiv.org/abs/2610.24372"}
{"aliases": [], "authors": ["Zhanhui Zhou", "Weichen Zhang", "Haoran Liu", "Lingjie Chen", "Tianneng Shi", "Kevin Lin", "Zhengyuan Yang", "Lijuan Wang", "Dawn Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0a297401f7ad4dcd2f485ade6bab295e365280bb803dcecc111aebf0693dd203", "submitted_at": "2026-10-19T09:30:54Z", "updated_at": "2026-10-19T09:30:54Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26071v1"}}, "environments": ["desktop", "linux", "windows", "web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:28:55.967592+00:00", "role": "primary", "sha256": "d4231fa9677e916ca8cc27493dfbbe9fb34c7430207150e6b9be10c2bad43bfb", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26071v1"}], "id": "arxiv:2610.26071", "identifiers": {"arxiv": "2610.26071"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26071"}], "name": "CUA-Lite: An Open Platform for Computer Use Agents", "notes": "Reviewed the complete official abstract and selected v1 methods, training and limitations sections. WebVoyager uses offline mirrors; VM-free sandboxes are Linux-only, and reported desktop/mobile gains do not establish unseen-app transfer. No results or released artifacts were independently reproduced.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "CUA-Lite unifies environment interfaces, supervised data formats and per-model harnesses for evaluation, reinforcement learning and supervised fine-tuning across desktop, browser and mobile tasks.", "topics": ["tool-integration", "data-generation", "evaluation", "imitation-learning", "reinforcement-learning", "sandbox"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26071"}
{"aliases": [], "authors": ["Zhizhou Wang", "Yaxi Lu", "Zhong Zhang", "Zikun Hu", "Xin Cong", "Yankai Lin", "Fangming Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "08ef7c9cdda652f60d3f2d4ab89e7218ad5a2b85f5c76f0277c289db06fd2796", "submitted_at": "2026-10-19T14:17:33Z", "updated_at": "2026-10-19T14:17:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26469v1"}}, "environments": ["desktop", "linux", "web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:28:59.073204+00:00", "role": "primary", "sha256": "f13f0225c721d5aa430bb8aea1d1e7a7da5973760c4c3b82f26ba648aade82ba", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26469v1"}], "id": "arxiv:2610.26469", "identifiers": {"arxiv": "2610.26469"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26469"}], "name": "VeriCUA: Learning Proactive Verification for Reliable GUI Agent Rewards", "notes": "Reviewed the complete abstract and selected v1 method/evaluation sections. The Judge is trained on OSWorld and frozen for MobileWorld/WebArena transfer; downstream Actor RL is separate. Evaluation uses screened terminal outcomes, and reported cost counts Probe tokens. No independent reproduction.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "VeriCUA trains a Judge to select verification checks and stopping decisions while a frozen Probe inspects the post-task GUI. Single-attempt reinforcement learning balances verdict correctness and Probe use.", "topics": ["reward-models", "reinforcement-learning", "evaluation", "efficiency"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26469"}
{"aliases": [], "authors": ["Zhicheng Zheng", "Cheng Zhang", "Shengping Yu", "Tianxiang Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7dc2ceb9aadbf7506efc068df99326fea5b60fdc6feb340c490841b88aa19c30", "submitted_at": "2026-10-19T14:38:41Z", "updated_at": "2026-10-19T14:38:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26503v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:29:02.225995+00:00", "role": "primary", "sha256": "0554aa531a5f94f51cba6c51bd1b13fb05b0f462497afa5d1dae7ec6fc6e375a", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26503v1"}], "id": "arxiv:2610.26503", "identifiers": {"arxiv": "2610.26503"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26503"}], "name": "ScopeMem: Latent Memory That Learns Where to Revise Frozen GUI Agents", "notes": "Reviewed the complete abstract and selected v1 controlled-study, method and limitations sections. Results are offline, step-level action matching; online task success is explicitly left untested. The frozen actor is queried twice per step. No independent reproduction or redistribution-license claim.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "ScopeMem conditions latent-memory prefixes on a frozen GUI agent's initial action and trains them to correct errors while preserving correct decisions. Its evaluation separates action repairs from harmful revisions.", "topics": ["memory", "interaction-history", "imitation-learning", "robustness", "evaluation"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26503"}
{"aliases": [], "authors": ["Ananya Gubbi Mohanbabu", "Rosiana Natalie", "Anhong Guo", "Amy Pavel"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9858b5ec15ee953504bd00e26b455c1e81104f80429c2dd112de17f9d23977ab", "submitted_at": "2026-10-19T16:39:11Z", "updated_at": "2026-10-19T16:39:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26703v1"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T09:17:30.049627+00:00", "role": "primary", "sha256": "9a964cbf50123887ad20bf156e69c0bbe4567d6b2c28e4cf202d5538eab8294c", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26703v1"}], "id": "arxiv:2610.26703", "identifiers": {"arxiv": "2610.26703"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26703"}], "name": "Comparing Computer Use Behavior Across Multiple Levels of Abstraction", "notes": "Reviewed the complete abstract and selected v1 taxonomy, annotation, evaluation and limitations sections. MAP analyzes existing human/agent traces rather than deploying a new acting policy. Annotation and timing coverage depend on the source datasets; manual agreement reflects a particular labeling scheme. Behavioral findings do not establish a privacy guarantee. No independent reproduction.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAP annotates computer-use traces at linked operation, function/consequence, method and goal levels, then compares human and agent workflows through sequence, pattern and effort metrics.", "topics": ["evaluation", "observability"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26703"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["mobile GUI operations"], "counts": [{"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending", "population": "AppSim-Bench paper-described release", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2609.07712", "unit": "tasks", "value": 557}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Acrab-Agentic-Labs/AppSim/HEAD/README.md", "population": "published simulated app set", "qualifier": "exact", "unit": "applications", "value": 17}], "evaluation": "Controlled backend data and outcome-based task verification.", "evidence_urls": ["https://raw.githubusercontent.com/Acrab-Agentic-Labs/AppSim/HEAD/README.md", "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"], "focus": ["mobile task completion"], "notes": "Use the simulated apps and controlled backend state supplied by the benchmark. Desktop host instructions are not the platform being evaluated.", "observations": ["mobile interfaces"], "setting": "interactive"}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "12dc2afe339c1ed66e4a6a4ccf5ae13d27ce32f452960341529a4b9016f69059", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/Acrab-Agentic-Labs/AppSim/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "github:acrab-agentic-labs/appsim", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Acrab-Agentic-Labs/AppSim"}], "name": "AppSim-Bench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Acrab-Agentic-Labs/AppSim/HEAD/README.md", "target_id": "arxiv:2609.07712", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Simulated mobile applications and evaluation tooling for reproducible GUI-agent tasks with controlled application state.", "topics": ["evaluation", "sandbox"], "updated_at": null, "url": "https://github.com/Acrab-Agentic-Labs/AppSim"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["agent-specific Android actions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md", "population": "complete README task collection", "qualifier": "exact", "unit": "tasks", "value": 340}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md", "population": "single-app subset", "qualifier": "exact", "unit": "tasks", "value": 300}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md", "population": "cross-app subset", "qualifier": "exact", "unit": "tasks", "value": 40}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md", "population": "README total including third-party, Google, and system apps", "qualifier": "exact", "unit": "applications", "value": 66}], "evaluation": "A multimodal evaluator and annotated key-component matching assess completion; the framework also reports resource-use metrics.", "evidence_urls": ["https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md", "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/Documentation.md"], "focus": ["mobile task completion", "resource use", "cross-app workflows"], "notes": "The two task subsets partition the 340 tasks. The integrated agent and evaluator modes must accompany results. The preconfigured emulator snapshot is still marked coming soon in the inspected documentation.", "observations": ["screenshot trajectories"], "setting": "interactive"}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "5294a67728072a3658cb2073bdf9a084fe3c00d87a848923dff68adb35213072", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "5294a67728072a3658cb2073bdf9a084fe3c00d87a848923dff68adb35213072", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:23Z", "role": "primary", "sha256": "a55290d62f8823097c73d74bac4eb08853a3caff724382e60a6a9364e7ced987", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/Documentation.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:58Z", "role": "primary", "sha256": "6962bfc8ee16c6e985e4fb680b8fe46d5b34d30f7cb8a1e61fd4c55c93120ddf", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/LICENSE"}], "id": "github:ai-agents-2030/spa-bench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:58Z", "url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ai-agents-2030/SPA-Bench"}], "name": "SPA-Bench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ai-agents-2030/SPA-Bench/HEAD/README.md", "target_id": "arxiv:2410.15164", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "340 single-app and cross-app tasks over 66 Chinese and English Android apps, with plug-in support for 11 smartphone agents and automatic coarse-to-fine success detection.", "topics": ["evaluation", "mobile-navigation"], "updated_at": null, "url": "https://github.com/ai-agents-2030/SPA-Bench"}
{"aliases": ["AdvWeb"], "authors": [], "details": {"availability": "Public research source and workflow documentation; no standalone trained checkpoint was verified.", "repository_revision": "e01316e33d11ab145daa7cc7ec12d6ddb0bf38a0"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:18:18Z", "role": "primary", "sha256": "019d45dd4c383f919f3ac2bf84c672500d63751a49de2dd2bac9c018ffc99881", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/AI-secure/AdvAgent/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:18:17Z", "role": "primary", "sha256": "3ea44a202a677c0beb744ab05bd687da93a83365b7eb0b7999c3a4c22d7e2608", "supports": ["identity"], "url": "https://api.github.com/repos/AI-secure/AdvAgent"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:18:33Z", "role": "primary", "sha256": "95929797eb0717ce26b26ccbdc752bfda18fc0070975b32f09827b4e9fec6ecf", "supports": ["release_status", "licenses.code"], "url": "https://api.github.com/repos/AI-secure/AdvAgent/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:17:48Z", "role": "primary", "sha256": "3ea44a202a677c0beb744ab05bd687da93a83365b7eb0b7999c3a4c22d7e2608", "supports": ["identity"], "url": "https://api.github.com/repos/AI-secure/AdvWeb"}], "id": "github:ai-secure/advagent", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T09:18:18Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/AI-secure/AdvAgent"}], "name": "AdvAgent (formerly AdvWeb)", "notes": "The official GitHub API resolves AI-secure/AdvWeb to AI-secure/AdvAgent. Public data-generation, SFT/DPO training and evaluation scripts are present, with upstream data and external model APIs required. No repository license was found, and no permission for third-party data or model weights is inferred. The README citation block is empty, so no paper identifier is bound from title similarity alone.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Research code for constructing and evaluating adversarial inputs to web agents, with a black-box red-team pipeline using Mind2Web data and SeeAct.", "topics": ["safety-security", "robustness", "evaluation"], "updated_at": null, "url": "https://github.com/AI-secure/AdvAgent"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["app launch/stop", "click/double-click/long-click", "text input", "swipe/drag and Android navigation", "finish"], "counts": [], "evaluation": "The released evaluator compares trajectories with reference actions using longest-common-subsequence-based task reward and completion. Its reversed redundancy metric is the reference/predicted action-length ratio (with 1e-6 denominator smoothing), multiplied by the LLM success flag. LLM success judgments and instruction-constraint checks are separate metrics. Full reference-dependent evaluation coverage cannot be assumed because complete annotated action sequences are not released.", "evidence_urls": ["https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/README.md", "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/android_env2/README.md", "https://api.github.com/repos/AndroidArenaAgent/AndroidArena/git/trees/main?recursive=1", "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/android_env2/observation.py", "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/android_env2/actions.py", "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/run_evaluator.py"], "focus": ["multi-step Android tasks", "cross-app workflows and instruction constraints"], "notes": "README states that task instructions are released but annotated action sequences, apart from the calendar example, will come later. Environment setup targets Android API 33 with uiautomator2 and manual Google-account setup. Public task YAML and evaluator code exist, but benchmark execution and full reference coverage were not verified. No task total is derived from app YAML file counts. Code and data licenses remain unknown.", "observations": ["UI hierarchy text", "screenshots", "combined text/image observations"], "setting": "interactive"}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "206bd0dc5220cf754cffc97767cce6a3cd27660ef050a569366bef2331241c60", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "a1e7dc0c5facae030c2cf19aad6511996e7e1198d8d6619268d10c29a1b57bcf", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/android_env2/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:05Z", "role": "primary", "sha256": "edc03388b67143644db4a412e4c888f8c796068daff29f63615f0b6a3a04262a", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/AndroidArenaAgent/AndroidArena/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:24:52Z", "role": "primary", "sha256": "cfa9d272c85398f54c18d6e7445d680009e8c16d128962235d899592d3094c2f", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/android_env2/observation.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:24:53Z", "role": "primary", "sha256": "d25f2edf9202f59615341b62fc6046ac23a6f4cb072a6038ac562506cbdc891c", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/android_env2/actions.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "257c42e980a0b14bd74bfcc381a0ecbef2b538c0f1cc1a94d22925efe354fb8e", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/run_evaluator.py"}], "id": "github:androidarenaagent/androidarena", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/AndroidArenaAgent/AndroidArena"}], "name": "AndroidArena", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/AndroidArenaAgent/AndroidArena/HEAD/README.md", "target_id": "arxiv:2402.06596", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Android environment for evaluating language-model-driven agents.", "topics": ["evaluation", "mobile-navigation"], "updated_at": null, "url": "https://github.com/AndroidArenaAgent/AndroidArena"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["list of UI element boxes [x1,y1,x2,y2] and text descriptions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/README.md", "population": "English annotation subset reported by README", "qualifier": "exact", "unit": "page-annotation entries", "value": 1000}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/README.md", "population": "Chinese annotation subset reported by README", "qualifier": "exact", "unit": "page-annotation entries", "value": 1000}], "evaluation": "Metrics use greedy one-to-one matching in descending IoU order, then micro precision/recall/F1 and lowercased difflib.SequenceMatcher description similarity on matched pairs. Actual metrics.py accepts IoU >= 0.5 at its default threshold; README prose says > 0.5. Predictions align to annotations by image_path; callers must supply complete prediction coverage because missing predictions are not independently enumerated by evaluate_dataset.", "evidence_urls": ["https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/README.md", "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/src/metrics.py", "https://api.github.com/repos/ant-research/GUIHalluBench/git/trees/master?recursive=1", "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/LEGAL.md", "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/.gitattributes"], "focus": ["GUI page-element detection", "element-description consistency"], "notes": "Each annotation entry represents a page, not one UI element. README’s illustrative num_samples=4796 output is not adopted as a corpus count. Public tree contains both annotation JSON files and an images.tar.gz Git-LFS pointer; payload availability/integrity was not tested by downloading images. LEGAL.md only governs comment-language precedence, not a code/data license grant; both slots remain unknown. No benchmark run was performed.", "observations": ["page screenshot"], "setting": "offline"}}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:51Z", "role": "primary", "sha256": "51df9dbcf52f46e4774207f2e6ad84dc8a6d117cdab6f77a38e9897ded1d0963", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://github.com/ant-research/GUIHalluBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "d9c476d6960a4ef99e20d4c3b5d35a63f3ae5e61fc20e9dea00b11e64b74c9e9", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:09Z", "role": "primary", "sha256": "f36707dfcd3e83fc9d7f57b50d0c27007801e5029cc1ac255e000560ecf778fc", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/src/metrics.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:07Z", "role": "primary", "sha256": "c0896bc90d515b7bd8fe86b062aeb199bd47570e6eca8474986f5b1ef1d98dbd", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/ant-research/GUIHalluBench/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:09Z", "role": "primary", "sha256": "adba50d9794b9ef3f7ec8cbc680f7f1fa3fbf9df0ac8d1f9b9ccab6d941bc11b", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/LEGAL.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:34:56Z", "role": "primary", "sha256": "8f35fa75be42a7a53e4f96de1b62be84962d2cf1beb8b23dfcfaecfaea4a6b90", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ant-research/GUIHalluBench/HEAD/.gitattributes"}], "id": "github:ant-research/guihallubench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:40:51Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ant-research/GUIHalluBench"}], "name": "GUIHalluBench", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Measures hallucination in UI-element parsing by matching the boxes and descriptions a model lists for 1,000 English and 1,000 Chinese screenshots against human annotations.", "topics": ["evaluation", "robustness"], "updated_at": null, "url": "https://github.com/ant-research/GUIHalluBench"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md", "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/LICENSE"], "reason": "The retained README restricts SheetCopilot and its dataset to noncommercial use, while the retained repository LICENSE contains the GNU GPL version 3 text. These declarations do not establish one unambiguous code license."}}}, "environments": ["desktop", "windows", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:56Z", "role": "primary", "sha256": "beb8c811bf2d070845aa6fcc7e94ffd43ca4f9f60ce42c72d958598790fbdffc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "licenses.code", "notes"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:47Z", "role": "primary", "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:47Z", "role": "primary", "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/LICENSE"}], "id": "github:bravegroup/sheetcopilot", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-18T10:59:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/BraveGroup/SheetCopilot"}], "name": "SheetCopilot", "notes": "The original agent and the later Ubuntu evaluator have different roles; the repository does not establish that the original agent runs on Ubuntu. Code licensing is unresolved because the retained README and LICENSE contain conflicting declarations.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md", "target_id": "arxiv:2305.19308", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Spreadsheet automation agent for Windows/Excel, with benchmark assets and a separately released headless Ubuntu evaluation backend.", "topics": ["workflow", "tool-integration", "evaluation"], "updated_at": null, "url": "https://github.com/BraveGroup/SheetCopilot"}
{"aliases": [], "authors": [], "details": {}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "dc60f683b0f44ccee4e54e5cbe1552ee3454e1ec4ed5c04635fedb83a325f51f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/bz-lab/AUITestAgent/HEAD/README.md"}], "id": "github:bz-lab/auitestagent", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-16T10:43:11Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/bz-lab/AUITestAgent"}], "name": "AUITestAgent", "notes": "Its README describes mobile apps without naming an operating system.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/bz-lab/AUITestAgent/HEAD/README.md", "target_id": "arxiv:2407.09018", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Generates mobile GUI interactions from natural-language test requirements and checks application behavior against those requirements.", "topics": ["automation", "evaluation"], "updated_at": null, "url": "https://github.com/bz-lab/AUITestAgent"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["Python-defined environment actions"], "counts": [{"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50", "population": "CRAB Benchmark-v0 (paper)", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2407.01511", "unit": "tasks", "value": 120}], "evaluation": "Graph-based evaluators measure progress through task requirements.", "evidence_urls": ["https://raw.githubusercontent.com/camel-ai/crab/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"], "focus": ["cross-environment tasks"], "notes": "CRAB is also an extensible framework. The 120-task count belongs to Benchmark-v0, not every environment that can be created with the framework.", "observations": [], "setting": "interactive"}}, "environments": ["desktop", "android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "7b894c5474ba78294427cfd4ef0eea8cb22d76dfa0c2e8c7f01e31ad927064e9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/camel-ai/crab/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "github:camel-ai/crab", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/camel-ai/crab"}], "name": "CRAB", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/camel-ai/crab/HEAD/README.md", "target_id": "arxiv:2407.01511", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Cross-environment benchmark framework for multimodal agents.", "topics": ["evaluation", "cross-app"], "updated_at": null, "url": "https://github.com/camel-ai/crab"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["READ text answer", "ACT selector, text/label, accessible role/name, DOM index, Stagehand reference or screenshot point, according to the representation"], "counts": [], "evaluation": "READ uses normalized exact match against an answer or allowed alias. ACT resolves predicted identifiers to ground-truth elements, or checks screenshot points inside the target box. Representation-token accounting subtracts fixed-prompt tokens; screenshots use billed input usage minus the fixed prompt. These are answer/target correctness and token metrics, not completed-task success.", "evidence_urls": ["https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/README.md", "https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/METHODOLOGY.md", "https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/src/wpbench/scoring.py"], "focus": ["Browser-representation token use", "Factual READ accuracy", "ACT target-element selection"], "notes": "Fixed local-page observations; the reported run uses claude-sonnet-5 and a shared prompt, without scrolling or an agent loop. Browser Use disables its viewport cutoff and sends untruncated text; Stagehand uses a separate matched browser load. Short-string representability labels are approximate. Offline does not mean API-free. Third-party snapshot redistribution remains unresolved; results were not reproduced.", "observations": ["Top-document HTML", "Markdown converted from HTML", "Playwright accessibility snapshot", "Browser Use indexed DOM", "Stagehand hybrid snapshot", "Viewport screenshot"], "setting": "offline"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/LICENSE"], "reason": "The inspected root LICENSE contains complete MIT terms for repository software. The methodology explicitly leaves redistribution permission for captured third-party practice-site pages unresolved; code MIT is not a blanket snapshot/data license."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:51:30.874397Z", "role": "primary", "sha256": "14775f6e3daba69d4684380fd33bf717bdc3093b65be796eb863bbd27d0c0bab", "supports": ["name", "kind", "summary", "topics", "scope", "notes", "environments", "details.benchmark_protocol.focus", "details.benchmark_protocol.setting", "details.benchmark_protocol.observations", "details.benchmark_protocol.actions", "details.benchmark_protocol.notes"], "url": "https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/README.md"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:05:35.063004+00:00", "role": "primary", "sha256": "3e22e1cbfa47bcf98f6f70a5f6f162a31a509a9f2091a84679bb5231977e29d9", "supports": ["id", "url"], "url": "https://api.github.com/search/repositories?q=topic%3Aweb-agents+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-20&per_page=100&page=1&sort=updated&order=asc"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:50:54.545491Z", "role": "primary", "sha256": "fd30e94e78e7167a49579922259fde1c7d7f8259e9f00afba15faf3a71966a94", "supports": ["id", "url", "notes"], "url": "https://github.com/charlypoly/web-context-bench-python"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:52:30.131541Z", "role": "primary", "sha256": "1c3c0cbf9f9d9f7cd9e7a2f027ecb217fd5c16fa157916281a15bb7f403f6f1c", "supports": ["summary", "environments", "topics", "scope", "notes", "licenses.data", "details.benchmark_protocol.focus", "details.benchmark_protocol.setting", "details.benchmark_protocol.observations", "details.benchmark_protocol.actions", "details.benchmark_protocol.evaluation", "details.benchmark_protocol.notes"], "url": "https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/METHODOLOGY.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:53:28.567861Z", "role": "primary", "sha256": "8cada5d71b5a61e7aff2e1ac2b951c266eb2db7c082c06c5a13aa127d42b3fc5", "supports": ["summary", "topics", "scope", "notes", "details.benchmark_protocol.evaluation", "details.benchmark_protocol.notes"], "url": "https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/src/wpbench/scoring.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:51:41.499735Z", "role": "primary", "sha256": "9b64ef672f5091b6074d02502f6d3c161d94308a3d9d142d7437dcddf611bbc4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/LICENSE"}], "id": "github:charlypoly/web-context-bench-python", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-20T11:51:30.874397Z", "licenses": {"code": {"basis": "Complete MIT license text in the repository root at the inspected commit; repository code only.", "declared": "MIT", "retrieved_at": "2026-10-20T11:51:41.499735Z", "url": "https://raw.githubusercontent.com/charlypoly/web-context-bench-python/58d604bfaaac60ad42f02013f53f9fb301c1dcda/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/charlypoly/web-context-bench-python"}], "name": "Web Context Bench (Python)", "notes": "This is a single-snapshot representation evaluation on local fixtures and captured pages, not a live multi-step agent benchmark. Its shared-prompt, one-model setup differs from complete Browser Use or Stagehand agents. The methodology notes unresolved redistribution licensing for third-party practice-site snapshots; MIT is recorded for repository code only. Selected scoring code was read, but no models, browser tasks or claimed results were run or reproduced.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Local-page benchmark comparing six browser representations on token use, factual reading and target-element selection, with deterministic answer scoring.", "topics": ["evaluation", "grounding", "efficiency", "screen-understanding"], "updated_at": null, "url": "https://github.com/charlypoly/web-context-bench-python"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/LICENSE"], "reason": "The inspected root LICENSE is complete MIT text for repository code. The local data README says same as parent repository, but the linked Hub dataset/model cards and artifacts were not inspected; their licenses are not inferred."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:51:34.292633Z", "role": "primary", "sha256": "7cda5207d8882bc8dedaefb93d2ccf54b310354917283b8f4890fffdc5764958", "supports": ["name", "kind", "summary", "topics", "scope", "notes", "environments", "relations"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/README.md"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:04:53.470789+00:00", "role": "primary", "sha256": "3d76d934e98641746085772b98ee17059749f980a616435498a7ad483bd4cec1", "supports": ["id", "url"], "url": "https://api.github.com/search/repositories?q=topic%3Agui-grounding+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-20&per_page=100&page=1&sort=updated&order=asc"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:50:58.795155Z", "role": "primary", "sha256": "08b89c5e6143afa4c0cdde95f8ec016d9033ee004f7631fadd4457144c2b87c9", "supports": ["id", "url", "notes"], "url": "https://github.com/ChenghengLi/WinDOM"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:52:41.278660Z", "role": "primary", "sha256": "09488140d35408cc11bb9c12b1bb435b21814cf6525937a8309475c9326c4004", "supports": ["environments", "notes", "licenses.data"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/data/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:52:45.091369Z", "role": "primary", "sha256": "0ee33b1e416ea4c5d8834688e6f42fbd15ef493f1b82398d3bb985838cbdffb9", "supports": ["summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/src/windom/training/sfd_4b.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:52:48.511963Z", "role": "primary", "sha256": "c481b2f91f2d82bf801632da6ab6cce01bc42c780104d743156c272942d2cef3", "supports": ["summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/src/windom/evaluation/screenspot.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:51:44.940106Z", "role": "primary", "sha256": "d5a37ff86df9b231440ab1bcf3ecf77304a0a53e772182cdb622a3ab5d043447", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/LICENSE"}], "id": "github:chenghengli/windom", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-20T11:51:34.292633Z", "licenses": {"code": {"basis": "Complete MIT license text in the repository root at the inspected commit; repository code only.", "declared": "MIT", "retrieved_at": "2026-10-20T11:51:44.940106Z", "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ChenghengLi/WinDOM"}], "name": "WinDOM code", "notes": "The Windows-style environment is win11React in Playwright, not native Windows control. The repository documents its predecessor trajectory pipeline but explicitly omits the final corpus harvester. The inspected ScreenSpot evaluator defaults to a privileged ground-truth-center demonstration; --no-demo and --official-prompt select other modes, so evaluation results require their prompt and coordinate settings. Linked datasets, weights and reported scores were not independently verified; MIT applies to inspected repository code.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/README.md", "target_id": "arxiv:2606.25964", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "GUI-grounding training and evaluation code with same-family distillation, GRPO recipes and point-in-box scoring, using a Windows-style web environment.", "topics": ["grounding", "imitation-learning", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://github.com/ChenghengLi/WinDOM"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "93defc2b31d0f8e6144e77cda4a74abeafd5253ae18a626bfc625f429050770a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/chessbean1/GUI-Robust/HEAD/README.md"}], "id": "github:chessbean1/gui-robust", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:43:11Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/chessbean1/GUI-Robust"}], "name": "GUI-Robust", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/chessbean1/GUI-Robust/HEAD/README.md", "target_id": "arxiv:2506.14477", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation scripts and linked data for comparing GUI action predictions and task execution under ordinary and anomalous interface conditions.", "topics": ["evaluation", "robustness"], "updated_at": null, "url": "https://github.com/chessbean1/GUI-Robust"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "a2df62f9b4359f7839c7d6c6917ff654f9eadeafa92f96311495b71e37abd9ba", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/convergence-ai/webgames/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:00Z", "role": "primary", "sha256": "0e03b68c7acb0b340e47f9000ac3808950e2ac8c4cd05960437a9e46e3c09a62", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/convergence-ai/webgames/HEAD/LICENSE"}], "id": "github:convergence-ai/webgames", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:00Z", "url": "https://raw.githubusercontent.com/convergence-ai/webgames/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/convergence-ai/webgames"}], "name": "WebGames", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "150 client-side web challenges in five capability domains, built to be easy for humans but hard for browsing agents, each revealing a password when solved.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/convergence-ai/webgames"}
{"aliases": [], "authors": [], "details": {"availability": "Public inference/scoring source; benchmark data access remains gated.", "evaluation_caution": "Missing screenshots are marked AUTO_PASS and credited as correct according to the README.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/LICENSE"], "reason": "Repository code license only; no data-license or gated-access permission is inferred."}}, "repository_revision": "0c2da7002baa5b0ee71f80dd6fbd90e33071e259"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:21Z", "role": "primary", "sha256": "6681ce1bd98c856fed4193e239e234a69eb2a74fa937e9c38db89f48db40f452", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:20Z", "role": "primary", "sha256": "f2b7c762383746a56e62273f756ca7ed3223726535763a41fda3ac5e3467aa82", "supports": ["identity"], "url": "https://api.github.com/repos/CVC2233/AndroTMem"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:59Z", "role": "primary", "sha256": "302c804dc65ebb3e8558651e0c7cd0a79c1caf29fc9fb11868d26fba36afdf48", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/CVC2233/AndroTMem/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:00Z", "role": "primary", "sha256": "6d572ccef64746baac78441c0b26488f525c70c2b7453f985bd4f19cbec00634", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:07Z", "role": "primary", "sha256": "e82623b7dc889eb7bff45c33f15a47c8c5caac3de0a3c56a7a9975ca9b2d6694", "supports": ["notes", "release_status"], "url": "https://huggingface.co/api/datasets/CVC2233/AndroTMem-Bench"}], "id": "github:cvc2233/androtmem", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T11:18:21Z", "licenses": {"code": {"basis": "Repository code license only; no data-license or gated-access permission is inferred.", "declared": "MIT", "retrieved_at": "2026-10-17T11:23:00Z", "url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/CVC2233/AndroTMem"}], "name": "AndroTMem inference and evaluation code", "notes": "The source tree releases inference backends and evaluation scripts under evaluation/. The separately hosted benchmark is gated=auto and its raw card returned 401; no access request was made and it is not approved here as a directly accessible dataset. The README explicitly credits missing screenshots as AUTO_PASS/correct, so incomplete local image inventories can inflate scores. Its TCR is an offline anchor/dependency score and is not presented here as independently verified live task success.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/README.md", "target_id": "arxiv:2603.18429", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A public Android GUI evaluation framework with action-history, summary and anchored-memory prompt modes, plus step-action and dependency-aware task scoring.", "topics": ["memory", "long-horizon", "evaluation"], "updated_at": null, "url": "https://github.com/CVC2233/AndroTMem"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:34Z", "role": "primary", "sha256": "9c938f8f3f3ee04f459362a8b7d399588fd030ce476cda111e5f23fbf1e31f10", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/facebookresearch/ai-agent-privacy/HEAD/README.md"}], "id": "github:facebookresearch/ai-agent-privacy", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:34Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/facebookresearch/ai-agent-privacy"}], "name": "AgentDAM", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/facebookresearch/ai-agent-privacy/HEAD/README.md", "target_id": "arxiv:2503.09780", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Web-agent benchmark for measuring unnecessary use of sensitive personal information during end-to-end task execution.", "topics": ["safety-security", "evaluation"], "updated_at": null, "url": "https://github.com/facebookresearch/ai-agent-privacy"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["click", "type", "scroll", "configured BrowserGym/AgentLab actions"], "counts": [], "evaluation": "Tasks define check_if_task_is_complete(initial_state,current_state); app-specific reward logic evaluates underlying state, with custom checks for code-editor file contents. Observation/history flags are configurable. No universal single fixed task total or published-run reproduction is implied.", "evidence_urls": ["https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/README.md", "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/docs/tasks.md", "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/docs/agents.md", "https://api.github.com/repos/facebookresearch/OpenApps/git/trees/main?recursive=1"], "focus": ["UI-agent reliability under environment variation", "state-based browser task evaluation"], "notes": "Preserve the code/tag, app state, theme/layout/content choices, goal variant and action/observation configuration. Current main refactors the paper-era appearance surface, and the v1.0-paper tag includes post-publication fixes that may move rewards. The environments run as browser apps; phone/tablet viewports are not native Android/iOS environments.", "observations": ["screenshots", "accessibility trees", "set-of-marks and coordinates (configurable)"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/LICENSE"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate."}}, "repository_tree_sha": "6e6782794433c60468d65ac75bc3f03c56007f8c", "repository_url": "https://github.com/facebookresearch/OpenApps"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:18Z", "role": "primary", "sha256": "1df0f66584b37e7dba90811a7732d3b0c13c96cd7da4aec3970a18be36a6cec1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:19Z", "role": "primary", "sha256": "2c57b627cb344e360480c10e74dcdd1d1ed321d9a7a4e3891bc970355b15f4cd", "supports": ["identity"], "url": "https://api.github.com/repos/facebookresearch/OpenApps"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:18Z", "role": "primary", "sha256": "34aa76fbd6cca6ac2f3949792144662b90d85d7a18988afabcb570167ea5c1d6", "supports": ["identity", "release_status", "details.repository_tree_sha", "details.benchmark_protocol"], "url": "https://api.github.com/repos/facebookresearch/OpenApps/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:40:40Z", "role": "primary", "sha256": "af030b93a7604c063b08f003b0a0124a6e95e560649d0306bfe40dd87fed563d", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/docs/tasks.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:40:40Z", "role": "primary", "sha256": "aede4dfd965bd3484336f40ddb1bd6db48aed00ab45e27c048b76cc4601d71b3", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/docs/agents.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:20Z", "role": "primary", "sha256": "336255dc30193e8e15d689d9481bb05673d89055718f3a96923a7ffb99adbbaf", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/LICENSE"}], "id": "github:facebookresearch/openapps", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T04:35:18Z", "licenses": {"code": {"basis": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate.", "declared": "CC-BY-NC-4.0", "retrieved_at": "2026-10-18T04:35:20Z", "url": "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/facebookresearch/OpenApps"}], "name": "OpenApps", "notes": "Public code includes app/task/configuration and agent-launching implementations. README distinguishes moving main from v1.0-paper and explicitly says even that tag is not a byte-exact snapshot of the published runs. CC-BY-NC-4.0 is the repository’s declared code license; third-party WebShop, map and icon content retain separate terms. No commercial-use permission or fixed downloadable task total is inferred.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/facebookresearch/OpenApps/HEAD/README.md", "target_id": "arxiv:2511.20766", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Configurable browser apps and task rewards for testing computer-use agents under changes in content, layout and appearance.", "topics": ["evaluation", "robustness", "cross-app", "sandbox"], "updated_at": null, "url": "https://github.com/facebookresearch/OpenApps"}
{"aliases": ["STEVE", "STEVE-R1"], "authors": [], "details": {}, "environments": ["desktop", "windows", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T13:14:29Z", "role": "primary", "sha256": "52aa383fdf5d91951f5d9a73f4ebed4454a82b35edae296918116e9f0895dd83", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/FanbinLu/STEVE/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["relations", "environments"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "github:fanbinlu/steve-r1", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T13:14:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/FanbinLu/STEVE-R1"}], "name": "STEVE / STEVE-R1 research code", "notes": "The repository has moved to STEVE-R1. Its README reports an SFT release, with RL tuning in progress, and describes evaluations using multiple attempts per task. These results are not independently reproduced here.", "published_at": null, "relations": [{"evidence_url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending", "target_id": "arxiv:2503.12532", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Research code and linked models for step-verified computer-use training; the current repository describes STEVE-R1 and its trajectory releases.", "topics": ["data-generation", "automation", "evaluation"], "updated_at": null, "url": "https://github.com/FanbinLu/STEVE-R1"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["configurable coordinate actions", "element-reference actions", "keyboard input"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md", "population": "README collection size", "qualifier": "more-than", "unit": "web environments", "value": 100}], "evaluation": "", "evidence_urls": ["https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md", "https://miniwob.farama.org/content/observation_space/", "https://miniwob.farama.org/content/action_space/"], "focus": ["small web interaction tasks"], "notes": "The README gives a lower bound, not an exact active-environment inventory. Action presets expose different capabilities, so comparisons must identify the preset. Environment-specific reward details are not normalized here. The repository is in maintenance mode.", "observations": ["task instructions", "screenshots", "visible DOM-element attributes"], "setting": "interactive"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:54Z", "role": "primary", "sha256": "90499776fa1b23e08ac85190ad3be709caa9172ff4f2a0a106b0fa350568e86f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:54Z", "role": "primary", "sha256": "90499776fa1b23e08ac85190ad3be709caa9172ff4f2a0a106b0fa350568e86f", "supports": ["details.benchmark_protocol", "relations", "notes"], "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:23Z", "role": "primary", "sha256": "511039e0e3c26b117e7e0a001746c26ec6d94a8ab8c467f0be0bc1a0b01e870a", "supports": ["details.benchmark_protocol"], "url": "https://miniwob.farama.org/content/observation_space/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:23Z", "role": "primary", "sha256": "ae7bf6a3b98b198b985d7ca7b41cb2bdab3cf48e6445467ad9a8d4a2c8df240f", "supports": ["details.benchmark_protocol"], "url": "https://miniwob.farama.org/content/action_space/"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:07Z", "role": "primary", "sha256": "fd6529fb8f648d4130fd969e3d7fc740c9992a0bd94a1998f52555a873311511", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/LICENSE"}], "id": "github:farama-foundation/miniwob-plusplus", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:40:54Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:07Z", "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Farama-Foundation/miniwob-plusplus"}], "name": "MiniWoB++", "notes": "The maintainer README identifies MiniWoB++ as the extension introduced in the 2018 workflow-guided exploration paper. It separately links World of Bits as the source for the original MiniWoB predecessor. The current Farama interface, environment inventory, and maintenance status belong to the retained repository snapshot.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md", "target_id": "arxiv:1802.08802", "type": "introduced_by"}], "review_status": "curated", "scope": "core", "summary": "Over 100 web-interaction environments extending OpenAI's MiniWoB, controlled from Python through a Gymnasium API that drives the browser with Selenium.", "topics": ["evaluation", "reinforcement-learning"], "updated_at": null, "url": "https://github.com/Farama-Foundation/miniwob-plusplus"}
{"aliases": [], "authors": [], "details": {"availability": "Public evaluation harness and ungated web-application archive manifest.", "benchmark_protocol": {"actions": ["browser interaction", "test checklist generation", "defect reports"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/friedrichor/WebTestBench/HEAD/README.md", "population": "publisher benchmark overview", "qualifier": "exact", "unit": "application categories", "value": 7}, {"basis": "computed", "evidence_url": "https://huggingface.co/api/datasets/friedrichor/WebTestBench", "population": "Hub file manifest", "qualifier": "exact", "unit": "web-application archives", "value": 100}], "evaluation": "The supplied WebTester pipeline separates checklist generation from interactive defect detection and provides a scoring workflow.", "evidence_urls": ["https://raw.githubusercontent.com/friedrichor/WebTestBench/HEAD/README.md", "https://huggingface.co/api/datasets/friedrichor/WebTestBench", "https://huggingface.co/datasets/friedrichor/WebTestBench/raw/main/README.md"], "focus": ["web functionality testing", "latent permission and business constraints"], "notes": "The 100 archive files are not counted as 100 independent test tasks or defects. Backend/tool versions and deployment settings affect reproducibility.", "observations": ["web application interface", "development instruction"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/friedrichor/WebTestBench/HEAD/LICENSE"], "reason": "Repository source LICENSE; external tools and services retain their own licenses."}, "data": {"evidence_urls": ["https://huggingface.co/datasets/friedrichor/WebTestBench/raw/main/README.md"], "reason": "Publisher Hub-card license declaration for the released benchmark dataset."}}, "repository_revision": "04873065bdb0938ed53d809559d4ab473c7ccd31"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:18Z", "role": "primary", "sha256": "b41472026cd7d4f81a01d58bf39193dc1fabc7875c61e370090a20cc77500cdd", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/friedrichor/WebTestBench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:18Z", "role": "primary", "sha256": "7203ecc2aae5f1697eeeb804ca8f21d47b9f22a8b8fa43a4064c4e1c49f5e72c", "supports": ["identity"], "url": "https://api.github.com/repos/friedrichor/WebTestBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:55Z", "role": "primary", "sha256": "bd1f09149b00d59a0078d414216bd2fb78e09ce7212d15bbda84ddec7052c8af", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/friedrichor/WebTestBench/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:08Z", "role": "primary", "sha256": "db1e97f40107038ae03835dc20f17b13b37e74b5c8587f8c39b441271618f017", "supports": ["identity", "release_status", "details.benchmark_protocol"], "url": "https://huggingface.co/api/datasets/friedrichor/WebTestBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:08Z", "role": "primary", "sha256": "6f2f94a25ed66a14779f6b4e6686044d0510eb32a0187645cc473978d93f7c5b", "supports": ["licenses.data", "details.benchmark_protocol"], "url": "https://huggingface.co/datasets/friedrichor/WebTestBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:55Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/friedrichor/WebTestBench/HEAD/LICENSE"}], "id": "github:friedrichor/webtestbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T11:18:18Z", "licenses": {"code": {"basis": "Repository source LICENSE; external tools and services retain their own licenses.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T11:22:55Z", "url": "https://raw.githubusercontent.com/friedrichor/WebTestBench/HEAD/LICENSE"}, "data": {"basis": "Publisher Hub-card license declaration for the released benchmark dataset.", "declared": "apache-2.0", "retrieved_at": "2026-10-17T11:23:08Z", "url": "https://huggingface.co/datasets/friedrichor/WebTestBench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/friedrichor/WebTestBench"}], "name": "WebTestBench", "notes": "Public evaluation/deployment scripts accompany an ungated Hub release with JSON/JSONL metadata and 100 application ZIP files. The ZIP count is an artifact inventory, not an asserted test-case or defect count. The documented baseline uses Claude Agent SDK and a pinned Playwright MCP version with external model credentials; these were not installed or run. Code and data Apache-2.0 declarations are retained separately.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/friedrichor/WebTestBench/HEAD/README.md", "target_id": "arxiv:2603.25226", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "An end-to-end web-testing benchmark where agents derive test checklists from development instructions, interact with deployed applications and report functional or constraint-related defects.", "topics": ["evaluation", "web-navigation", "tool-integration"], "updated_at": null, "url": "https://github.com/friedrichor/WebTestBench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["PyAutoGUI"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/README.md", "population": "full WindowsAgentArena-V2 suite reported by README", "qualifier": "exact", "unit": "tasks", "value": 141}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/README.md", "population": "full suite coverage reported by README", "qualifier": "exact", "unit": "Windows applications", "value": 11}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/src/win-arena-container/client/evaluation_examples_windows/test_custom.json", "population": "shipped test_custom.json selection", "qualifier": "exact", "unit": "task IDs", "value": 20}], "evaluation": "Runner exposes four observation choices and uses PyAutoGUI. count.py sums the last score in each results.json and divides by the number of traj.jsonl files, multiplying by 100; it separately reports result/trajectory/error-file coverage. This denominator is not automatically the full benchmark size.", "evidence_urls": ["https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/README.md", "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/src/win-arena-container/client/run.py", "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/src/win-arena-container/client/evaluation_examples_windows/test_custom.json", "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/count.py", "https://api.github.com/repos/GAIR-NLP/WindowsAgentArena-V2/git/trees/main?recursive=1"], "focus": ["Windows desktop task execution", "revised task evaluators"], "notes": "README sample config uses max_steps=30. The shipped selected-task index has 20 IDs, not 141; preserve the selected task list, trial ID, score coverage and VM snapshot when comparing runs. Docker/Windows-11 setup is required. No VM or benchmark was executed.", "observations": ["screenshot", "accessibility tree", "screenshot plus accessibility tree", "set-of-marks"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/LICENSE"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate."}}, "repository_tree_sha": "2927fe55005d1be75d5a9188c0045f73ba28d192", "repository_url": "https://github.com/GAIR-NLP/WindowsAgentArena-V2"}, "environments": ["desktop", "windows"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:17Z", "role": "primary", "sha256": "1dd744ce565ec00b20ea5637eb7bc6112e97fb9044bd9a471f7cc4b7227ca344", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:20Z", "role": "primary", "sha256": "0b0af02d3c2c4916dd786b868b99d60b4681927d0273a23661eb2b6db7cf9404", "supports": ["identity"], "url": "https://api.github.com/repos/GAIR-NLP/WindowsAgentArena-V2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:16Z", "role": "primary", "sha256": "638ef9ee593a3a560c1d0b3fb2bc4a9d31a3e904c0111971924a6d66e1763aa6", "supports": ["identity", "release_status", "details.repository_tree_sha", "details.benchmark_protocol"], "url": "https://api.github.com/repos/GAIR-NLP/WindowsAgentArena-V2/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:40:39Z", "role": "primary", "sha256": "494de9e74629ac2c55358629c0d9ba13703f36317057a3089388fbb6e80ca28f", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/src/win-arena-container/client/run.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:40:39Z", "role": "primary", "sha256": "e45ba7fdc7555f46ad1b4cfaab79d00754909d053fc49d5bab651b0fb593439b", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/src/win-arena-container/client/evaluation_examples_windows/test_custom.json"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:40:39Z", "role": "primary", "sha256": "f12147ac78c39c72eb75f9c37b293ba7fff3209fc5758fd51c69abbd96d80a97", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/count.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:20Z", "role": "primary", "sha256": "2a2d1e46940b570d60f5e2198ed27fcec6316fc8489e9660da8777dbd75aaeec", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/LICENSE"}], "id": "github:gair-nlp/windowsagentarena-v2", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T04:35:17Z", "licenses": {"code": {"basis": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate.", "declared": "MIT", "retrieved_at": "2026-10-18T04:35:20Z", "url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/GAIR-NLP/WindowsAgentArena-V2"}], "name": "WindowsAgentArena-V2", "notes": "Repository contains executable evaluation code and task definitions. README links a separate Windows 11 VM snapshot; that image and its software licenses/access were not validated by downloading it. MIT is recorded for repository code only. Default task selection and the reported full suite have different scopes.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/GAIR-NLP/WindowsAgentArena-V2/HEAD/README.md", "target_id": "arxiv:2505.13909", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Windows benchmark revision with task/evaluator fixes and VM-initialization support, documented alongside PC-Agent-E.", "topics": ["evaluation", "cross-app", "robustness"], "updated_at": null, "url": "https://github.com/GAIR-NLP/WindowsAgentArena-V2"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/ghadinehme/VideoCAD/2d69eeaf0ac44210c76dcab2d08b4d4610be236c/LICENSE"], "reason": "MIT declaration in the author repository LICENSE; separate data, weights and document scopes are not inferred."}}, "repository_id": 1079441024, "repository_revision": "2d69eeaf0ac44210c76dcab2d08b4d4610be236c"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T12:02:53.536297+00:00", "role": "primary", "sha256": "cbb08803922b4b7736979ca0a5fe552b111fcd428fa9317401bc090b14bb4e54", "supports": ["name", "kind", "summary", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/ghadinehme/VideoCAD/2d69eeaf0ac44210c76dcab2d08b4d4610be236c/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T12:02:30.768857+00:00", "role": "primary", "sha256": "ff80afd35c79d8dd27dd7b9284ce082fe238c13a113aa19400ca0dc2d18ff194", "supports": ["identity", "details.repository_id"], "url": "https://api.github.com/repos/ghadinehme/VideoCAD"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T12:02:52.564274+00:00", "role": "primary", "sha256": "9ded7c6ef25b88879e9d46939ca7ff3bef97362e76dbdb808e57c67008541557", "supports": ["release_status", "details.repository_revision", "notes"], "url": "https://api.github.com/repos/ghadinehme/VideoCAD/git/trees/2d69eeaf0ac44210c76dcab2d08b4d4610be236c?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T12:02:34.311949+00:00", "role": "primary", "sha256": "a303d95a811331cd6ee23b0340f1b9ef2dd996835bfbc66656a871ff94eead50", "supports": ["environments", "summary", "relations", "notes"], "url": "https://arxiv.org/html/2505.24838v2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T12:03:24.765415+00:00", "role": "primary", "sha256": "8e02753a1016dc67c97b395934313027b5f91b74209c7a24438de70fdc19cad1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ghadinehme/VideoCAD/2d69eeaf0ac44210c76dcab2d08b4d4610be236c/LICENSE"}], "id": "github:ghadinehme/videocad", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-18T12:02:53.536297+00:00", "licenses": {"code": {"basis": "The author repository at the inspected commit contains the standard MIT license. This declaration does not establish terms for externally hosted data or weights.", "declared": "MIT", "retrieved_at": "2026-10-18T12:03:24.765415+00:00", "url": "https://raw.githubusercontent.com/ghadinehme/VideoCAD/2d69eeaf0ac44210c76dcab2d08b4d4610be236c/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ghadinehme/VideoCAD"}], "name": "VideoCAD training and evaluation code", "notes": "The paper links this author repository. Training, preprocessing, model and evaluation source files are released; runtime behavior has not been independently reproduced. Onshape is a browser-based CAD environment; supported operating systems are not inferred. The full dataset and a checkpoint are linked separately, and their payloads were not downloaded for this review. MIT is recorded only for this repository code; external data and weight licenses remain unknown.", "published_at": null, "relations": [{"evidence_url": "https://arxiv.org/html/2505.24838v2", "target_id": "arxiv:2505.24838", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "VideoCADFormer training, action-log preprocessing and evaluation code for learning CAD interface action sequences from screen recordings of browser-based Onshape.", "topics": ["grounding", "imitation-learning", "evaluation"], "updated_at": null, "url": "https://github.com/ghadinehme/VideoCAD"}
{"aliases": [], "authors": [], "details": {"availability": "Public archived TFRecord annotation files for both downstream tasks.", "dataset_profile": {"counts": [], "evidence_urls": ["https://raw.githubusercontent.com/google-research-datasets/uibert/HEAD/README.md"], "formats": ["TFRecord"], "modalities": ["text", "bounding boxes", "UI hierarchy metadata", "Rico image identifiers"], "notes": "Original UI images are retrieved separately by Rico IDs. No release-specific sample count is established.", "splits": ["AppSim train/dev/test", "RefExp train/dev/test"], "supervision": ["similar UI element retrieval", "referring-expression component retrieval"]}, "license_review": {"data": {"evidence_urls": ["https://raw.githubusercontent.com/google-research-datasets/uibert/HEAD/README.md"], "reason": "The README explicitly says the datasets are licensed under CC BY 4.0; original Rico materials keep their source terms."}}, "repository_revision": "eceafbde7567959899bfb7f8d9e6fd71a52aaa12"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "45d1ca0f5c727cf58b2671dc5579f0628f4d78e07b977a5d8b018295c6fe1f58", "supports": ["name", "kind", "summary", "topics", "scope", "notes", "relations", "details.dataset_profile", "licenses.data"], "url": "https://raw.githubusercontent.com/google-research-datasets/uibert/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "24a7838d425bdae363a7d5872702996b22068c1f0e47be274fdfbc4b0e20144f", "supports": ["identity"], "url": "https://api.github.com/repos/google-research-datasets/uibert"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:05Z", "role": "primary", "sha256": "9a6fc19c6b407309943fe645f4addb3d032f153856976c4f8d7a8cfb6d2ffe6b", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/google-research-datasets/uibert/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:36:55Z", "role": "primary", "sha256": "f752ebfb2598b7f848d343f428502b28e0f591f27849ab707f14bee398d4eb8d", "supports": ["environments", "notes"], "url": "https://interactionmining.org/rico"}], "id": "github:google-research-datasets/uibert", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T12:30:54Z", "licenses": {"code": null, "data": {"basis": "The README explicitly says the datasets are licensed under CC BY 4.0; original Rico materials keep their source terms.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-17T12:30:54Z", "url": "https://raw.githubusercontent.com/google-research-datasets/uibert/HEAD/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/google-research-datasets/uibert"}], "name": "UIBert AppSim and RefExp annotations", "notes": "The archived repository retains TFRecord train/dev/test files for AppSim and RefExp. These are annotation and UI-structure records with Rico image identifiers, not a republished complete screenshot collection or a UIBert checkpoint. The README’s 72k figure describes the upstream Rico source, not the number of released AppSim/RefExp examples; no sample total is inferred. The publisher explicitly assigns CC-BY-4.0 to these datasets.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/google-research-datasets/uibert/HEAD/README.md", "target_id": "url:0b134b082a7bd507fc6a", "type": "uses"}, {"evidence_url": "https://raw.githubusercontent.com/google-research-datasets/uibert/HEAD/README.md", "target_id": "arxiv:2107.13731", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Two Rico-derived UI component retrieval datasets: AppSim pairs semantically similar elements across screens, while RefExp links referring expressions to target elements.", "topics": ["grounding", "screen-understanding", "evaluation"], "updated_at": null, "url": "https://github.com/google-research-datasets/uibert"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/google-research-datasets/uicrit/HEAD/README.md", "population": "current three-annotator release", "qualifier": "exact", "unit": "mobile UI screens", "value": 1000}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/google-research-datasets/uicrit/HEAD/README.md", "population": "current three-annotator release", "qualifier": "exact", "unit": "design critiques", "value": 11344}], "evidence_urls": ["https://raw.githubusercontent.com/google-research-datasets/uicrit/HEAD/README.md"], "formats": ["CSV"], "modalities": ["text", "bounding boxes", "quality ratings"], "notes": "Publisher-reported current release, separate from the smaller original-paper population. Screenshot pixels are obtained through referenced RICO IDs.", "splits": [], "supervision": ["design critique", "region localization", "quality rating"]}}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:45Z", "role": "primary", "sha256": "345ff7003120f146016741099dd57b9b1a9ffc54ce1fc90f4d16d61e31d379e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/google-research-datasets/uicrit/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:45Z", "role": "primary", "sha256": "5579bbbc622a2f9fa1f77b509c1ece17af023b8c14b53aa48dcc41288a5852f2", "supports": ["identity"], "url": "https://api.github.com/repos/google-research-datasets/uicrit"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:45Z", "role": "primary", "sha256": "8e790a53de8c40aa14f9e49aebf42b077e6a172a6841f38c7c6851af57b0617a", "supports": ["release_status"], "url": "https://api.github.com/repos/google-research-datasets/uicrit/contents"}], "id": "github:google-research-datasets/uicrit", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:02:45Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/google-research-datasets/uicrit"}], "name": "UICrit", "notes": "The published CSV contains critique text, normalized boxes, ratings and RICO IDs; screenshots are referenced through RICO. The README explicitly licenses the work CC-BY-4.0. Its current 1,000-screen, 11,344-critique release differs from the original paper population. The README cites an ACM DOI rather than an arXiv ID, so no paper relation is asserted here. Its README establishes mobile interfaces without naming an operating system.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Mobile-interface critiques and quality ratings with region coordinates and RICO screen references, including a larger three-annotator release than the original paper dataset.", "topics": ["screen-understanding", "ui-design", "evaluation"], "updated_at": null, "url": "https://github.com/google-research-datasets/uicrit"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["Android device interactions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/google-research/android_world/HEAD/README.md", "population": "parameterized AndroidWorld tasks", "qualifier": "exact", "unit": "task definitions", "value": 116}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/google-research/android_world/HEAD/README.md", "population": "documented AndroidWorld app set", "qualifier": "exact", "unit": "applications", "value": 20}], "evaluation": "Programmatic task-specific initialization, system-state success checks, and teardown.", "evidence_urls": ["https://raw.githubusercontent.com/google-research/android_world/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"], "focus": ["mobile task completion"], "notes": "Random parameters produce many instances from the same task definitions. Preserve seeds and parameterization; 116 is not a count of all possible task instances.", "observations": ["screenshots", "accessibility information"], "setting": "interactive"}, "evaluation_type": "programmatic state checks", "leaderboard_url": "https://docs.google.com/spreadsheets/d/1cchzP9dlTZ3WXQTfYNhh3avxoLipqHN75v1Tb86uhHo/edit?gid=0#gid=0", "metrics": ["task success"], "task_count": 116, "task_unit": "parameterized task definitions"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "93201de5e21f2b12283a75b1e2c357781b2e6cc27eab20cc7dd7aa0f5919d839", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.task_count", "details.task_unit", "details.metrics", "details.evaluation_type", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/google-research/android_world/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "b09fcaf4fbc265223c2479c410beef2b98c5e687c78ecedf3452aa12ffc465f1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/google-research/android_world/HEAD/LICENSE"}], "id": "github:google-research/android_world", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/google-research/android_world/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/google-research/android_world"}], "name": "AndroidWorld", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/google-research/android_world/HEAD/README.md", "target_id": "arxiv:2405.14573", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Android environment with programmatic task evaluation for mobile agents.", "topics": ["evaluation", "mobile-navigation"], "updated_at": null, "url": "https://github.com/google-research/android_world"}
{"aliases": [], "authors": [], "details": {}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "63349e7557bd99240d4628de3994a53b69e54e3bd809895be11e9879bfba40cb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "f872ff2b1a064f46a45eee86f2cb2562fc0513285b80036c6cf7df9a5ffc6c70", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/LICENSE"}], "id": "github:gydpku/pptc", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T10:48:41Z", "url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/gydpku/PPTC"}], "name": "PPTC", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/README.md", "target_id": "arxiv:2311.01767", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "PowerPoint creation and editing benchmark with API-based execution and checks of the resulting presentation files.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://github.com/gydpku/PPTC"}
{"aliases": [], "authors": [], "details": {"availability": "Public benchmark code and public Zenodo dataset release metadata.", "benchmark_protocol": {"actions": ["generate procedures", "segment demonstrations", "answer workflow questions", "validate and rank procedures"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/HazyResearch/wonderbread/HEAD/README.md", "population": "WONDERBREAD BPM benchmark", "qualifier": "exact", "unit": "task types", "value": 6}, {"basis": "reported", "evidence_url": "https://zenodo.org/api/records/14094162", "population": "full dataset release 14094162", "qualifier": "exact", "unit": "human demonstrations", "value": 2928}, {"basis": "reported", "evidence_url": "https://zenodo.org/api/records/14094162", "population": "full dataset release 14094162", "qualifier": "exact", "unit": "web navigation workflows", "value": 598}, {"basis": "reported", "evidence_url": "https://zenodo.org/api/records/14094162", "population": "gold subset used for the original paper evaluations", "qualifier": "exact", "unit": "human demonstrations", "value": 724}], "evaluation": "Task-specific scripts evaluate six workflow-understanding task types; model outputs are not treated as successful interactive execution.", "evidence_urls": ["https://raw.githubusercontent.com/HazyResearch/wonderbread/HEAD/README.md", "https://zenodo.org/api/records/14094162"], "focus": ["workflow documentation", "workflow quality assessment", "workflow knowledge transfer"], "notes": "Counts are publisher-reported. The gold subset covers 162 workflows and differs from the full collection; scores require the exact task and subset.", "observations": ["screen recordings", "action traces", "screenshots", "written standard operating procedures"], "setting": "offline"}, "dataset_release": "10.5281/zenodo.14094162", "license_review": {"data": {"evidence_urls": ["https://zenodo.org/api/records/14094162"], "reason": "The linked Zenodo dataset release declares CC-BY-4.0; this does not assign a code license to the repository."}}, "repository_revision": "ed052c67aeada04167cdfe92ff8de454aa94627a"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:11Z", "role": "primary", "sha256": "1e4308d59a22b39543c28b745a58d3d7c47be1ab9df9a4c5c0fe868a810350b2", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/HazyResearch/wonderbread/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:11Z", "role": "primary", "sha256": "5ba27e24455b36b52bf65b3c9f589e760f97a576bb01d622d89836c5bb53b38c", "supports": ["identity"], "url": "https://api.github.com/repos/HazyResearch/wonderbread"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:04Z", "role": "primary", "sha256": "accb518a5362948cf96e9f0b684fb6fcc9b26e43a691016ae70a3283c0e1e2c2", "supports": ["release_status", "licenses.code"], "url": "https://api.github.com/repos/HazyResearch/wonderbread/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:15:34Z", "role": "primary", "sha256": "68b4b952c1ab90484d457b392de1d0c95a24f6cc1674c82125a6bd26f1cb1016", "supports": ["release_status", "licenses.data", "details.benchmark_protocol"], "url": "https://zenodo.org/api/records/14094162"}], "id": "github:hazyresearch/wonderbread", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T09:12:11Z", "licenses": {"code": null, "data": {"basis": "The linked Zenodo dataset release declares CC-BY-4.0; this does not assign a code license to the repository.", "declared": "cc-by-4.0", "retrieved_at": "2026-10-17T09:15:34Z", "url": "https://zenodo.org/api/records/14094162"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/HazyResearch/wonderbread"}], "name": "WONDERBREAD", "notes": "The repository contains evaluation and collection code; the linked Zenodo release exposes the full, gold and debug demonstration archives and metadata. This measures workflow understanding rather than interactive agent task completion. Zenodo declares CC-BY-4.0 for that dataset; no repository-level code license was found. Archives and human demonstration records were not downloaded.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/HazyResearch/wonderbread/HEAD/README.md", "target_id": "arxiv:2406.13264", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "An offline benchmark of workflow documentation, quality assessment and knowledge transfer using human web-navigation demonstrations, action traces and written procedures.", "topics": ["evaluation", "workflow", "video", "screen-understanding"], "updated_at": null, "url": "https://github.com/HazyResearch/wonderbread"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "68cb0deb8a2eb49fb9236ae3d885b043d240dded0fce31cf7425130a56ff5819", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/HeimingX/TAG/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "a33bee17685d64d23fb4d65b30b9c52f3a6dd5b21789d7b3509eb2e15145de6f", "supports": ["identity"], "url": "https://api.github.com/repos/HeimingX/TAG"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "fac4f5e38011bbcfef916063d495c21d49830759af4e8713a60d7bae992d2710", "supports": ["release_status"], "url": "https://api.github.com/repos/HeimingX/TAG/git/trees/main?recursive=1"}], "id": "github:heimingx/tag", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T02:02:06Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/HeimingX/TAG"}], "name": "TAG attention-driven GUI grounding", "notes": "The release contains grounding and benchmark evaluation code. The inspected root repository metadata does not declare a code license. It uses external pretrained models rather than publishing a new trained checkpoint, and the README restricts redistribution of unzipped benchmark data.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/HeimingX/TAG/HEAD/README.md", "target_id": "arxiv:2412.10840", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Grounding and evaluation code that aggregates a pretrained vision-language model’s attention maps to localize requested interface elements without fine-tuning.", "topics": ["grounding", "evaluation", "training-free"], "updated_at": null, "url": "https://github.com/HeimingX/TAG"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["PyAutoGUI", "UiPath Computer_13 action interface"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/HITsz-TMG/WindowsWorld/HEAD/README.md", "population": "WindowsWorld benchmark described in the retained README", "qualifier": "exact", "unit": "tasks", "value": 181}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/HITsz-TMG/WindowsWorld/HEAD/README.md", "population": "WindowsWorld coverage described in the retained README", "qualifier": "exact", "unit": "desktop applications", "value": 17}], "evaluation": "Publisher distinguishes S_int (intermediate checkpoint score on L1–L3) and S_final (final completion on L1–L4). Documented step budgets are 15/25/40/20 for L1/L2/L3/L4. Preserve level, observation mode, agent interface and checkpoint/final metric when reporting a run.", "evidence_urls": ["https://raw.githubusercontent.com/HITsz-TMG/WindowsWorld/HEAD/README.md", "https://api.github.com/repos/HITsz-TMG/WindowsWorld/git/trees/main?recursive=1"], "focus": ["multi-application Windows workflows", "intermediate-state and final-goal completion"], "notes": "Interactive VMware/Windows environment. The public repository includes benchmark.json, task/evaluator code and monitoring support; the README refers VM import to Installation Guide.md, whose distribution links and payloads were not inspected for this profile. Checkpoint counts and application counts are not extra tasks. Apache-2.0 is recorded for repository code; Windows, apps and VM artifacts retain separate terms.", "observations": ["screenshot", "accessibility tree", "screenshot plus accessibility tree", "set-of-marks"], "setting": "interactive"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "9857bd8ae9f521f152761ae3bd85debe4e7471aad34c9eb47618332948bca6c9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/HITsz-TMG/WindowsWorld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:05Z", "role": "primary", "sha256": "490c9257d56f0a5214e6424e02490379b0b00ca2f7d4626fd5d14e17239d05ed", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/HITsz-TMG/WindowsWorld/git/trees/main?recursive=1"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "a8d1d88a0f9cce3cdbff73b03a80dc5e5d69e3ba2443946c268815a550c9e0ec", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/HITsz-TMG/WindowsWorld/HEAD/LICENSE"}], "id": "github:hitsz-tmg/windowsworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/HITsz-TMG/WindowsWorld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/HITsz-TMG/WindowsWorld"}], "name": "WindowsWorld", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/HITsz-TMG/WindowsWorld/HEAD/README.md", "target_id": "arxiv:2604.27776", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Cross-application Windows benchmark with intermediate task evaluation.", "topics": ["evaluation", "cross-app"], "updated_at": null, "url": "https://github.com/HITsz-TMG/WindowsWorld"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "3ef331612819a006d4279ac59203971a22a56fb6d9a0047f85a82668d09c327b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/hud-evals/hud-python/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c88cc185e5547fce850b56f3b4072bed161658dc7277707e509437b9d760d470", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/hud-evals/hud-python/HEAD/LICENSE"}], "id": "github:hud-evals/hud-python", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/hud-evals/hud-python/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/hud-evals/hud-python"}], "name": "HUD", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "SDK for RL environments and evals spanning coding, browser, computer-use and robotics agents, whose environments expose SSH, MCP, CDP or VNC capabilities behind a minimal start-and-grade protocol.", "topics": ["evaluation", "reinforcement-learning"], "updated_at": null, "url": "https://github.com/hud-evals/hud-python"}
{"aliases": [], "authors": [], "details": {"availability": "Public benchmark-adapter source and running instructions.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/huggingface/screensuite/HEAD/LICENSE"], "reason": "Inspected repository LICENSE; not a license for linked benchmark datasets, pretrained weights or services."}}, "repository_revision": "9ceda5070d5664dd9376d5f8e7b43f02188e01ff"}, "environments": ["web", "android", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:13Z", "role": "primary", "sha256": "edcaa6f8355be6a8774ebeeb9bdb67648e8eeddb813d6be0fa2433116d9def36", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/huggingface/screensuite/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:12Z", "role": "primary", "sha256": "e372458e48d0ebbbe0f7887b66e2d5bb4db612fd96db4dc53404c0319891f7e8", "supports": ["identity"], "url": "https://api.github.com/repos/huggingface/screensuite"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:05Z", "role": "primary", "sha256": "93b685a13bfba4d9569dbfc0418de93c7b9f8a3f7c101af053aadeebbdf8b2c0", "supports": ["release_status"], "url": "https://api.github.com/repos/huggingface/screensuite/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:06Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/huggingface/screensuite/HEAD/LICENSE"}], "id": "github:huggingface/screensuite", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T09:12:13Z", "licenses": {"code": {"basis": "Inspected repository LICENSE; not a license for linked benchmark datasets, pretrained weights or services.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T09:14:06Z", "url": "https://raw.githubusercontent.com/huggingface/screensuite/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/huggingface/screensuite"}], "name": "ScreenSuite", "notes": "The publisher states that the suite compares the underlying multimodal models rather than full agent implementations. Public source includes benchmark adapters and registry code. Multi-step environments require Docker and KVM; constituent datasets, remote model services and OSWorld account-dependent tasks retain their own setup and terms. Apache-2.0 is scoped to this repository code.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A shared evaluation harness for GUI-model perception, offline action prediction and multi-step interaction, using simple smolagents-based agents across existing benchmarks.", "topics": ["evaluation", "grounding", "action-prediction"], "updated_at": null, "url": "https://github.com/huggingface/screensuite"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://github.com/iMeanAI/WebCanvas/blob/main/Mind2web-live_Leaderboard.md"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "5b2b4b4ce1241db7fd776f2ba0856ec2e0e441a56fa6461234abe0c6516f643e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/iMeanAI/WebCanvas/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:00Z", "role": "primary", "sha256": "cbaab6bcbc9c3e28d0de902e12f92c2466103c775cb2a5af1e6d4110e82223a6", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/iMeanAI/WebCanvas/HEAD/LICENSE"}], "id": "github:imeanai/webcanvas", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:00Z", "url": "https://raw.githubusercontent.com/iMeanAI/WebCanvas/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/iMeanAI/WebCanvas"}], "name": "WebCanvas", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/iMeanAI/WebCanvas/HEAD/README.md", "target_id": "arxiv:2406.12373", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Framework and evaluation resources for real-world web agents.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/iMeanAI/WebCanvas"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:51Z", "role": "primary", "sha256": "dfcbaa088cc21dd4e3c67d6eb3b3a1dc476217165ce2ca0daa2facfda71b5a38", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://github.com/jacklishufan/MobileWorld"}], "id": "github:jacklishufan/mobileworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:40:51Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/jacklishufan/MobileWorld"}], "name": "MobileWorldBench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://github.com/jacklishufan/MobileWorld", "target_id": "arxiv:2512.14014", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Benchmark of semantic world modeling for mobile agents, with question-answering tasks and GPT-judged generation tasks, plus training data built on AitW and AndroidControl screens.", "topics": ["evaluation", "mobile-navigation"], "updated_at": null, "url": "https://github.com/jacklishufan/MobileWorld"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["GUI interactions", "command-line operations"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/JetAstra/MacAgentBench/HEAD/README.md", "population": "documented MacAgentBench suite", "qualifier": "exact", "unit": "tasks", "value": 676}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/JetAstra/MacAgentBench/HEAD/README.md", "population": "documented application coverage", "qualifier": "exact", "unit": "applications", "value": 25}], "evaluation": "Deterministic rule-based evaluators with intermediate checkpoints.", "evidence_urls": ["https://raw.githubusercontent.com/JetAstra/MacAgentBench/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"], "focus": ["macOS task completion", "cross-application workflows"], "notes": "The benchmark evaluates multiple agent frameworks and GUI/CLI workflows. Framework skills and interfaces affect the task setting; the inspected README does not define one universal observation representation.", "observations": [], "setting": "interactive"}, "leaderboard_url": "https://jetastra.github.io/MacAgentBench/"}, "environments": ["desktop", "macos"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "f1f8c0cd5d984e4026abea858b54f0f1427c018cad7e75322e744f1e070225a9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/JetAstra/MacAgentBench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["details.benchmark_protocol", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "adf69997028988ba1eec34666afa119e007771dfc8e830c5dc50ad650896e897", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/JetAstra/MacAgentBench/HEAD/LICENSE"}], "id": "github:jetastra/macagentbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/JetAstra/MacAgentBench/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/JetAstra/MacAgentBench"}], "name": "MacAgentBench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48", "target_id": "arxiv:2606.22557", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "676 macOS tasks across 25 applications, each run in its own Docker container and scored by deterministic rule-based checks with multiple checkpoints.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/JetAstra/MacAgentBench"}
{"aliases": ["github:gimme1dollar/b-moca", "gimme1dollar/b-moca", "https://github.com/gimme1dollar/b-moca"], "authors": [], "details": {"benchmark_protocol": {"actions": ["normalized touch/lift coordinate pairs for tap or swipe"], "counts": [{"basis": "computed", "evidence_url": "https://api.github.com/repos/jylee425/b-moca/git/trees/release/ver.3?recursive=1", "population": "asset/tasks/**/*.textproto in retained release/ver.3 tree; not a paper test-task total", "qualifier": "exact", "unit": "task-definition files", "value": 131}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/asset/environments/config/environments_train.csv", "population": "environments_train.csv, retained release/ver.3 snapshot", "qualifier": "exact", "unit": "environment configuration rows", "value": 35}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/asset/environments/config/environments_test.csv", "population": "environments_test.csv, retained release/ver.3 snapshot", "qualifier": "exact", "unit": "environment configuration rows", "value": 10}], "evaluation": "Task success is detected using Appium; positive task reward terminates the episode. Environment variants change device/locale/wallpaper/DPI/dark mode and are loaded as snapshots. The numeric action interface is (touch_y, touch_x, lift_y, lift_x); wrapper/tanh transformations must be kept with the evaluated agent.", "evidence_urls": ["https://raw.githubusercontent.com/jylee425/b-moca/HEAD/readme.md", "https://api.github.com/repos/jylee425/b-moca/git/trees/release/ver.3?recursive=1", "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/bmoca/environment/coordinator.py", "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/bmoca/environment/environment.py", "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/asset/environments/config/environments_train.csv", "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/asset/environments/config/environments_test.csv"], "focus": ["Android control across visual/device variation", "generalization across environment snapshots"], "notes": "Current source inventory differs in grain from published experiment subsets: do not multiply 131 definitions by 35/10 configuration rows or label the product as evaluated tasks. Setup requires Android API 29, Appium, external app versions and generated AVD snapshots. No apps/snapshots were downloaded or run. Code Apache-2.0 does not establish data or app rights.", "observations": ["screenshot pixels", "XML view hierarchy", "goal instruction"], "setting": "interactive"}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "bd1354c8008d7d79d4883c0c037b7918392f86fab61391b39b52e507b04672ce", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/readme.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:05Z", "role": "primary", "sha256": "06ba40a5e0d5fb52e7bf5e5df636cb8e8bbbc56a8308dc6d528012489f7650ee", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/jylee425/b-moca/git/trees/release/ver.3?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:24:53Z", "role": "primary", "sha256": "34bc0c23949732e0a8091e334d59d2d41cae49a82cde850247aa0e6d739d2464", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/bmoca/environment/coordinator.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "d866af344c0cff5255b813576f2e9e9c5c2964e6809058b39e2b5180af71d946", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/bmoca/environment/environment.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:23:14Z", "role": "primary", "sha256": "c278cc28f8a0667b7fdc6ba638704cfbe43783cd496181231c45c865087543f3", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/asset/environments/config/environments_train.csv"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:23:15Z", "role": "primary", "sha256": "4b92933fe4f4f66e71a694b579940db11bedd5ea8f153752a44e633825b06fa3", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/asset/environments/config/environments_test.csv"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:58Z", "role": "primary", "sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T14:14:03Z", "role": "primary", "sha256": "9d4706306b04960f0eb4298887496961efdee6b7d41445e82abcac489d63779f", "supports": ["aliases"], "url": "https://api.github.com/repos/gimme1dollar/b-moca"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T14:14:04Z", "role": "primary", "sha256": "9d4706306b04960f0eb4298887496961efdee6b7d41445e82abcac489d63779f", "supports": ["aliases"], "url": "https://api.github.com/repos/jylee425/b-moca"}], "id": "github:jylee425/b-moca", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:58Z", "url": "https://raw.githubusercontent.com/jylee425/b-moca/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/jylee425/b-moca"}], "name": "B-MoCA", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Mobile control benchmark with varied interface configurations.", "topics": ["evaluation", "robustness"], "updated_at": null, "url": "https://github.com/jylee425/b-moca"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["tap/long-tap/swipe", "navigation buttons and text editing", "refuse, ask-consent and complete", "task-specific shortcut actions"], "counts": [], "evaluation": "Appium-backed task evaluators track goal achievement and harm prevention as separate booleans, plus risk-detection step and completion progress. The base evaluator defaults to max_step=15; task-specific overrides must be retained. Results are not a single generic success-rate metric.", "evidence_urls": ["https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/README.md", "https://api.github.com/repos/jylee425/mobilesafetybench/git/trees/release/ver.3?recursive=1", "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/mobile_safety/environment.py", "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/mobile_safety/component/parser.py", "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/mobile_safety/evaluator/_base_evaluator.py"], "focus": ["mobile task helpfulness", "harm prevention during Android interaction"], "notes": "Inspected release/ver.3 branch exposes task definitions, Android environment and per-task evaluators. It requires Android API 34, Appium/ADB and documented third-party app versions; those APKs/accounts were not acquired. The action parser includes semantic shortcuts in addition to screen gestures, so this is not a strictly coordinate-only action setting. No corpus task total is inferred from evaluator-module or file counts. Apache-2.0 applies to repository code, not automatically app/data assets.", "observations": ["screenshot pixels", "parsed UI hierarchy text", "task instruction and interaction history"], "setting": "interactive"}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "e733115d9b2fbe910cb3891162f04a89af8cbda25255c8f0bf60f78101e4309a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:05Z", "role": "primary", "sha256": "8dc66f6c5087a32c5c9fb7b3104ea6f9e0082255e2ba3cf9ba8244ce5a24da6f", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/jylee425/mobilesafetybench/git/trees/release/ver.3?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:09Z", "role": "primary", "sha256": "f56d6345546a9c67a2113cfba5b7148f5bd6b48fac5c34f1a5f4080e2d81ae2f", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/mobile_safety/environment.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "120487c3a6492371321dd3eebcd65dc59de6a82454d622d7fed3a7d86b8fa6c5", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/mobile_safety/component/parser.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:23:14Z", "role": "primary", "sha256": "918e3aa64ab3f578fb54580da123c1ff4508022b64ce74f88ce28c7a612756a4", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/mobile_safety/evaluator/_base_evaluator.py"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:58Z", "role": "primary", "sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/LICENSE.txt"}], "id": "github:jylee425/mobilesafetybench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:58Z", "url": "https://raw.githubusercontent.com/jylee425/mobilesafetybench/HEAD/LICENSE.txt"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/jylee425/mobilesafetybench"}], "name": "MobileSafetyBench", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Android emulator testbed that scores mobile device-control agents on both safety and helpfulness, using ADB and Appium for observations, actions and success checks.", "topics": ["safety-security", "evaluation"], "updated_at": null, "url": "https://github.com/jylee425/mobilesafetybench"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/LICENSE", "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/README.md"], "reason": "The README declares Apache-2.0, but the inspected LICENSE is a short notice and link rather than the complete terms; the code license remains unconfirmed."}}, "repository_id": 1278022559}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:54:57.854138+00:00", "role": "primary", "sha256": "3f7545f466407ed5336f47d18b12e8354ec352ff83dcf973b1545f81d9a629ac", "supports": ["name", "kind", "summary", "topics", "scope", "relations", "notes", "licenses.code"], "url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:26.237759+00:00", "role": "primary", "sha256": "4c92c584b6885a3bde8ee89d04ccf48287e7f6119889f75cde0c034c90bd8217", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/evaluation/memgui3k_offline_eval/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:01.061844+00:00", "role": "primary", "sha256": "d4e344274155be1d0a192c8ae2940169abb0bd8546d045b96b45810593860636", "supports": ["licenses.code", "notes"], "url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-19T19:10:26.954787+00:00", "role": "primary", "sha256": "0ae4477b885b0915e43874399c2a6b91f7d07954697370204550df931f7db82c", "supports": ["details.repository_id"], "url": "https://api.github.com/search/repositories?q=topic%3Agui-agent+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-18&per_page=100&page=1&sort=updated&order=asc"}], "id": "github:kwai/memgui-agent", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-20T03:54:57.854138+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/kwai/MemGUI-Agent"}], "name": "MemGUI-Agent", "notes": "The inspected release documentation describes dataset utilities, a LoRA SFT template and gold-step offline evaluation; live Android rollouts are explicitly outside that evaluator and linked to separate benchmark repositories. Executable source files, external datasets/checkpoints and reported results were not independently verified in this bounded review.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/README.md", "target_id": "arxiv:2606.19926", "type": "introduced_by"}], "review_status": "curated", "scope": "core", "summary": "MemGUI-Agent training and offline evaluation tools provide trajectory conversion, LoRA SFT templates, and step-level checks of GUI actions, memory updates and context folding.", "topics": ["memory", "interaction-history", "long-horizon", "imitation-learning", "evaluation"], "updated_at": null, "url": "https://github.com/kwai/MemGUI-Agent"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/kwai/MobileForge/main/LICENSE", "https://raw.githubusercontent.com/kwai/MobileForge/main/README.md"], "reason": "The complete root LICENSE declares Apache-2.0 for repository code. Linked data and model weights, and third-party dependencies, retain separate licensing requirements."}}, "repository_id": 1278021888}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:04.230792+00:00", "role": "primary", "sha256": "f78f58a6e3b8dd2ae3e8c49a15473782ae258ecd70719f7eeddfa43a10441adf", "supports": ["name", "kind", "summary", "topics", "scope", "relations", "notes", "environments", "licenses.code"], "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:29.533431+00:00", "role": "primary", "sha256": "58187fe25d8398ef2ae31de02debd4bc3b932ff1925db86a87f5de65b016aed1", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/training/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:07.420029+00:00", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code", "notes"], "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-19T19:10:26.954787+00:00", "role": "primary", "sha256": "0ae4477b885b0915e43874399c2a6b91f7d07954697370204550df931f7db82c", "supports": ["details.repository_id"], "url": "https://api.github.com/search/repositories?q=topic%3Agui-agent+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-18&per_page=100&page=1&sort=updated&order=asc"}], "id": "github:kwai/mobileforge", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-20T03:55:04.230792+00:00", "licenses": {"code": {"basis": "Complete repository root Apache License 2.0 text; code declaration only, not linked datasets or weights.", "declared": "Apache-2.0", "retrieved_at": "2026-10-20T03:55:07.420029+00:00", "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/kwai/MobileForge"}], "name": "MobileForge", "notes": "The root README describes MobileGym exploration/feedback and HiFPO optimization. The training README limits its public training path to the Qwen3-VL-8B experiment. Linked model and dataset collections, executable source completeness and benchmark results were not independently verified; the repository code license is not a model or dataset license.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/kwai/MobileForge/main/README.md", "target_id": "arxiv:2606.19930", "type": "introduced_by"}], "review_status": "curated", "scope": "core", "summary": "Android GUI-agent adaptation framework combining app exploration, generated curricula and rollout feedback with hint-conditioned step-level GRPO; the documented public training path targets Qwen3-VL-8B.", "topics": ["mobile-navigation", "data-generation", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://github.com/kwai/MobileForge"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "07431058d8b9a71b80e75b7a74a40a8346c1503eec240628459cc4d4e2f0802f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/kyle8581/Web-Shepherd/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "52925efd66489696c4e1f31f0f8353997fe534df9fcd42e50232cc78f0506903", "supports": ["identity"], "url": "https://api.github.com/repos/kyle8581/Web-Shepherd"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "ba7d28866cec4fb889499fc011473aedcdcc36fd8175c73527eafd7ef6b6dbba", "supports": ["release_status"], "url": "https://api.github.com/repos/kyle8581/Web-Shepherd/git/trees/main?recursive=1"}], "id": "github:kyle8581/web-shepherd", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T02:44:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/kyle8581/Web-Shepherd"}], "name": "Web-Shepherd inference and evaluation", "notes": "The tree contains src/webprm modules and a WebRewardBench evaluation script. The README’s illustrative folder layout differs from the actual tree, so a complete training pipeline is not claimed. No root code license is declared. Some old checkpoint links are unavailable; current model releases are separately identified through the official Hub collection.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Process-reward inference and benchmark-evaluation utilities that use task checklists to assess individual web-agent decisions.", "topics": ["reward-models", "evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/kyle8581/Web-Shepherd"}
{"aliases": [], "authors": [], "details": {}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:56Z", "role": "primary", "sha256": "95184255b6ec021aac8ca062b944f3ed76d1e01405e5771d9e31368021523eec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "notes"], "url": "https://raw.githubusercontent.com/lgy0404/LearnAct/HEAD/README.md"}], "id": "github:lgy0404/learnact", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-18T10:59:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/lgy0404/LearnAct"}], "name": "LearnAct", "notes": "At the retained README snapshot, LearnGUI-Online code and the real-world deployment implementation were marked coming soon. The mobile operating system is not established by the reviewed release instructions; no code was executed.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/lgy0404/LearnAct/HEAD/README.md", "target_id": "arxiv:2504.13805", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Released inference and evaluation code for demonstration-conditioned mobile GUI agents on LearnGUI-Offline. The repository describes a multi-agent adaptation method and links the LearnGUI demonstrations.", "topics": ["mobile-navigation", "imitation-learning", "evaluation"], "updated_at": null, "url": "https://github.com/lgy0404/LearnAct"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["predicted target coordinates"], "counts": [{"basis": "computed", "evidence_url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md", "population": "sum of icon/text counts in the published card table", "qualifier": "exact", "unit": "target annotations", "value": 1581}], "evaluation": "For positive targets, the predicted point must fall inside the annotated bounding box.", "evidence_urls": ["https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/readme.md", "https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/eval_screenspot_pro.py", "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md"], "focus": ["GUI grounding"], "notes": "Language, instruction style, and positive/negative target configuration are explicit evaluator options. This profile describes target localization, not end-to-end workflow completion. The target count is computed from the card table, not from expanded language/style evaluation instances.", "observations": ["high-resolution screenshots", "target instructions"], "setting": "offline"}, "leaderboard_url": "https://gui-agent.github.io/grounding-leaderboard"}, "environments": ["desktop", "windows", "macos", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "7b64ccad9cfec24c2c9cb35f55923bff93d53729cef3cc7463c79b29bdb1de1a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/readme.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:38:05Z", "role": "primary", "sha256": "a77c859775b33591b5619e8520c94702a53a1f9a29ada93e66744dbe23afe9dc", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/eval_screenspot_pro.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:35:55Z", "role": "primary", "sha256": "76ae2231564fe84af029f8625c02d9cd7fb6cf85c5f8af1b2a2593e2107719d9", "supports": ["details.benchmark_protocol", "environments"], "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "c78681e624c748b7a9980b6a9fd1514cec5140c7b3c7f08fe21b36bc7468f2ce", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/LICENSE"}], "id": "github:likaixin2000/screenspot-pro-gui-grounding", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding"}], "name": "ScreenSpot Pro", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI grounding benchmark focused on professional high-resolution interfaces.", "topics": ["evaluation", "grounding"], "updated_at": null, "url": "https://github.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["touch actions", "MCP tools (optional)"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ljang0/iOSWorld/HEAD/README.md", "population": "tasks.json benchmark", "qualifier": "exact", "unit": "tasks", "value": 133}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ljang0/iOSWorld/HEAD/README.md", "population": "bundled SwiftUI apps", "qualifier": "exact", "unit": "applications", "value": 26}], "evaluation": "Rubric satisfaction per task; full-task pass requires every criterion to be satisfied.", "evidence_urls": ["https://raw.githubusercontent.com/ljang0/iOSWorld/HEAD/README.md"], "focus": ["personalized phone tasks", "cross-app workflows"], "notes": "Vision-only, vision-plus-XML, and MCP modes expose different information and tools. The documented default evaluator is model-based; compare the same mode and judge.", "observations": ["screenshots", "accessibility XML (optional)"], "setting": "interactive"}, "evaluation_type": "rubric-based", "metrics": ["mean rubric score", "full-task pass rate"], "task_count": 133}, "environments": ["ios"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "706f9d770b65a1238e56f74f7dd5089c7756cc88e0ac09d80c7b2b3085a09b27", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.task_count", "details.metrics", "details.evaluation_type", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ljang0/iOSWorld/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "fa9fe1e31ed8da21d0b51621d3a3669ced4f0e42353aaf4bd8be1615d200eaac", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ljang0/iOSWorld/HEAD/LICENSE"}], "id": "github:ljang0/iosworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T10:48:41Z", "url": "https://raw.githubusercontent.com/ljang0/iOSWorld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ljang0/iOSWorld"}], "name": "iOSWorld", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Native iPhone-simulator benchmark with shared user data, application workflows, agent runners, and rubric-based scoring.", "topics": ["evaluation", "cross-app", "memory"], "updated_at": null, "url": "https://github.com/ljang0/iOSWorld"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["device actions through AgentEnv"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md", "population": "complete task collection", "qualifier": "exact", "unit": "tasks", "value": 495}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md", "population": "AITW-derived subset", "qualifier": "exact", "unit": "tasks", "value": 102}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md", "population": "self-constructed subset", "qualifier": "exact", "unit": "tasks", "value": 393}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md", "population": "self-constructed subset", "qualifier": "exact", "unit": "Android applications", "value": 46}], "evaluation": "Checks whether the recorded execution visits all annotated essential states rather than requiring an identical action sequence.", "evidence_urls": ["https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md"], "focus": ["mobile task completion", "essential-state evaluation"], "notes": "Application coverage refers to the self-constructed subset. Task execution and essential-state annotation are separate components; counts are publisher-reported, not a fresh task-file recount.", "observations": ["screenshots", "view hierarchies", "Android activity names"], "setting": "interactive"}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "4c4fe0fa92961f7286b6553b13117b2942e8482a48b7120a79e57dce2f8d42b9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "4c4fe0fa92961f7286b6553b13117b2942e8482a48b7120a79e57dce2f8d42b9", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md"}], "id": "github:llamatouch/llamatouch", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/LlamaTouch/LlamaTouch"}], "name": "LlamaTouch", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md", "target_id": "arxiv:2404.16054", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Mobile UI task evaluation using application state information.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/LlamaTouch/LlamaTouch"}
{"aliases": [], "authors": [], "details": {"availability": "Public graph/task manifests, evaluation source and ungated screenshot archive manifest.", "benchmark_protocol": {"actions": ["mobile GUI operations along graph transitions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "tasks", "value": 175}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "cross-app tasks", "value": 101}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "single-app tasks", "value": 74}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "mobile applications", "value": 21}], "evaluation": "The publisher reports success rate, completion rate and atomic-capability diagnostics using graph-based evaluation.", "evidence_urls": ["https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md"], "focus": ["long-horizon mobile task completion", "multi-path graph interaction"], "notes": "The interaction environment replays graph states rather than running real user accounts. The 101 and 74 subsets partition the reported 175 tasks; do not add the three counts.", "observations": ["mobile screenshots represented as graph states"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/LICENSE"], "reason": "Repository source LICENSE."}, "data": {"evidence_urls": ["https://huggingface.co/datasets/Norsheep/ColorBench/raw/main/README.md"], "reason": "Publisher metadata for the separately hosted screenshot archive; underlying third-party content retains applicable rights."}}, "repository_revision": "1f26a84230d26f99d84eab97d05032582b39018e"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:16Z", "role": "primary", "sha256": "5c0190fffb68b91c1855a408f6a5e252871d94000b95fa258504a3c100600cec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:16Z", "role": "primary", "sha256": "2f6940503cc0d6cb97bdf142d6b8c6ab116de82227f2fa6d97fd56d7e5aebec0", "supports": ["identity"], "url": "https://api.github.com/repos/MadeAgents/ColorBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:54Z", "role": "primary", "sha256": "6a00c093a6794ec4a172d30b70fdabc17400584a9e25a157a47a7661dbdd20fa", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/MadeAgents/ColorBench/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:06Z", "role": "primary", "sha256": "c0c3286b5150e128e55034ba0393ef85f7fcf73e4b73f4276389a04c9bcec919", "supports": ["identity", "release_status"], "url": "https://huggingface.co/api/datasets/Norsheep/ColorBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:06Z", "role": "primary", "sha256": "4bcf87ecfbbb8e07a01b21415a970c8b53a5283bf6872b657040d3f45c9241f7", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/Norsheep/ColorBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:54Z", "role": "primary", "sha256": "130976c026bcafa9dcc479697fe6595dbefb73cdb375fc47b8478acc16e6ea73", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/LICENSE"}], "id": "github:madeagents/colorbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T11:18:16Z", "licenses": {"code": {"basis": "Repository source LICENSE.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T11:22:54Z", "url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/LICENSE"}, "data": {"basis": "Publisher metadata for the separately hosted screenshot archive; underlying third-party content retains applicable rights.", "declared": "apache-2.0", "retrieved_at": "2026-10-17T11:23:06Z", "url": "https://huggingface.co/datasets/Norsheep/ColorBench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/MadeAgents/ColorBench"}], "name": "ColorBench", "notes": "The source tree contains graph and task JSON files plus construction/evaluation code; the linked ungated Hub provides the screenshot archive. The environment is static but interactive and graph-based, not a live-device deployment. Counts are publisher-reported task/app inventories. Apache-2.0 declarations for repository code and the Hub dataset are separately retained; third-party application visuals are not granted additional rights by this catalog.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "target_id": "arxiv:2510.14621", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A graph-based mobile GUI benchmark that evaluates long tasks through screenshot states and multiple valid or erroneous transitions, including reflection, replanning and backtracking.", "topics": ["evaluation", "mobile-navigation", "long-horizon", "cross-app"], "updated_at": null, "url": "https://github.com/MadeAgents/ColorBench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["judge verdicts"], "counts": [{"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending", "population": "paper-reported expert-reviewed set", "qualifier": "exact", "unit": "trajectories", "value": 1302}], "evaluation": "Compare automatic judgments with expert answers on task success, side effects and repetitiveness.", "evidence_urls": ["https://raw.githubusercontent.com/McGill-NLP/agent-reward-bench/HEAD/README.md", "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"], "focus": ["automatic evaluator agreement", "task success", "side effects", "repetition"], "notes": "Counts refer to the paper study; current Hub package files are not converted into trajectory totals.", "observations": ["web trajectories", "screenshots", "actions"], "setting": "offline"}, "leaderboard_url": "https://huggingface.co/spaces/McGill-NLP/agent-reward-bench-leaderboard", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/McGill-NLP/agent-reward-bench/raw/main/README.md"], "reason": "The dataset card supplies custom Terms of Use and preserves third-party conditions. The repository has no root code license, and no dataset license is inherited from its benchmarks."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "fb753e41c3eb19be172e9e7fccbc6bcadf44288507f82b190ca99e64144294b6", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/McGill-NLP/agent-reward-bench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "7dfac983a6e443d5f66e1fc0fc3706ebea4c93005f03457cc0a2b13a2bd89601", "supports": ["identity"], "url": "https://api.github.com/repos/McGill-NLP/agent-reward-bench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "79a1639cd4738719e51fc7f32313d63c0483914517cee96175459f63513e8c07", "supports": ["release_status"], "url": "https://api.github.com/repos/McGill-NLP/agent-reward-bench/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:18Z", "role": "primary", "sha256": "20c910fad3ad45a0c8c39b182c53632a821d3b6b546278bea6b7a78fbcb3f6cb", "supports": ["release_status"], "url": "https://huggingface.co/api/datasets/McGill-NLP/agent-reward-bench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:20Z", "role": "primary", "sha256": "e16adf817e3950441d70c7bb1bb3f44bd8dc65344938247f23fea4aaff9f654e", "supports": ["licenses.data", "summary"], "url": "https://huggingface.co/datasets/McGill-NLP/agent-reward-bench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "github:mcgill-nlp/agent-reward-bench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T03:59:54Z", "licenses": {"code": null, "data": {"basis": "Publisher dataset-card Terms of Use, including third-party restrictions; no standard SPDX or unrestricted permission is asserted.", "declared": "Custom AgentRewardBench Terms of Use", "retrieved_at": "2026-10-17T04:03:20Z", "url": "https://huggingface.co/datasets/McGill-NLP/agent-reward-bench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/McGill-NLP/agent-reward-bench"}], "name": "AgentRewardBench", "notes": "The public repository contains judges, evaluation utilities and annotation files; the linked Hub hosts trajectory/screenshot manifests. No root code license is declared. Dataset Terms of Use are custom and retain third-party restrictions rather than declaring an SPDX license. The paper-reported 1,302 trajectories describe its reviewed study population, not a count recomputed from current Hub files.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/McGill-NLP/agent-reward-bench/HEAD/README.md", "target_id": "arxiv:2504.08942", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A benchmark and evaluation toolkit for comparing automatic web-agent judges against expert judgments of task success, side effects and repetition.", "topics": ["evaluation", "reward-models", "web-navigation"], "updated_at": null, "url": "https://github.com/McGill-NLP/agent-reward-bench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["web interactions"], "counts": [{"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending", "population": "safe split", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2503.04957", "unit": "tasks", "value": 250}, {"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending", "population": "harmful split", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2503.04957", "unit": "tasks", "value": 250}], "evaluation": "Task outcomes and the Agent Risk Assessment categories evaluate agent behavior on benign and harmful requests.", "evidence_urls": ["https://raw.githubusercontent.com/McGill-NLP/safearena/HEAD/README.md", "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"], "focus": ["web-agent safety", "task completion"], "notes": "Use SafeArena-specific containers rather than ordinary WebArena instances. The documented task splits require access approval; the profile does not imply unrestricted dataset access.", "observations": [], "setting": "interactive"}, "leaderboard_url": "https://huggingface.co/spaces/McGill-NLP/safearena-leaderboard"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T09:36:33Z", "role": "primary", "sha256": "ae50121aa60ac26b32d8d6ad6f578f9091a81a3cb9ca28b98050bddee16c929a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/McGill-NLP/safearena/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "github:mcgill-nlp/safearena", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T09:36:33Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/McGill-NLP/safearena"}], "name": "SafeArena", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/McGill-NLP/safearena/HEAD/README.md", "target_id": "arxiv:2503.04957", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Web-agent benchmark pairing benign and harmful tasks to study deliberate misuse risks.", "topics": ["safety-security", "evaluation"], "updated_at": null, "url": "https://github.com/McGill-NLP/safearena"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["predicted browser actions"], "counts": [{"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50", "population": "original paper benchmark", "qualifier": "approximately", "unit": "interactions", "value": 100000}, {"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50", "population": "original paper benchmark", "qualifier": "approximately", "unit": "expert demonstrations", "value": 2300}, {"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50", "population": "original paper benchmark", "qualifier": "more-than", "unit": "websites", "value": 150}], "evaluation": "Offline evaluation compares action predictions against recorded human demonstrations.", "evidence_urls": ["https://raw.githubusercontent.com/mcgill-nlp/weblinx/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50", "https://huggingface.co/datasets/McGill-NLP/WebLINX/raw/main/README.md"], "focus": ["conversational web navigation", "next-action prediction"], "notes": "The figures are rounded descriptions from the original paper. WebLINX 1.1 / BrowserGym removes some demonstrations and changes evaluated steps, including tab actions; do not mix those results or counts with the original release. This profile does not treat replayed demonstration steps as live task completion.", "observations": ["dialogue history", "HTML", "screenshots", "action history"], "setting": "offline"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "15c3f1656afd861894b180bca7d9214b4cf78a6ad9cd0601c80f5d67d2839528", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/mcgill-nlp/weblinx/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "15c3f1656afd861894b180bca7d9214b4cf78a6ad9cd0601c80f5d67d2839528", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/mcgill-nlp/weblinx/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:23Z", "role": "primary", "sha256": "9bd7512e257f5914d66b091530d7eb711a6f90bc21dc89cbba53015c7136c527", "supports": ["details.benchmark_protocol"], "url": "https://huggingface.co/datasets/McGill-NLP/WebLINX/raw/main/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "cc0156153933fb96f6f3bf8d736c15e24d78d0838945c2ed456e3ada96cf67c4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/mcgill-nlp/weblinx/HEAD/LICENSE"}], "id": "github:mcgill-nlp/weblinx", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/mcgill-nlp/weblinx/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/mcgill-nlp/weblinx"}], "name": "WebLINX", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/mcgill-nlp/weblinx/HEAD/README.md", "target_id": "arxiv:2402.05930", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Demonstrations and evaluation for conversational web navigation.", "topics": ["evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/mcgill-nlp/weblinx"}
{"aliases": [], "authors": [], "details": {"availability": "Public fine-tuning/evaluation tooling, examples and results viewer.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/McGill-NLP/webllama/HEAD/LICENSE", "https://raw.githubusercontent.com/McGill-NLP/webllama/HEAD/README.md"], "reason": "README explicitly scopes MIT to code unless a file header states otherwise; models/data/images retain original licenses."}}, "repository_revision": "1b7b41327b2222312e07818ce239c04307beaf9b"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:55Z", "role": "primary", "sha256": "121546c7a73f4d7604ae5340bd78856e8170dc8aaa8e57f7e62efcfb85d8d4c9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "licenses.code"], "url": "https://raw.githubusercontent.com/McGill-NLP/webllama/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:55Z", "role": "primary", "sha256": "6605b7b09e8be77cce9d5bfd1a6199c2256cbfc1f067b93b2b952c1d19e74f10", "supports": ["identity"], "url": "https://api.github.com/repos/McGill-NLP/webllama"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:06Z", "role": "primary", "sha256": "915620d7a155f5c5fe099cdb0461d9614e1894b38559f0d2f9a6a138afe06212", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/McGill-NLP/webllama/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "e69268828e67ffd1d3393a2d3530f97df6deaa0b961039fc971abea28cec8715", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/McGill-NLP/webllama/HEAD/LICENSE"}], "id": "github:mcgill-nlp/webllama", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T12:30:55Z", "licenses": {"code": {"basis": "README explicitly scopes MIT to code unless a file header states otherwise; models/data/images retain original licenses.", "declared": "MIT", "retrieved_at": "2026-10-17T12:34:10Z", "url": "https://raw.githubusercontent.com/McGill-NLP/webllama/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/McGill-NLP/webllama"}], "name": "WebLlama", "notes": "The manifest contains modeling/training code, BrowserGym examples, documentation and a results viewer. The README distinguishes currently available components from planned additional benchmarks and deployment integrations. It requests citation of WebLINX as the foundation for data, training and evaluation; this is not a claim that the WebLINX paper first introduced every later WebLlama component. MIT applies to repository code, subject to file-header exceptions; weights and datasets have their own terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/McGill-NLP/webllama/HEAD/README.md", "target_id": "arxiv:2402.05930", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Training, evaluation and integration tooling for dialogue-guided web agents, including Llama fine-tuning workflows and BrowserGym examples built around WebLINX.", "topics": ["web-navigation", "imitation-learning", "evaluation", "tool-integration"], "updated_at": null, "url": "https://github.com/McGill-NLP/webllama"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["pixel (x,y) cursor coordinate"], "counts": [], "evaluation": "The inspected evaluator scales target boxes from a 1000×1000 reference grid into image pixels and tests pixel predictions, allowing ±10 pixels around the center along any target dimension narrower than 1 pixel after scaling. It reports cumulative hit accuracy by turn (earlier hits remain correct), miss distances, any-turn hit rate, and a correction rate whose denominator is all successfully processed samples. The default budget is five turns; retain prompt/feedback variant, turn budget, and processed/skipped coverage.", "evidence_urls": ["https://raw.githubusercontent.com/microsoft/precision-cua-bench/HEAD/README.md", "https://raw.githubusercontent.com/microsoft/precision-cua-bench/HEAD/eval/test_allturns_api.py", "https://api.github.com/repos/microsoft/precision-cua-bench/git/trees/main?recursive=1"], "focus": ["pixel-precise editor cursor placement", "multi-turn visual self-correction"], "notes": "The release has four editor/theme JSONL partitions (Cursor/VS Code × dark/light) and screenshot files. Partitions are not train/test splits or independent task totals. README’s initial field table calls bbox pixel coordinates, but the evaluation-format section and actual ratio_h/ratio_w=dimension/1000 code use normalized target boxes; profile follows the inspected evaluator and retains the inconsistency. Source/images were not executed or downloaded as a dataset. MIT is recorded for repository code only.", "observations": ["code-editor screenshot", "cursor-location instruction", "red-cross feedback marking previous incorrect prediction"], "setting": "offline"}}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:52Z", "role": "primary", "sha256": "23d0dc8ea0a10c080cf8ae32121eade2785ae923a6b0d853d166c94c195e789b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://github.com/microsoft/precision-cua-bench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "84ad4fb798abb776dce813cf4dab296d6e03248d8ae29388e1f06cccd27d1241", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/microsoft/precision-cua-bench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:08Z", "role": "primary", "sha256": "1fda7c3980fe7a7c373478eda6ecfabff227e92c2b1f19c6fe903b408f583797", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/microsoft/precision-cua-bench/HEAD/eval/test_allturns_api.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:06Z", "role": "primary", "sha256": "ede84e3dda47f80fada4234caa2423db08e21ff7667a37599c03aae657b2f84b", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/microsoft/precision-cua-bench/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["relations"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "c2cfccb812fe482101a8f04597dfc5a9991a6b2748266c47ac91b6a5aae15383", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/microsoft/precision-cua-bench/HEAD/LICENSE"}], "id": "github:microsoft/precision-cua-bench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:40:52Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/microsoft/precision-cua-bench/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/microsoft/precision-cua-bench"}], "name": "Precision CUA Bench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50", "target_id": "arxiv:2604.13019", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation resources for pixel-precise cursor grounding in code editors.", "topics": ["evaluation", "grounding"], "updated_at": null, "url": "https://github.com/microsoft/precision-cua-bench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["agent-defined Windows interactions"], "counts": [{"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/src/win-arena-container/client/evaluation_examples_windows/test_all.json", "population": "evaluation_examples_windows/test_all.json", "qualifier": "exact", "unit": "task definitions", "value": 154}], "evaluation": "Task configuration selects an evaluator and expected application or system outcome; evaluators return a reward for the resulting state.", "evidence_urls": ["https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/README.md", "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/docs/Develop-Tasks.md", "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/src/win-arena-container/client/evaluation_examples_windows/test_all.json"], "focus": ["Windows task completion", "desktop workflows"], "notes": "Counted unique identifiers in the inspected manifest. Normal and hard modes differ in how much task setup the agent must perform. Record the difficulty and observation/backend configuration with each result; no VM was launched.", "observations": ["screenshots", "accessibility information", "screen-element detections"], "setting": "interactive"}}, "environments": ["windows", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "e481abd422a4cc8b819ad84b32684e4af641e18a7c0d277dd608cfd5d3c5e6ad", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "e481abd422a4cc8b819ad84b32684e4af641e18a7c0d277dd608cfd5d3c5e6ad", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:06:30Z", "role": "primary", "sha256": "c565dc02ab0049bf9710b87e2760dda8d7c92268ea78f0354952695ab59c0de3", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/docs/Develop-Tasks.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:06:30Z", "role": "primary", "sha256": "902fb24cf0a9e146bf6c29288e74a7254fe40294e49dd47363e0b15d65608fef", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/src/win-arena-container/client/evaluation_examples_windows/test_all.json"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "9906940f61b1f0b533fa7d99baf55178b2808fbe113ea51dfbfad8572ccd5f2b", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/LICENSE"}], "id": "github:microsoft/windowsagentarena", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/microsoft/WindowsAgentArena"}], "name": "Windows Agent Arena", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/microsoft/WindowsAgentArena/HEAD/README.md", "target_id": "arxiv:2409.08264", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Windows 11 benchmark for multimodal desktop agents in a reproducible VM that runs locally or in parallel on Azure ML, with a harder mode where agents must set tasks up themselves.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/microsoft/WindowsAgentArena"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["Selenium-backed browser interactions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/README.md", "population": "WebVoyager task collection", "qualifier": "exact", "unit": "task queries", "value": 643}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/README.md", "population": "WebVoyager task collection", "qualifier": "exact", "unit": "websites", "value": 15}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/README.md", "population": "separate GAIA validation-derived web subset", "qualifier": "exact", "unit": "tasks", "value": 90}], "evaluation": "The repository supplies an automated multimodal-model evaluation protocol.", "evidence_urls": ["https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/README.md"], "focus": ["live-web task completion"], "notes": "The GAIA-derived subset is separate from the 643 WebVoyager queries. Live sites and time-sensitive travel tasks can change; record the task snapshot and evaluation date.", "observations": ["webpage screenshots", "textual task instructions"], "setting": "interactive"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "b5523941b2773634fdcf359b5b89bf0aca589f33ad9db30692c8b1912ed5dda5", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "43070e2d4e532684de521b885f385d0841030efa2b1a20bafb76133a5e1379c1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/LICENSE"}], "id": "github:minorjerry/webvoyager", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/MinorJerry/WebVoyager"}], "name": "WebVoyager", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/MinorJerry/WebVoyager/HEAD/README.md", "target_id": "arxiv:2401.13919", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Multimodal web navigation agent and evaluation tasks on real websites.", "topics": ["evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/MinorJerry/WebVoyager"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "e1d75e8a196f19af6050c53694383f069561244c810ca259c5f8075ee91f72b1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/MobileAgentBench/mobile-agent-bench/HEAD/README.md"}], "id": "github:mobileagentbench/mobile-agent-bench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/MobileAgentBench/mobile-agent-bench"}], "name": "MobileAgentBench", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Automated benchmark for mobile LLM agents whose default tasks use ten Simple Mobile Tools Android apps, installed as a Python library that detects task completion.", "topics": ["evaluation", "mobile-navigation"], "updated_at": null, "url": "https://github.com/MobileAgentBench/mobile-agent-bench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": [], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/open-compass/MMBench-GUI/HEAD/README.md", "population": "reported multi-level benchmark", "qualifier": "more-than", "unit": "tasks", "value": 8000}], "evaluation": "Level-specific evaluation; Efficiency–Quality Area (EQA) combines navigation quality and efficiency.", "evidence_urls": ["https://raw.githubusercontent.com/open-compass/MMBench-GUI/HEAD/README.md"], "focus": ["screen understanding", "grounding", "task automation", "task collaboration"], "notes": "The README documents four levels, but its release notes and TODOs distinguish released L1/L2 evaluation from pending L3/L4 integration. Do not treat every level as the same evaluator or a complete implemented interface.", "observations": ["screenshots"], "setting": "mixed"}}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "35db9fca3ed1c1a3513924fefff4ccd224bc70cd7eb4bf8c54fcb9048851988e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/open-compass/MMBench-GUI/HEAD/README.md"}], "id": "github:open-compass/mmbench-gui", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/open-compass/MMBench-GUI"}], "name": "MMBench-GUI", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/open-compass/MMBench-GUI/HEAD/README.md", "target_id": "arxiv:2507.19478", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Four-level GUI-agent benchmark across six platforms, from content understanding and element grounding to task automation and collaboration, scored with an efficiency-quality area metric.", "topics": ["evaluation", "grounding"], "updated_at": null, "url": "https://github.com/open-compass/MMBench-GUI"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["GUI gestures", "foreground-app tool calls in hybrid mode"], "counts": [{"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/bench_env/benchmark/official.json", "population": "official.json", "qualifier": "exact", "unit": "task definitions", "value": 215}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md", "population": "complete simulator installation", "qualifier": "exact", "unit": "simulated clients", "value": 63}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md", "population": "new clients in this release", "qualifier": "exact", "unit": "simulated clients", "value": 35}], "evaluation": "Programmatic checkers inspect changes in application state; both modes share task definitions and initial states.", "evidence_urls": ["https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md", "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/bench_env/benchmark/official.json"], "focus": ["task completion", "cross-application workflows", "GUI and app-native tool comparison"], "notes": "The task count was checked against unique task_id values in the published manifest. The 35 new clients are a subset of 63 total clients. This simulator runs in a browser; Android-emulator apps are a separate part of OpenMobile-2.", "observations": ["screenshots"], "setting": "interactive"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:10Z", "role": "primary", "sha256": "d9f850ec196e715650e679cb81515adaa64ddc5a60a882bfe97b0f6a098ddc83", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:46:06Z", "role": "primary", "sha256": "9b385e9f6aa2f7169f616a76887df3b76df605cb2bea071d99ee7ca3f9b659cc", "supports": ["relations"], "url": "https://os-copilot.github.io/OpenMobile2-Home/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:50:11Z", "role": "primary", "sha256": "35c1641b349201c68917fbe4c4036814487678e9588d9b1c903af26917800316", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/bench_env/benchmark/official.json"}, {"method": "license_header", "retrieved_at": "2026-10-17T01:52:51Z", "role": "primary", "sha256": "7618cd7de039c264759ca8ca0b9130c6de63a4536dcf436baedd4cfbf639073e", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/LICENSE"}], "id": "github:openmobile2/mobilegym-plusplus", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T01:48:10Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T01:52:51Z", "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OpenMobile2/MobileGym-plusplus"}], "name": "MobileGym++ / MobileGym++ Bench", "notes": "These clients simulate named commercial applications and are not those commercial products. The code and simulator data have distinct license files. No benchmark run was performed. The README describes a phone that runs in the browser, and the Android-emulator playground is a separate OpenMobile2 release, so the platform facet is web rather than a mobile operating system.", "published_at": null, "relations": [{"evidence_url": "https://os-copilot.github.io/OpenMobile2-Home/", "target_id": "arxiv:2610.14157", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Browser-rendered mobile simulation with resettable app state and a shared task suite for GUI-only and GUI-plus-tool agents.", "topics": ["mobile-navigation", "evaluation", "cross-app", "tool-integration"], "updated_at": null, "url": "https://github.com/OpenMobile2/MobileGym-plusplus"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:10Z", "role": "primary", "sha256": "485279038266d9454df75e27a80a0a0d71c24e90700126d2dcfa0da921b1273d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T01:52:51Z", "role": "primary", "sha256": "b09fcaf4fbc265223c2479c410beef2b98c5e687c78ecedf3452aa12ffc465f1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/LICENSE"}], "id": "github:openmobile2/openmobile2-code", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T01:48:10Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T01:52:51Z", "url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OpenMobile2/OpenMobile2-Code"}], "name": "OpenMobile-2 code", "notes": "The README marks evaluation code, checkpoints, environments, and data as released; data-construction scripts and other resources remain pending. Each benchmark runtime is started separately. Code terms do not establish dataset permissions.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/README.md", "target_id": "arxiv:2610.14157", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation adapters and training guidance for OpenMobile-2 across Android emulators and browser-rendered mobile simulation.", "topics": ["mobile-navigation", "tool-integration", "evaluation"], "updated_at": null, "url": "https://github.com/OpenMobile2/OpenMobile2-Code"}
{"aliases": [], "authors": [], "details": {"availability": "Public task definitions, execution harness and an ungated VM-file manifest.", "benchmark_protocol": {"actions": ["actions executed by the OSWorld-derived DesktopEnv"], "counts": [{"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/all.json", "population": "evaluation_examples/all.json", "qualifier": "exact", "unit": "category/task references", "value": 416}], "evaluation": "Task configurations provide evaluators for application/system state, using an OSWorld-derived runtime.", "evidence_urls": ["https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md", "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/all.json", "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/README.md"], "focus": ["breadth of desktop task coverage", "depth of desktop task execution"], "notes": "416 category/task pairs are unique, but only 412 bare ID strings are unique across categories. This is a manifest-reference count, not a claim of 416 semantically distinct tasks. Main runtime documentation specifies Ubuntu; the subdirectory README retains older Windows-only wording.", "observations": ["desktop observations from an Ubuntu virtual machine"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/LICENSE"], "reason": "Repository code LICENSE; no license is inferred for the VM image or bundled third-party applications."}}, "repository_revision": "191efd0e6766acab444a31685b437565e1a94ac2"}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:14Z", "role": "primary", "sha256": "fc9f3c8d9fffdff8b51e312b013be0d9fdfc10e57fdc2ec4a8a8d012c4b42475", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:14Z", "role": "primary", "sha256": "d6c91b4910e92f40cc694b4cb6467c1a20f42c0c14047b95a1e777aca0892677", "supports": ["identity"], "url": "https://api.github.com/repos/OS-Copilot/OS-Map"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:07Z", "role": "primary", "sha256": "41342157c53f28ad2729137491930225ac11da922e7b16762e84ac958a2e9d28", "supports": ["release_status"], "url": "https://api.github.com/repos/OS-Copilot/OS-Map/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:07Z", "role": "primary", "sha256": "1dd924135cbf0a85ce72d75025259adb7b849970d400af4ed854327a34e7ee41", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:17:22Z", "role": "primary", "sha256": "4199af4c8bc60451079d8a2e7598ab7386bd82b0d5c9ad7e6927d6cb8e03ad6f", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/all.json"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:17:22Z", "role": "primary", "sha256": "a6b60d5721cfdd26416e7d0ae3192360896427a97ccb3793812cc150e9cd7126", "supports": ["details.benchmark_protocol", "notes"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:16Z", "role": "primary", "sha256": "d2da837872b123f2927a069a020aeb14fb4b22148599c9f2e8b6689ee444ec88", "supports": ["release_status", "details"], "url": "https://huggingface.co/api/datasets/os-map/OSMAP"}], "id": "github:os-copilot/os-map", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T09:12:14Z", "licenses": {"code": {"basis": "Repository code LICENSE; no license is inferred for the VM image or bundled third-party applications.", "declared": "MIT", "retrieved_at": "2026-10-17T09:14:07Z", "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OS-Copilot/OS-Map"}], "name": "OS-Map", "notes": "Public task manifests, agent runner and evaluator code are present, and the linked ungated Hub manifest lists Ubuntu0.zip and task files. The main setup supports x86 VMware Workstation and says ARM/Docker alternatives are still in development. An inherited evaluation-examples README still mentions Windows 10 and under-construction status; those statements are retained as a documentation inconsistency, not interpreted as verified Windows support. The manifest has 416 category/task references but 412 globally distinct ID strings.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md", "target_id": "github:xlang-ai/osworld", "type": "uses"}, {"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md", "target_id": "arxiv:2507.19132", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "An OSWorld-derived desktop benchmark with an Ubuntu virtual-machine environment and expanded task configurations for measuring breadth and depth of computer interaction.", "topics": ["evaluation", "long-horizon", "cross-app"], "updated_at": null, "url": "https://github.com/OS-Copilot/OS-Map"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["PyAutoGUI desktop actions", "code/tool commands and answer/stop actions according to configured agent mode"], "counts": [{"basis": "computed", "evidence_url": "https://api.github.com/repos/OS-Copilot/ScienceBoard/git/trees/main?recursive=1", "population": "tasks/VM/**/*.json at tree 21b5eb0f; excludes tasks/Raw and tasks/Dual", "qualifier": "exact", "unit": "VM task definitions", "value": 169}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/verifier-audit.md", "population": "VM suite: Celestia, ChimeraX, GrassGIS, KAlgebra, Lean, TeXstudio", "qualifier": "exact", "unit": "application families", "value": 6}], "evaluation": "Application-specific verifiers inspect state, files, database values, equations, compilation or answer/stop conditions. The official audit records 242 verifier entries across its VM task suite. This profile covers the released VM task definitions; Docker smoke results and Raw/Dual alternatives are separate populations.", "evidence_urls": ["https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/verifier-audit.md", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/known-issues.md", "https://api.github.com/repos/OS-Copilot/ScienceBoard/git/trees/main?recursive=1", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/sci/base/task.py"], "focus": ["scientific-software desktop workflows", "GUI and code-assisted scientific tasks"], "notes": "Manifest count 169 matches the official verifier audit. Other Raw/Dual definitions can represent alternative execution modes of the same problems and are not added. Official Docker migration documents a 20-task acceptance subset, not a full-suite performance result; legacy VMware remains. External service sign-in/rate limits, amd64 binaries and ChimeraX redistribution terms limit portability. Only code MIT is established; VM, scientific data and bundled apps need their own terms.", "observations": ["desktop screenshots", "accessibility-tree observations", "set-of-marks with accessibility tree"], "setting": "interactive"}}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "4c3fe93aa740dbf4a279a246ddd836f2590a2a36f39b820abe166e38cb484577", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "d276e11127abc5826bd1b2aa2caf543fcd45a8a60477774e8c8b1054067997ec", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/verifier-audit.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "4f90270617c0c1d58f523c0a7f2e4603fa67a49560d68e2f88cfa080d764313b", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/known-issues.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:07Z", "role": "primary", "sha256": "3398670a517a5d1d71f420dd866493d3f8a77853e3933fd4928103f9a8be1e4f", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/OS-Copilot/ScienceBoard/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:35:24Z", "role": "primary", "sha256": "d6a1481b453c074e34739fd29def56efa35a1dc10e856418f557f067e9ff08c5", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/sci/base/task.py"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "cbb1d28b515a39c294963d4bfc24077e669a483fc5ebba9b84a727afd4187426", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/LICENSE"}], "id": "github:os-copilot/scienceboard", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OS-Copilot/ScienceBoard"}], "name": "ScienceBoard", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md", "target_id": "arxiv:2505.19897", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Scientific-workflow tasks for multimodal agents in desktop software such as ChimeraX, KAlgebra, Celestia, Grass GIS and Lean 4, with apps modified to expose state for evaluation.", "topics": ["evaluation", "science"], "updated_at": null, "url": "https://github.com/OS-Copilot/ScienceBoard"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:54Z", "role": "primary", "sha256": "61eb6109404f266fd22c6fa8ca58ef773a9d502567403f42edc096958ab2e430", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/EIA_against_webagent/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:46Z", "role": "primary", "sha256": "953ff47dbbc0c0563d1ec2d245b06d1206f2f38c4ed2f0da728e027ae2e5a973", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/EIA_against_webagent/HEAD/LICENSE"}], "id": "github:osu-nlp-group/eia_against_webagent", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T10:59:54Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-18T11:03:46Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/EIA_against_webagent/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/EIA_against_webagent"}], "name": "EIA", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/EIA_against_webagent/HEAD/README.md", "target_id": "arxiv:2409.11295", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation suite for Environmental Injection Attacks that embed invisible malicious HTML elements into benign websites to test privacy leakage in generalist web agents.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/EIA_against_webagent"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["CLICK", "TYPE", "SELECT"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md", "population": "train split", "qualifier": "exact", "unit": "task instances", "value": 1009}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md", "population": "cross-task test split", "qualifier": "exact", "unit": "task instances", "value": 252}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md", "population": "cross-website test split", "qualifier": "exact", "unit": "task instances", "value": 177}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md", "population": "cross-domain test split", "qualifier": "exact", "unit": "task instances", "value": 912}], "evaluation": "Offline candidate and action evaluation against recorded demonstrations; the repository distinguishes macro from micro averaging.", "evidence_urls": ["https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md"], "focus": ["web action prediction", "generalization to unseen sites and domains"], "notes": "These are reported split inventories, not online task-success measurements. Original Mind2Web, Multimodal-Mind2Web, and Online-Mind2Web have different inputs/protocols. The authors require macro-average metrics for comparison with the paper; test archives remain upstream.", "observations": ["task instructions", "HTML", "recorded action history"], "setting": "offline"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "157dbd840264ea29f958c0b7383d63783845c573b3e19a1f39b773be22e0b32d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "157dbd840264ea29f958c0b7383d63783845c573b3e19a1f39b773be22e0b32d", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "953ff47dbbc0c0563d1ec2d245b06d1206f2f38c4ed2f0da728e027ae2e5a973", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/LICENSE"}], "id": "github:osu-nlp-group/mind2web", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/Mind2Web"}], "name": "Mind2Web", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web/HEAD/README.md", "target_id": "arxiv:2306.06070", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Human demonstrations and tasks for generalist web navigation research.", "topics": ["evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/Mind2Web"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://osu-nlp-group.github.io/Mind2Web-2/#leaderboard"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T17:18:21Z", "role": "primary", "sha256": "5cc72b992cfc11094dcf341c284d5baadc6c66a302fdfd474b27913fe1c91f6f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web-2/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T17:18:56Z", "role": "primary", "sha256": "93b43d692b033b76129504448695bfe76ef22d18b11e51352bfab4b5622e5aaa", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web-2/HEAD/LICENSE"}], "id": "github:osu-nlp-group/mind2web-2", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T17:18:21Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T17:18:56Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web-2/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/Mind2Web-2"}], "name": "Mind2Web 2", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Long-horizon agentic web search benchmark whose Agent-as-a-Judge framework scores time-varying, citation-backed answers, with evaluation scripts for its dev and test sets.", "topics": ["evaluation", "web-navigation", "long-horizon"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/Mind2Web-2"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["web interactions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/Online-Mind2Web/HEAD/README.md", "population": "documented Online-Mind2Web suite", "qualifier": "exact", "unit": "tasks", "value": 300}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/Online-Mind2Web/HEAD/README.md", "population": "documented site coverage", "qualifier": "exact", "unit": "websites", "value": 136}], "evaluation": "Human evaluation and WebJudge-based evaluation are documented as distinct assessment paths.", "evidence_urls": ["https://raw.githubusercontent.com/OSU-NLP-Group/Online-Mind2Web/HEAD/README.md"], "focus": ["live-web task completion"], "notes": "Tasks are actively replaced when websites or requirements become invalid. Record the task revision and submission schema; a constant headline task count does not imply an unchanged test set.", "observations": [], "setting": "interactive"}, "leaderboard_url": "https://huggingface.co/spaces/osunlp/Online_Mind2Web_Leaderboard"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "49d7cc317fd2ba0df65abf65a0357d1fa2384e59019a96d4e24b784888599eaa", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Online-Mind2Web/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "93675f9d13697ae98889d703b8fae86a7200623e7ae8d844c9ee2200ad480583", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Online-Mind2Web/HEAD/LICENSE"}], "id": "github:osu-nlp-group/online-mind2web", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Online-Mind2Web/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/Online-Mind2Web"}], "name": "Online Mind2Web", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "300 tasks on 136 live websites for evaluating web agents online, with outdated tasks replaced over time, human-reviewed submissions and the WebJudge automatic evaluator.", "topics": ["evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/Online-Mind2Web"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "782dea8635b417bc89d21489cacea215ee6878ba5a1d4b9e18ca7b2913545ab6", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/RedTeamCUA/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "10a9bb268e611da40ae6edba6777917abb6e97d43dcb480e772ae8169b687e10", "supports": ["identity"], "url": "https://api.github.com/repos/OSU-NLP-Group/RedTeamCUA"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "712f1b7bc13b593cf23ecfab72cec33d1c256afe23a383e49d7d5b2996a0f55c", "supports": ["release_status"], "url": "https://api.github.com/repos/OSU-NLP-Group/RedTeamCUA/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/RedTeamCUA/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/RedTeamCUA/HEAD/LICENSE"}], "id": "github:osu-nlp-group/redteamcua", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:44:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T02:45:20Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/RedTeamCUA/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/RedTeamCUA"}], "name": "RedTeamCUA", "notes": "The release includes environment code, adversarial scenario configuration and RTC-Bench task-generation assets. The project declares Apache-2.0, while incorporated environments retain their licenses. It distinguishes full task runs from tests initialized at injection locations; those settings must not be treated as interchangeable results.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/RedTeamCUA/HEAD/README.md", "target_id": "arxiv:2505.21936", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A controlled testing framework that combines desktop virtual machines and replicated websites to evaluate computer-use agents against indirect prompt injection.", "topics": ["safety-security", "evaluation", "sandbox"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/RedTeamCUA"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:53Z", "role": "primary", "sha256": "be53561dcf69cb2295a26eaebf965eead10b3504fca30f71f73dc5e5728fc5b3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/WebGuard/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:46Z", "role": "primary", "sha256": "93b43d692b033b76129504448695bfe76ef22d18b11e51352bfab4b5622e5aaa", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/WebGuard/HEAD/LICENSE"}], "id": "github:osu-nlp-group/webguard", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T10:59:53Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-18T11:03:46Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/WebGuard/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/WebGuard"}], "name": "WebGuard", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/WebGuard/HEAD/README.md", "target_id": "arxiv:2507.14293", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Codebase, annotation toolkit, and action-level risk dataset for training and evaluating guardrail models that flag unsafe browser-agent actions before execution.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/WebGuard"}
{"aliases": [], "authors": [], "details": {}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:51:38.123692Z", "role": "primary", "sha256": "3226df7608039522072e59d459094fc45be9c5a6be7ea0d7ad76ddf9723a3aea", "supports": ["name", "kind", "summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/PatrickStarL/gui-grounding-eval/d92b059caa0e0acf8db76f2164dd7be2617ccc72/README.md"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:04:53.470789+00:00", "role": "primary", "sha256": "3d76d934e98641746085772b98ee17059749f980a616435498a7ad483bd4cec1", "supports": ["id", "url"], "url": "https://api.github.com/search/repositories?q=topic%3Agui-grounding+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-20&per_page=100&page=1&sort=updated&order=asc"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:51:05.389053Z", "role": "primary", "sha256": "4fe6f65a21586dde0bf8b64f8c49e5c583b320bf5e0c2b8253baf2cd3fcfc5d2", "supports": ["id", "url", "notes"], "url": "https://github.com/PatrickStarL/gui-grounding-eval"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:53:32.022035Z", "role": "primary", "sha256": "c3eb1708b7d57dcde7478ca2f5063241b1ab567b2ba2ba28b9956b043da2b6c2", "supports": ["summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/PatrickStarL/gui-grounding-eval/d92b059caa0e0acf8db76f2164dd7be2617ccc72/ssp_eval/metrics.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:53:35.461491Z", "role": "primary", "sha256": "66fa5940f0c4f6265cb162cb0596402848dd2a49fd48b07ee950dc866b8730a5", "supports": ["summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/PatrickStarL/gui-grounding-eval/d92b059caa0e0acf8db76f2164dd7be2617ccc72/ssp_eval/geometry.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:53:38.931130Z", "role": "primary", "sha256": "f83bc0632a616e7b01c8a46b0e435733f0cc47b4a19f0c709b221eaee8a0db5f", "supports": ["summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/PatrickStarL/gui-grounding-eval/d92b059caa0e0acf8db76f2164dd7be2617ccc72/ssp_eval/runner.py"}], "id": "github:patrickstarl/gui-grounding-eval", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-20T11:51:38.123692Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/PatrickStarL/gui-grounding-eval"}], "name": "GUI Grounding Evaluation Harness", "notes": "This is an independent harness using ScreenSpot-Pro, not a new dataset or verified leaderboard entry. The read code maps points back to original-image coordinates and scores point-in-box hits; documentation limits comparisons by prompt, model, subset and provider behavior. No OS support, code license or external data/weight license was established in this bounded review. Model calls, tests and reported results were not executed or reproduced.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "ScreenSpot-Pro evaluation harness with explicit coordinate transforms, prompt and crop-based inference comparisons, and per-sample grounding accuracy and cost records.", "topics": ["grounding", "evaluation", "efficiency"], "updated_at": null, "url": "https://github.com/PatrickStarL/gui-grounding-eval"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["search queries", "click actions", "product option selection"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/README.md", "population": "crowd-sourced instruction collection", "qualifier": "exact", "unit": "instructions", "value": 12087}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/README.md", "population": "full product inventory", "qualifier": "approximately", "unit": "products", "value": 1180000}], "evaluation": "", "evidence_urls": ["https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/README.md"], "focus": ["instruction-conditioned product selection", "web navigation"], "notes": "This is a simulated shop. The default preview loads 1,000 products, while the full inventory is much larger; report the loaded subset and observation mode. Instructions and products are different populations. Detailed reward semantics are not normalized in this profile.", "observations": ["HTML pages", "simplified text views"], "setting": "interactive"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "79b7a90a1413a52deb53142b5bbd81ecf40f61743bd4601c2835e91375898ec0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "79b7a90a1413a52deb53142b5bbd81ecf40f61743bd4601c2835e91375898ec0", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:58Z", "role": "primary", "sha256": "8872dbf8660b00890b5e07ce1ea1f7a44fa7ae4e1857da56caba172b51dab3cb", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/LICENSE.md"}], "id": "github:princeton-nlp/webshop", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:58Z", "url": "https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/LICENSE.md"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/princeton-nlp/WebShop"}], "name": "WebShop", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/princeton-nlp/WebShop/HEAD/README.md", "target_id": "arxiv:2207.01206", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Simulated shopping environment for instruction-following web agents.", "topics": ["evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/princeton-nlp/WebShop"}
{"aliases": [], "authors": [], "details": {"availability": "Public simulator, task/judge and online-RL source, plus official release metadata for the companion data archive.", "benchmark_protocol": {"actions": ["tap", "type", "swipe", "back", "home", "drag", "complete"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "population": "publisher benchmark overview", "qualifier": "exact", "unit": "task templates", "value": 416}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "population": "publisher benchmark overview", "qualifier": "exact", "unit": "simulated apps", "value": 28}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/train.txt", "population": "bench_env/splits/train.txt", "qualifier": "exact", "unit": "task identifiers", "value": 160}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/test.txt", "population": "bench_env/splits/test.txt", "qualifier": "exact", "unit": "task identifiers", "value": 256}], "evaluation": "Deterministic task judges inspect state changes; reported metrics distinguish success, progress, false completion and unexpected side effects.", "evidence_urls": ["https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/train.txt", "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/test.txt"], "focus": ["mobile GUI task completion in simulation", "state verification and parallel training"], "notes": "The split-ID inventory is not a count of sampled trajectories, live Android apps or real user accounts. Extra runtime/data dependencies and third-party content terms remain distinct.", "observations": ["simulated mobile screenshots", "structured simulator state for judges"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE"], "reason": "Core source LICENSE; mobilegym-rl has a separate matching Apache-2.0 license and retains vendored notices."}, "data": {"evidence_urls": ["https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE-DATA"], "reason": "Limited to maintainers’ dataset compilation/curation and authored synthetic content; third-party user content remains subject to platform terms and its authors’ rights."}}, "release_tag": "v0.1.0", "repository_revision": "5fbd3f34be0e9780ae4b1e9d5089963ddbd05d38"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:19Z", "role": "primary", "sha256": "815294dd2aa4bf8264978c3bfedee85a3a2c835c511860c732314a4c282a974a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:18Z", "role": "primary", "sha256": "48df04a2fbf479f956d5c7f22d7006c22f591b8857cab8a1d64c196349045495", "supports": ["identity"], "url": "https://api.github.com/repos/Purewhiter/mobilegym"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:56Z", "role": "primary", "sha256": "68f1fa16c0999079bccc3aa1b6c3c22b72b570a5b971837c5026ea5408853375", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/Purewhiter/mobilegym/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:08Z", "role": "primary", "sha256": "cb1057b0a0a2ef7ab781414c168bb41f0f3da88afc042ce266d6063493561bde", "supports": ["release_status", "details.release_tag"], "url": "https://api.github.com/repos/Purewhiter/mobilegym/releases/tags/v0.1.0"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:09Z", "role": "primary", "sha256": "ad0c86a6cdde2caba8bd1f98b121803817d93718319a863b0b833da2760c23aa", "supports": ["release_status"], "url": "https://api.github.com/repos/Purewhiter/mobilegym/releases/tags/data-v0.1.0"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:59Z", "role": "primary", "sha256": "7618cd7de039c264759ca8ca0b9130c6de63a4536dcf436baedd4cfbf639073e", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:27:50Z", "role": "primary", "sha256": "58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/mobilegym-rl/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:59Z", "role": "primary", "sha256": "e222fa518c5003d425ee9b371e987c95ba10e52b54fea078571254ae6a8eaf23", "supports": ["licenses.data", "notes"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE-DATA"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:27:49Z", "role": "primary", "sha256": "0ecef460fecfd8aa91dcbb2cd4cca17b033f620f85a50d7f0b598ed19a27c0eb", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/train.txt"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:27:48Z", "role": "primary", "sha256": "a3d485015d480f4f1ebbd12dbbf02c25c0aa667d1afcf5f6fa35f0263a82c567", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/test.txt"}], "id": "github:purewhiter/mobilegym", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T11:18:19Z", "licenses": {"code": {"basis": "Core source LICENSE; mobilegym-rl has a separate matching Apache-2.0 license and retains vendored notices.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T11:22:59Z", "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE"}, "data": {"basis": "Limited to maintainers’ dataset compilation/curation and authored synthetic content; third-party user content remains subject to platform terms and its authors’ rights.", "declared": "CC-BY-NC-4.0", "retrieved_at": "2026-10-17T11:22:59Z", "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE-DATA"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Purewhiter/mobilegym"}], "name": "MobileGym", "notes": "The inspected code release is v0.1.0, with a separately listed data-v0.1.0 archive. The interaction target is a browser-hosted Android-like simulation; real-device transfer results are separate publisher-reported experiments. Source split manifests contain 160 training IDs and 256 test IDs with no duplicates or overlap. Platform/training code declares Apache-2.0. LICENSE-DATA applies CC-BY-NC-4.0 only to maintainer-created compilation, curation and synthetic content, and explicitly leaves third-party UGC/API terms and author rights separate. No runtime, archive or model was executed/downloaded.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "target_id": "arxiv:2605.26114", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A browser-hosted mobile-app simulator with programmable state, parameterized tasks and deterministic state-checking judges for parallel GUI-agent evaluation and online reinforcement learning.", "topics": ["evaluation", "mobile-navigation", "reinforcement-learning", "sandbox"], "updated_at": null, "url": "https://github.com/Purewhiter/mobilegym"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/RadishSoftware/mobster/738cecdd3eacb9df88d48efb9efa53507bb4aeb4/LICENSE"], "reason": "Complete root MIT license text inspected at the pinned commit. Applies to repository software; data, weights, external WebDriverAgent code and separately distributed products do not inherit this declaration."}}, "repository_revision": "738cecdd3eacb9df88d48efb9efa53507bb4aeb4"}, "environments": ["ios"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:57:18.994340+00:00", "role": "primary", "sha256": "dfffdc4b5850ed6bfe0bcfe705fa5918d61420f8c80df2e38f61b7d1fb3819ec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/RadishSoftware/mobster/738cecdd3eacb9df88d48efb9efa53507bb4aeb4/README.md"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:06:00.276978+00:00", "role": "primary", "sha256": "01f97a863952373bd10a1f849d79706344be63f5774fe7e0cfe9ba774e41d213", "supports": ["identity"], "url": "https://api.github.com/search/repositories?q=topic%3Amobile-agent+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-20&per_page=100&page=1&sort=updated&order=asc"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:54:17.260430+00:00", "role": "primary", "sha256": "fde31e2fdae142655f09fa14fb3114bf39a22cc1ff7994204a66622f5cbfe2c7", "supports": ["identity", "details.repository_revision"], "url": "https://api.github.com/repos/RadishSoftware/mobster/commits/main"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:56:37.457839+00:00", "role": "primary", "sha256": "8c2b9647b33a93b1a33e9256307e09b0f3e6aa62608019defa6dcdc5d7bff3dc", "supports": ["identity", "details.repository_revision", "notes"], "url": "https://api.github.com/repos/RadishSoftware/mobster/git/trees/81f2c1990c07a3aaf001a5c3eb4f37a457f0fb3d?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:57:25.539870+00:00", "role": "primary", "sha256": "f9ad743798f16491100ea692069e346f82bbdb5a5df91e4e01c1d6168f2513cf", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/RadishSoftware/mobster/738cecdd3eacb9df88d48efb9efa53507bb4aeb4/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T12:05:06.711971+00:00", "role": "primary", "sha256": "4296a186cc2c00b9de93ababae490dc99cba596d858eff195e40283d58d73a0a", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/RadishSoftware/mobster/738cecdd3eacb9df88d48efb9efa53507bb4aeb4/mobile_agent/drivers.py"}], "id": "github:radishsoftware/mobster", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-20T11:57:18.994340+00:00", "licenses": {"code": {"basis": "Complete root MIT license text inspected at the pinned commit. Applies to repository software; data, weights, external WebDriverAgent code and separately distributed products do not inherit this declaration.", "declared": "MIT", "retrieved_at": "2026-10-20T11:57:25.539870+00:00", "url": "https://raw.githubusercontent.com/RadishSoftware/mobster/738cecdd3eacb9df88d48efb9efa53507bb4aeb4/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/RadishSoftware/mobster"}], "name": "Mobster engine", "notes": "This entry covers the published engine, not the separate Mobster for Mac product. The documented reference example requires an already-serving WebDriverAgent runner for an iPhone and a model-service key; it performs a fixed read-only Settings task and declines approvals. The repository lists benchmark/evaluation tooling, but dated simulator subsets, real-device trials and earlier engine results are different evaluation populations. The README explicitly places several product modules outside this repository. Source inspection confirms WDA dispatch paths, not task success, security guarantees or reproducible performance; no phone, simulator, model or benchmark was run. MIT applies to the inspected repository software, not automatically to external WDA code, images or other data.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "iOS agent engine organized around an observe, decide, verify and act loop, using WebDriverAgent screen state and phone actions such as taps, typing, swipes and app switches.", "topics": ["mobile-navigation", "evaluation"], "updated_at": null, "url": "https://github.com/RadishSoftware/mobster"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "5559ca58f8ee06d2812c6cbc8c3e49ef112f4ef41fdcf4e95568b268bb24bf14", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["relations", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "5559ca58f8ee06d2812c6cbc8c3e49ef112f4ef41fdcf4e95568b268bb24bf14", "supports": ["notes"], "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "d7a04bf11ce24a7b79cd44b4a78dcc9971eaeb3b565821de437e0ecb7723e634", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/LICENSE"}], "id": "github:saagent/mcpworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/SAAgent/MCPWorld"}], "name": "MCPWorld", "notes": "The retained paper reports 201 tasks, whereas the retained repository README describes approximately 170. These are separate source descriptions; their release/version difference has not been established.", "published_at": null, "relations": [{"evidence_url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending", "target_id": "arxiv:2506.07672", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation environment combining computer interaction with tool access.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://github.com/SAAgent/MCPWorld"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/README.md", "https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/LICENSE"], "reason": "The retained README states that the project is MIT licensed, while the retained LICENSE contains Creative Commons Attribution-NonCommercial-NoDerivatives 4.0. These declarations conflict, so a settled code license is not asserted."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:54Z", "role": "primary", "sha256": "a9d33536fae38485eb6aac9698ae22ed3533510d43f783d6c0e9f9cf95d9c58f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "licenses.code", "notes"], "url": "https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:46Z", "role": "primary", "sha256": "f7f28b8c7a1af76b9874ca6d040e8b1eb6768fd1043106c9d315090ea96754e7", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:46Z", "role": "primary", "sha256": "f7f28b8c7a1af76b9874ca6d040e8b1eb6768fd1043106c9d315090ea96754e7", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/LICENSE"}], "id": "github:scaleapi/browser-art", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T10:59:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/scaleapi/browser-art"}], "name": "BrowserART", "notes": "The retained README and LICENSE contain conflicting code-license declarations. The catalog leaves the code-license value unknown pending an upstream clarification; it does not infer data or model permissions.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/README.md", "target_id": "arxiv:2410.13886", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Red-teaming benchmark and test suite from Scale AI evaluating whether refusal-trained language models can be jailbroken when deployed as browser agents.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/scaleapi/browser-art"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:54Z", "role": "primary", "sha256": "e712feb15886d5a2aba1051d419dc69dc6e58c9a86a9d2d566090bbbf67492fc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/segev-shlomov/ST-WebAgentBench/HEAD/README.md"}], "id": "github:segev-shlomov/st-webagentbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T10:59:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/segev-shlomov/ST-WebAgentBench"}], "name": "ST-WebAgentBench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/segev-shlomov/ST-WebAgentBench/HEAD/README.md", "target_id": "arxiv:2410.06703", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "BrowserGym-based benchmark for evaluating the safety and policy compliance of enterprise web agents across six trustworthiness risk dimensions.", "topics": ["safety-security", "evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/segev-shlomov/ST-WebAgentBench"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "ca312434db1db852f96fb49bbc0a61f6a93cffe6c0dd3aa2b7ac814681ccc19e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ServiceNow/AgentLab/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "b192c58991e8ff585cc574615d40e74185404d4b96c1109d423071ab1367344b", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ServiceNow/AgentLab/HEAD/LICENSE"}], "id": "github:servicenow/agentlab", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/ServiceNow/AgentLab/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ServiceNow/AgentLab"}], "name": "AgentLab", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ServiceNow/AgentLab/HEAD/README.md", "target_id": "arxiv:2412.05467", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Experiment framework for developing and evaluating browser agents.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/ServiceNow/AgentLab"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["element-ID actions", "coordinate mouse and keyboard actions", "tab/navigation/scroll operations", "Python/Playwright action mode"], "counts": [], "evaluation": "Gymnasium reset/step interface delegates reward, completion, messages and task info to each adapter’s validate() method. There is no universal BrowserGym success metric or shared task total; retain benchmark/adapter version and action subset for comparisons.", "evidence_urls": ["https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/README.md", "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/browsergym/core/src/browsergym/core/env.py", "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/docs/src/core/action_space.rst", "https://api.github.com/repos/ServiceNow/BrowserGym/git/trees/main?recursive=1"], "focus": ["shared browser-agent evaluation interface", "benchmark adapters with separate task definitions"], "notes": "A framework with interactive browser adapters and a static WebLINX evaluation integration, so the profile setting is mixed. README-listed integrations are not added as disjoint tasks or datasets. Core requires Playwright/Chromium; individual benchmarks have their own setup, credentials, licenses and validation rules. Apache-2.0 covers repository code only. No browser tasks or third-party project code were executed.", "observations": ["browser screenshot", "DOM snapshot", "accessibility tree", "tab/URL state", "previous action and its error", "chat messages and task goal"], "setting": "mixed"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "7b685115ef822989fac9093518d16c46f0202736d005eed56e7f9fc36b43f3ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:05Z", "role": "primary", "sha256": "867cc6efdfead356839fef8a26b492502b8b5e3d165d1b3c17d24aaeec268d63", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/browsergym/core/src/browsergym/core/env.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:11Z", "role": "primary", "sha256": "c4bbe6b7d0cb36a5d5effab1aa5568301cead97fe4015273e1b82c667da3823a", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/docs/src/core/action_space.rst"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:08Z", "role": "primary", "sha256": "418efbe8c65ad7ebab61d3a066339876c820a9c957506eb456b8078ab352f6bf", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/ServiceNow/BrowserGym/git/trees/main?recursive=1"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "b192c58991e8ff585cc574615d40e74185404d4b96c1109d423071ab1367344b", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/LICENSE"}], "id": "github:servicenow/browsergym", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ServiceNow/BrowserGym"}], "name": "BrowserGym", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ServiceNow/BrowserGym/HEAD/README.md", "target_id": "arxiv:2412.05467", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Unified interfaces for browser-agent tasks and benchmark environments.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/ServiceNow/BrowserGym"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["BrowserGym browser actions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/README.md", "population": "WorkArena-L1 atomic tasks", "qualifier": "exact", "unit": "task definitions", "value": 33}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/README.md", "population": "WorkArena-L1 documented instances", "qualifier": "exact", "unit": "task instances", "value": 19912}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/README.md", "population": "WorkArena++ compositional tasks", "qualifier": "exact", "unit": "task definitions", "value": 682}], "evaluation": "Task-specific checks on ServiceNow workflows; levels and task configurations define different evaluations.", "evidence_urls": ["https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/README.md", "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"], "focus": ["enterprise web workflows", "compositional tasks"], "notes": "Definitions and instantiated tasks are different units. ServiceNow instance access is gated. WorkArena-L1 and WorkArena++ should be reported separately.", "observations": ["multimodal browser observations"], "setting": "interactive"}, "leaderboard_url": "https://huggingface.co/spaces/ServiceNow/browsergym-leaderboard"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "834a48f0346036a0fc591a36f9eb6f513b44a2777a66afa0917f28fc63f29334", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "b192c58991e8ff585cc574615d40e74185404d4b96c1109d423071ab1367344b", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/LICENSE"}], "id": "github:servicenow/workarena", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ServiceNow/WorkArena"}], "name": "WorkArena", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/README.md", "target_id": "arxiv:2403.07718", "type": "documented_by"}, {"evidence_url": "https://raw.githubusercontent.com/ServiceNow/WorkArena/HEAD/README.md", "target_id": "arxiv:2407.05291", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Browser-based enterprise workflow tasks built around ServiceNow.", "topics": ["evaluation", "workflow"], "updated_at": null, "url": "https://github.com/ServiceNow/WorkArena"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "1a21631fb312cc7eac35547792611a84f13dcf756a3cf92209676ccdb92cbc6b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/showlab/videogui/HEAD/README.md"}], "id": "github:showlab/videogui", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/showlab/videogui"}], "name": "VideoGUI", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/showlab/videogui/HEAD/README.md", "target_id": "arxiv:2406.10227", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Benchmark for desktop GUI tasks derived from instructional videos.", "topics": ["evaluation", "video"], "updated_at": null, "url": "https://github.com/showlab/videogui"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "ea54ea82b86548ff38913ef328ec18152ace71a3fe9bb497bcabcaad24e6a678", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/showlab/WorldGUI/HEAD/README.md"}], "id": "github:showlab/worldgui", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/showlab/WorldGUI"}], "name": "WorldGUI", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/showlab/WorldGUI/HEAD/README.md", "target_id": "arxiv:2502.08047", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "611 desktop GUI tasks that agents must complete from varied starting points, shipped with instructional videos and project files, plus the self-reflective WorldGUI-Agent baseline.", "topics": ["evaluation", "robustness"], "updated_at": null, "url": "https://github.com/showlab/WorldGUI"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "abc848e7eb5181be22168dbc076446c5eb0f065fd57c7e10367284ca6e50bf5c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/shulin16/MMInA/HEAD/README.md"}], "id": "github:shulin16/mmina", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/shulin16/MMInA"}], "name": "MMInA", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/shulin16/MMInA/HEAD/README.md", "target_id": "arxiv:2404.09992", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Multimodal web tasks in a WebArena-based environment that chain from 1 to 10 hops across sites such as a shopping store and Wikipedia, testing long-chain reasoning.", "topics": ["evaluation", "long-horizon"], "updated_at": null, "url": "https://github.com/shulin16/MMInA"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:56Z", "role": "primary", "sha256": "5653db27f736c0c77afb63f81ca051bf31f1e1f0579ac77cf00dca4ea751c159", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/skyworkai/agent-studio/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:47Z", "role": "primary", "sha256": "8486a10c4393cee1c25392769ddd3b2d6c242d6ec7928e1414efff7dfb2f07ef", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/skyworkai/agent-studio/HEAD/LICENSE"}], "id": "github:skyworkai/agent-studio", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-18T10:59:56Z", "licenses": {"code": {"basis": "repository LICENSE header; GNU text does not say -only or -or-later; not a weights/data license", "declared": "AGPL-3.0", "retrieved_at": "2026-10-18T11:03:47Z", "url": "https://raw.githubusercontent.com/skyworkai/agent-studio/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/skyworkai/agent-studio"}], "name": "AgentStudio", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/skyworkai/agent-studio/HEAD/README.md", "target_id": "arxiv:2403.17918", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Toolkit and Ubuntu-based environment for building and evaluating generalist computer-use agents through GUI and API interaction.", "topics": ["sandbox", "evaluation", "grounding", "recording"], "updated_at": null, "url": "https://github.com/skyworkai/agent-studio"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/sqzhang-lazy/D-PoT/HEAD/README.md"], "reason": "No project-wide code license is established by the inspected README/API/tree; third-party code presence does not establish a license for this package."}}, "repository_tree_sha": "fc764f19802f4f59afa635bfeec5a50e4c3d5465", "repository_url": "https://github.com/sqzhang-lazy/D-PoT"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "c9c21102db512270079666531dbab71c1ab2ec2073add16be56f8ab3c6fe8abf", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.code"], "url": "https://raw.githubusercontent.com/sqzhang-lazy/D-PoT/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T15:37:02Z", "role": "primary", "sha256": "d62cfecbfccd120e53a7fc60e8d4b90b157e998cc5ddac467514c30c0cae1363", "supports": ["identity"], "url": "https://api.github.com/repos/sqzhang-lazy/D-PoT"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:25:24Z", "role": "primary", "sha256": "f75444a0252df5d6ec1ab70d805584cb490cdc38efa995aae55ffbde43ab5fb1", "supports": ["identity", "release_status", "details.repository_tree_sha"], "url": "https://api.github.com/repos/sqzhang-lazy/D-PoT/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "github:sqzhang-lazy/d-pot", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-18T04:27:14Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/sqzhang-lazy/D-PoT"}], "name": "D-PoT evaluation code", "notes": "README describes AITW test sampling and preprocessing reused from Auto-GUI, followed by prediction/action-extraction/evaluation scripts. This is not claimed as a ready-to-run interactive phone controller. No project-level license declaration appears in the inspected README/API/non-truncated tree; code, dataset and model rights remain unestablished here.", "published_at": null, "relations": [{"evidence_url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50", "target_id": "arxiv:2410.00467", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Offline prompting and evaluation code for Dynamic Planning of Thoughts on Android-in-the-Wild, including action extraction and comparison with reference actions.", "topics": ["planning", "action-prediction", "evaluation"], "updated_at": null, "url": "https://github.com/sqzhang-lazy/D-PoT"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["Python code", "application API calls"], "counts": [{"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50", "population": "original AppWorld benchmark (2024 paper)", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2407.18901", "unit": "tasks", "value": 750}, {"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50", "population": "original AppWorld environment (2024 paper)", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2407.18901", "unit": "applications", "value": 9}], "evaluation": "State-based unit tests check task outcomes and unintended changes.", "evidence_urls": ["https://raw.githubusercontent.com/stonybrooknlp/appworld/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"], "focus": ["interactive coding", "cross-app tool use"], "notes": "This is an adjacent API/coding benchmark, not a screenshot-only task suite. Preserve the named release; task counts here refer to the original paper.", "observations": ["API documentation", "execution feedback"], "setting": "interactive"}, "leaderboard_url": "https://appworld.dev/appworld/leaderboard"}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:45:03Z", "role": "primary", "sha256": "f4049141b6e520217395b7ccd9f27048c8e0f90331690dcbef300d2bc4ff7bdd", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/stonybrooknlp/appworld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:42Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/StonyBrookNLP/appworld/HEAD/LICENSE"}], "id": "github:stonybrooknlp/appworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:45:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T10:48:42Z", "url": "https://raw.githubusercontent.com/StonyBrookNLP/appworld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/StonyBrookNLP/appworld"}], "name": "AppWorld", "notes": "Primarily API and coding interaction; included as an adjacent comparison for GUI and hybrid agents.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/stonybrooknlp/appworld/HEAD/README.md", "target_id": "arxiv:2407.18901", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Simulated applications, user data, and state-based task checks for evaluating agents that interact through APIs and code.", "topics": ["evaluation", "tool-integration", "cross-app"], "updated_at": null, "url": "https://github.com/StonyBrookNLP/appworld"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["web browsing", "code execution", "coworker communication"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/TheAgentCompany/TheAgentCompany/HEAD/README.md", "population": "workspace images documented by the README", "qualifier": "exact", "unit": "task images", "value": 175}], "evaluation": "Task-specific evaluator scripts run within the prepared workplace environment.", "evidence_urls": ["https://raw.githubusercontent.com/TheAgentCompany/TheAgentCompany/HEAD/README.md"], "focus": ["professional workflows", "multi-tool tasks"], "notes": "Each task is packaged as a Docker image. Record task-image and server versions; the environment includes simulated coworkers and external application state.", "observations": [], "setting": "interactive"}, "leaderboard_url": "https://the-agent-company.com/#/leaderboard"}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:34Z", "role": "primary", "sha256": "d1646f67d49d38ef1e81bf7b5057dfc3d6f172b4c8a7413a2cc78741ccb08723", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/TheAgentCompany/TheAgentCompany/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:42Z", "role": "primary", "sha256": "2d6976a9b482bb14d9cd4f0d97ce6e14305496f1649b8667cb91d027c670930f", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/TheAgentCompany/TheAgentCompany/HEAD/LICENSE"}], "id": "github:theagentcompany/theagentcompany", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:34Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T10:48:42Z", "url": "https://raw.githubusercontent.com/TheAgentCompany/TheAgentCompany/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/TheAgentCompany/TheAgentCompany"}], "name": "TheAgentCompany", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/TheAgentCompany/TheAgentCompany/HEAD/README.md", "target_id": "arxiv:2412.14161", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Self-contained workplace environment for evaluating agents on professional tasks involving websites, files, code, and simulated coworkers.", "topics": ["evaluation", "workflow", "cross-app"], "updated_at": null, "url": "https://github.com/TheAgentCompany/TheAgentCompany"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web", "android", "cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:55Z", "role": "primary", "sha256": "d69da19edaed185ad3071a97db3bb3fcfbdb14c0b514b7d27794b39fbe8e2541", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/thu-ml/MLA-Trust/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:46Z", "role": "primary", "sha256": "240fb6820e7bcb0753617bcad6ffbd4a9ba83ca98e1e6aaa06c7b197b244bf16", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/thu-ml/MLA-Trust/HEAD/LICENSE"}], "id": "github:thu-ml/mla-trust", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T10:59:55Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-18T11:03:46Z", "url": "https://raw.githubusercontent.com/thu-ml/MLA-Trust/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/thu-ml/MLA-Trust"}], "name": "MLA-Trust", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/thu-ml/MLA-Trust/HEAD/README.md", "target_id": "arxiv:2506.01616", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Unified evaluation framework measuring truthfulness, controllability, safety, and privacy of multimodal LLM agents across interactive web and mobile GUI tasks.", "topics": ["safety-security", "evaluation", "robustness"], "updated_at": null, "url": "https://github.com/thu-ml/MLA-Trust"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://docs.google.com/spreadsheets/d/e/2PACX-1vRR3Wl7wsCgHpwUw1_eUXW_fptAPLL3FkhnW_rua0O1Ji_GIVrpTjY5LaKAhwO-WeARjnY_KNw0SYNJ/pubhtml"}, "environments": ["cross-platform", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "1b7d1f1cd7a74ac9497c639324ec342dd3f8769e9c0c4cd2cd92a4eb6740ac98", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/LICENSE"}], "id": "github:thudm/agentbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T10:48:41Z", "url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/AgentBench"}], "name": "AgentBench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/README.md", "target_id": "arxiv:2308.03688", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Agent evaluation suite with containerized tasks and a function-calling interface, including operating-system and shopping environments.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://github.com/THUDM/AgentBench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["Android operations through the agent framework"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/README.md", "population": "benchmark task collection", "qualifier": "exact", "unit": "tasks", "value": 138}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/README.md", "population": "benchmark application set", "qualifier": "exact", "unit": "applications", "value": 9}], "evaluation": "The evaluation pipeline produces success, subgoal success, reversed redundancy, and reasonable-operation metrics; result generation requires a configured judge model.", "evidence_urls": ["https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/README.md"], "focus": ["Android task completion", "subgoal completion", "operation efficiency"], "notes": "The benchmark uses predefined virtual-device images and apps chosen to work without network/login requirements. XML and SoM modes should be reported separately. These counts concern evaluation tasks, not the separately released Android Instruct training set.", "observations": ["UI XML", "set-of-marks views"], "setting": "interactive"}, "evaluation_type": "benchmark task evaluation", "leaderboard_url": "https://docs.google.com/spreadsheets/d/1Zv6mBfd4Ibt8mke24K6zAFrXe4AvEFaBLW3hpJqSgjw/edit?gid=0#gid=0", "metrics": ["task success"], "task_count": 138}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "340680a9a7833d103234b8367195deebd29bfec7e25ebc50c2dc04fc3627e55b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "340680a9a7833d103234b8367195deebd29bfec7e25ebc50c2dc04fc3627e55b", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:58Z", "role": "primary", "sha256": "1bbbee919161337e2456c587c6be809214406482cd9f1292a9eb62428f6961f3", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/LICENSE"}], "id": "github:thudm/android-lab", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:58Z", "url": "https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/Android-Lab"}], "name": "AndroidLab", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/Android-Lab/HEAD/README.md", "target_id": "arxiv:2410.24024", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Reproducible Android benchmark of 138 tasks across nine apps on predefined virtual devices, run via AVD on Mac or Docker on Linux, released with the Android Instruct training set.", "topics": ["evaluation", "mobile-navigation"], "updated_at": null, "url": "https://github.com/THUDM/Android-Lab"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform", "web", "android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "21afbc4086441dee131fe4b84986870f9b280413dd2a38df6e99406f1fc45586", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/THUDM/VisualAgentBench/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/THUDM/VisualAgentBench/HEAD/LICENSE"}], "id": "github:thudm/visualagentbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T10:48:41Z", "url": "https://raw.githubusercontent.com/THUDM/VisualAgentBench/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/VisualAgentBench"}], "name": "VisualAgentBench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/VisualAgentBench/HEAD/README.md", "target_id": "arxiv:2408.06327", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Evaluation framework spanning visual GUI, embodied, and design tasks, including mobile and WebArena-based environments.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/THUDM/VisualAgentBench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["OSWorld/PyAutoGUI desktop interaction"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/README.md", "population": "reported benchmark total; not a recount of unique released task IDs", "qualifier": "exact", "unit": "tasks as reported by the paper/README", "value": 150}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_misuse.json", "population": "released misuse task-ID list; before injection/goal expansion", "qualifier": "exact", "unit": "task IDs", "value": 50}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_injection.json", "population": "released injection task-ID list; before injection/goal expansion", "qualifier": "exact", "unit": "task IDs", "value": 10}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_misbehavior.json", "population": "released misbehavior task-ID list; before injection/goal expansion", "qualifier": "exact", "unit": "task IDs", "value": 50}], "evaluation": "Automated LLM judge scores both task accuracy and safety from saved execution traces. Keep judge model/prompt version and harm category with a result; the scripts separately compute predicted metrics. This is an interactive desktop safety benchmark, not a single offline grounding score.", "evidence_urls": ["https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/README.md", "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/run.py", "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_misuse.json", "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_injection.json", "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_misbehavior.json"], "focus": ["deliberate misuse", "prompt-injection safety", "unintended harmful behavior during normal tasks"], "notes": "README/paper reports 150 tasks, whereas retained misuse/injection/misbehavior lists contain 50/10/50 IDs. run.py expands injection vectors/goals, so the released ID inventory is not a direct recount of paper settings; the complete expanded total was not reconstructed. Judge agreement F1 values in the README measure agreement with human labels, not agent success. Experiments use OSWorld desktop apps, with VMware documented as tested and screenshot/a11y/SOM observation choices. Code Apache-2.0 does not establish VM/app/data rights. Publisher asks that the benchmark not be used directly as agent training data.", "observations": ["screenshot", "accessibility tree", "screenshot plus accessibility tree", "set-of-marks"], "setting": "interactive"}, "evaluation_type": "automated judgment of accuracy and safety", "metrics": ["task completion", "safety"], "task_count": 150, "task_unit": "tasks reported by the paper/README; not a recount of released task IDs"}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T09:36:33Z", "role": "primary", "sha256": "d001158f58ac37b01c8470224db1567a52feb8010877305f7966637cf7400706", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "details.benchmark_protocol", "notes"], "url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:23:14Z", "role": "primary", "sha256": "86dd59a252fd654fc3244268a75de37b2c7f97e16f6b1de8c4cf885458a422f7", "supports": ["details.benchmark_protocol", "notes"], "url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/run.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:04Z", "role": "primary", "sha256": "8ff6092a51cc924c5225ace01c011c6481a2461eb0a47b7360eddaf6619b898c", "supports": ["details.benchmark_protocol", "notes"], "url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_misuse.json"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:04Z", "role": "primary", "sha256": "b4c451236bb4ff36eb7e790dfa8cb517bffb245602bc441080314b13c8aba54f", "supports": ["details.benchmark_protocol", "notes"], "url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_injection.json"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:05Z", "role": "primary", "sha256": "8eb6bade72696c3c0b53160a05b006a773c9e553ab5e401418b54d9acc98c0f2", "supports": ["details.benchmark_protocol", "notes"], "url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/evaluation_examples/test_misbehavior.json"}, {"method": "license_header", "retrieved_at": "2026-10-16T09:39:34Z", "role": "primary", "sha256": "7c3ab1eeeb073d92238ecb65ffaebe58ab5922929b115a7c0b64eb4f9a9afc2f", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/LICENSE"}], "id": "github:tml-epfl/os-harm", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T09:36:33Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T09:39:34Z", "url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/tml-epfl/os-harm"}], "name": "OS-Harm", "notes": "The retained task-ID manifests have 50 misuse, 10 injection and 50 misbehavior IDs. Prompt-injection execution expands vectors/goals; the reported 150-task total must not be read as 150 unique released IDs.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/tml-epfl/os-harm/HEAD/README.md", "target_id": "arxiv:2506.14866", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Benchmark for safety of computer-use agents across misuse, prompt injection, and unintended model behavior.", "topics": ["safety-security", "evaluation"], "updated_at": null, "url": "https://github.com/tml-epfl/os-harm"}
{"aliases": [], "authors": [], "details": {"availability": "Public task definitions, evaluator and container orchestration source.", "benchmark_protocol": {"actions": ["Android GUI operations", "MCP calls when enabled", "agent-user interaction when enabled"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md", "population": "README benchmark overview", "qualifier": "exact", "unit": "tasks", "value": 201}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md", "population": "README benchmark overview", "qualifier": "exact", "unit": "mobile applications", "value": 20}], "evaluation": "The publisher documents text-answer matching, backend database verification, local-storage inspection and application callbacks.", "evidence_urls": ["https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md"], "focus": ["Android task completion", "agent-user interaction", "hybrid GUI and MCP use"], "notes": "Publisher-reported inventory. GUI-only, agent-user and MCP-augmented configurations have different dependencies and task eligibility; preserve enabled modes and step budget with results.", "observations": ["Android application interface", "user-agent interaction when enabled"], "setting": "interactive"}, "leaderboard_url": "https://tongyi-mai.github.io/MobileWorld/#leaderboard", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/LICENSE"], "reason": "Repository code license; self-hosted applications, model APIs and MCP services keep their own terms."}}, "repository_revision": "e7eee8981135ecbe39005a231edf63574defd79c"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:16Z", "role": "primary", "sha256": "fbc306b3414dfdb230171358efd6eb8d67573fbb94e11487a4a7e971e2abf0d9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:15Z", "role": "primary", "sha256": "266e1fe5b5b1a4c1540ab0fd5eba206395084ee283f0079f424930a156cf8e18", "supports": ["identity"], "url": "https://api.github.com/repos/Tongyi-MAI/MobileWorld"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:08Z", "role": "primary", "sha256": "14500ed2dfa14bb5c34211090f0c8827e4a16c3233e35a2d695c0d7261f9358b", "supports": ["release_status"], "url": "https://api.github.com/repos/Tongyi-MAI/MobileWorld/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:09Z", "role": "primary", "sha256": "b09fcaf4fbc265223c2479c410beef2b98c5e687c78ecedf3452aa12ffc465f1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/LICENSE"}], "id": "github:tongyi-mai/mobileworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T09:12:16Z", "licenses": {"code": {"basis": "Repository code license; self-hosted applications, model APIs and MCP services keep their own terms.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T09:14:09Z", "url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Tongyi-MAI/MobileWorld"}], "name": "MobileWorld (Tongyi-MAI)", "notes": "This is Tongyi-MAI/MobileWorld, distinct from jacklishufan/MobileWorld (MobileWorldBench). Public source includes task definitions, verification logic and environment orchestration. The README reports 201 tasks across 20 applications; this is not an independent count of the evolving task tree. Linux/KVM or WSL2/KVM is required by the documented host setup, with macOS support in progress. MCP and simulated-user settings can require separate external services.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md", "target_id": "arxiv:2512.19432", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A containerized Android evaluation environment with GUI-only, agent-user interaction and MCP-augmented tasks, plus checks on application and backend state.", "topics": ["evaluation", "mobile-navigation", "human-in-the-loop", "tool-integration"], "updated_at": null, "url": "https://github.com/Tongyi-MAI/MobileWorld"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "12830b2eeb3edea7af9953b27987f3467fa90803c641e786db09a33643c714d7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/trycua/cua/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c0779290c1d4783169aa3dbfb55feb505e563ef8a004bbf55298ceffcfbda8d9", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/trycua/cua/HEAD/LICENSE.md"}], "id": "github:trycua/cua", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/trycua/cua/HEAD/LICENSE.md"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/trycua/cua"}], "name": "Cua", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Monorepo behind Cua Spaces, which gives agents full macOS or Linux desktops, plus Cua Driver for operating native apps, Lume for Apple Silicon VMs, Cua Bench and the CUA-S1 decision models.", "topics": ["evaluation", "workflow"], "updated_at": null, "url": "https://github.com/trycua/cua"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "cfe9d8614b0b0558c37ba7c50371f6c47f119b914061ae815a9e3d796dd9ccfa", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:01:50Z", "role": "primary", "sha256": "7a4ca3daa6d011ff2d24d993191882d2b5545d8476c7eddf396581c0fef77496", "supports": ["identity"], "url": "https://api.github.com/repos/UbiquitousLearning/DroidCall"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "d0717d706359969376889eac73ea8764b659f41a2b4f3c1a409110267a912526", "supports": ["release_status"], "url": "https://api.github.com/repos/UbiquitousLearning/DroidCall/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:16Z", "role": "primary", "sha256": "2832cc5333f7166924a2823aa38a286d918b39b7fef2ca7af6521766c68a5303", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "08a6c5cfa655e1ecd774a41a9a815e845ad0afd47b289378bc1d6ee66c81dda9", "supports": ["relations"], "url": "https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:02:16Z", "role": "primary", "sha256": "2832cc5333f7166924a2823aa38a286d918b39b7fef2ca7af6521766c68a5303", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/LICENSE"}], "id": "github:ubiquitouslearning/droidcall", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:02:06Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T02:02:16Z", "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/UbiquitousLearning/DroidCall"}], "name": "DroidCall generation and evaluation code", "notes": "Source utilities are published under MIT. This component operates through Android intent functions and is marked adjacent to visual GUI agents. Training data is separately released under its own card declaration; no checkpoint release is inferred from the fine-tuning instructions.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md", "target_id": "arxiv:2412.00402", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Data-generation, fine-tuning and evaluation utilities for teaching small language models to invoke Android intents through predefined functions.", "topics": ["tool-integration", "data-generation", "evaluation"], "updated_at": null, "url": "https://github.com/UbiquitousLearning/DroidCall"}
{"aliases": ["github:eric-ai-lab/screen-point-and-read", "eric-ai-lab/Screen-Point-and-Read", "https://github.com/eric-ai-lab/Screen-Point-and-Read"], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "7567287ceee0ad2d3d5605abb44ac50463c1008f9b6309393695763cf2ce89bb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/eric-ai-lab/Screen-Point-and-Read/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:42Z", "role": "primary", "sha256": "4748fb7d92f7a3a2a4a93353db96f6a2df547e5ea9ae87afa811c5423e96f363", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/UCSB-AI/Screen-Point-and-Read/HEAD/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T14:14:01Z", "role": "primary", "sha256": "868e18717ff76355c6552a5629b15de9fed1add1262fa5801ecabd7a7fb333d4", "supports": ["aliases"], "url": "https://api.github.com/repos/eric-ai-lab/Screen-Point-and-Read"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T14:14:02Z", "role": "primary", "sha256": "868e18717ff76355c6552a5629b15de9fed1add1262fa5801ecabd7a7fb333d4", "supports": ["aliases"], "url": "https://api.github.com/repos/UCSB-AI/Screen-Point-and-Read"}], "id": "github:ucsb-ai/screen-point-and-read", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:43:11Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T10:48:42Z", "url": "https://raw.githubusercontent.com/UCSB-AI/Screen-Point-and-Read/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/UCSB-AI/Screen-Point-and-Read"}], "name": "Screen Point-and-Read / Tree-of-Lens", "notes": "The canonical repository now belongs to UCSB-AI. Its README lists additional setup documentation and evaluation scripts as unfinished.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/eric-ai-lab/Screen-Point-and-Read/HEAD/README.md", "target_id": "arxiv:2406.19263", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Screen-reading benchmark and Tree-of-Lens implementation for interpreting user-indicated regions and their layout context.", "topics": ["screen-understanding", "grounding", "evaluation"], "updated_at": null, "url": "https://github.com/UCSB-AI/Screen-Point-and-Read"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://unipat.ai/benchmarks/SaaS-Bench"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "8cfab6fca80090adb47c4ef634f89eedd49a6204dbdc2b5b8e0d6f04c32db289", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/UniPat-AI/SaaS-Bench/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "971246559df3f118e367cb299d3544f94e7c3d30db39f05c0c164589af4f3a88", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/UniPat-AI/SaaS-Bench/HEAD/LICENSE"}], "id": "github:unipat-ai/saas-bench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/UniPat-AI/SaaS-Bench/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/UniPat-AI/SaaS-Bench"}], "name": "SaaS-Bench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/UniPat-AI/SaaS-Bench/HEAD/README.md", "target_id": "arxiv:2605.15777", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "106 browser-driven business workflows across 23 self-hosted SaaS apps in six domains, scored by per-task verifiers that read each app's database and give partial credit.", "topics": ["evaluation", "workflow"], "updated_at": null, "url": "https://github.com/UniPat-AI/SaaS-Bench"}
{"aliases": ["VeriGUI", "VeriWeb"], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "e9a0866c8d9fa1759dedebd5ea3a4176009e51eb669f5877540524e53ac1077d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/VeriGUI-Team/VeriGUI/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/VeriGUI-Team/VeriGUI/HEAD/LICENSE"}], "id": "github:verigui-team/verigui", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/VeriGUI-Team/VeriGUI/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/VeriGUI-Team/VeriGUI"}], "name": "VeriWeb (formerly VeriGUI)", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Verifiable web information-seeking benchmark renamed from VeriGUI; desktop evaluation is described as future work.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/VeriGUI-Team/VeriGUI"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "262947081ddd17dbf5bb3d072574a19bbf30510c42c3ff42c9851a748fd37546", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/VisualWebBench/VisualWebBench/HEAD/README.md"}], "id": "github:visualwebbench/visualwebbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/VisualWebBench/VisualWebBench"}], "name": "VisualWebBench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/VisualWebBench/VisualWebBench/HEAD/README.md", "target_id": "arxiv:2404.05955", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation of multimodal models on visual web understanding tasks.", "topics": ["evaluation", "screen-understanding"], "updated_at": null, "url": "https://github.com/VisualWebBench/VisualWebBench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["browser element actions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md", "population": "full VWA suite", "qualifier": "exact", "unit": "tasks", "value": 910}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md", "population": "one task from each template type", "qualifier": "exact", "unit": "human trajectories", "value": 233}], "evaluation": "Execution-based completion checks for visually grounded web tasks.", "evidence_urls": ["https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md"], "focus": ["multimodal web task completion"], "notes": "The 233 demonstration trajectories are not the full 910-task evaluation set. Observation modes and captioning components differ across baseline agents.", "observations": ["screenshots with set-of-mark annotations", "caption-augmented accessibility tree"], "setting": "interactive"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "182529c2fce1e41f3a2162727205c54da8b3d2381b2273d726170ee9d8cf65cc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md"}], "id": "github:web-arena-x/visualwebarena", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/web-arena-x/visualwebarena"}], "name": "VisualWebArena", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md", "target_id": "arxiv:2401.13649", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Visually grounded web tasks extending the WebArena evaluation setting.", "topics": ["evaluation", "screen-understanding"], "updated_at": null, "url": "https://github.com/web-arena-x/visualwebarena"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["browser element actions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/web-arena-x/webarena/HEAD/README.md", "population": "canonical WebArena evaluation suite", "qualifier": "exact", "unit": "task examples", "value": 812}], "evaluation": "Functional correctness of completed web tasks.", "evidence_urls": ["https://raw.githubusercontent.com/web-arena-x/webarena/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"], "focus": ["web task completion"], "notes": "The canonical repository documents self-hosted websites and environment resets. Its README recommends BrowserGym/AgentLab for newer infrastructure; harness changes must be recorded.", "observations": ["HTML or accessibility tree"], "setting": "interactive"}, "leaderboard_url": "https://docs.google.com/spreadsheets/d/1M801lEpBbKSNwP-vDBkC_pF7LdyGU1f_ufZb_NWNBZQ/edit?usp=sharing"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "081c826497f9b7da4952771ee128c2215ec3532379db0fe9f7fcbd899d9c9287", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/web-arena-x/webarena/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/web-arena-x/webarena/HEAD/LICENSE"}], "id": "github:web-arena-x/webarena", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/web-arena-x/webarena/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/web-arena-x/webarena"}], "name": "WebArena", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/web-arena-x/webarena/HEAD/README.md", "target_id": "arxiv:2307.13854", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Self-hosted web environments and tasks for evaluating autonomous web agents.", "topics": ["evaluation", "web-navigation"], "updated_at": null, "url": "https://github.com/web-arena-x/webarena"}
{"aliases": [], "authors": [], "details": {"availability": "Public WebVLN training/evaluation and environment-wrapper source; external data/models unverified.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/WebVLN/WebVLN/HEAD/README.md", "https://api.github.com/repos/WebVLN/WebVLN/git/trees/main?recursive=1"], "reason": "README MIT badge is not backed by a project license file or component-scoped statement in the inspected release."}}, "repository_revision": "59aba284f5888fb7730e1ff07a7e682a1bd1968a"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "cb98feab09648856ab30d8eb739a29e837673cc5b4a6642b7ba3a8943ad71053", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "licenses.code"], "url": "https://raw.githubusercontent.com/WebVLN/WebVLN/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "112e67cc9f865b9640da3827e32eaba5a50f2feb8d85c20018dccbcafa6fd3e5", "supports": ["identity"], "url": "https://api.github.com/repos/WebVLN/WebVLN"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:05Z", "role": "primary", "sha256": "966adf983bae6052c271f7153c30776a29f7272cf55d05e0175713f02d1900fb", "supports": ["release_status", "details.repository_revision", "licenses.code"], "url": "https://api.github.com/repos/WebVLN/WebVLN/git/trees/main?recursive=1"}], "id": "github:webvln/webvln", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T12:30:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/WebVLN/WebVLN"}], "name": "WebVLN research implementation", "notes": "The released tree contains navigation models, an environment wrapper, scoring utilities and train/eval scripts. Full WebVLN-v1 data and pretrained files are linked through an external Drive folder; their complete contents and data/weight terms were not verified, so no separate benchmark download or checkpoint is approved here. The README MIT badge has no matching repository license file or explicit component scope; code permissions remain unestablished.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/WebVLN/WebVLN/HEAD/README.md", "target_id": "arxiv:2312.15820", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Training and evaluation code for question-guided website navigation that combines visual observations with web-specific HTML information.", "topics": ["web-navigation", "grounding", "evaluation"], "updated_at": null, "url": "https://github.com/WebVLN/WebVLN"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "813b7580cdae140917db573522b83a7986448025b31064442d34757292a6935c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/WebAgent-R1/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:19Z", "role": "primary", "sha256": "db1f71491c1d9482f403db30ea5599bd76e93d1db02c4ba8134d1a5cae5f0487", "supports": ["identity"], "url": "https://api.github.com/repos/weizhepei/WebAgent-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:39Z", "role": "primary", "sha256": "cc1a74037c4a2c4c094b11f6874ef67f88b24d14293ec09568ac6f378035c0b0", "supports": ["release_status"], "url": "https://api.github.com/repos/weizhepei/WebAgent-R1/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/LICENSE"}], "id": "github:weizhepei/webagent-r1", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:05:22Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T02:05:22Z", "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/weizhepei/WebAgent-R1"}], "name": "WebAgent-R1", "notes": "The implementation is inside the WebAgent-R1 subdirectory and includes training, evaluation and WebArena setup components. Root source licensing is Apache-2.0. No downloadable author checkpoint was established by the reviewed README, so this entry represents code.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/WebAgent-R1/README.md", "target_id": "arxiv:2505.16421", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Environment setup, multi-turn reinforcement-learning and evaluation tooling for browser agents interacting with WebArena.", "topics": ["reinforcement-learning", "web-navigation", "evaluation"], "updated_at": null, "url": "https://github.com/weizhepei/WebAgent-R1"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/LICENSE", "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/NOTICE"], "reason": "Record only the publisher's top-level Apache-2.0 declaration for Orbit-original code. NOTICE explicitly preserves per-component terms; neither bundled or external data nor model weights inherit this declaration. Its account of BrowserART licensing is not an independent resolution of upstream conflicting declarations."}}}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:41:22.467183+00:00", "role": "primary", "sha256": "b6743e3e9e5256d7afa8bb411baf246b60ba71bcfb84eda5a816fba3ddd7d78b", "supports": ["name", "kind", "summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/README.md"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T13:39:57.052202+00:00", "role": "primary", "sha256": "0436414b9bcc4307f3bcde476a544da6fe1dd5baa90939eed392942490fbfbf7", "supports": ["id", "url"], "url": "https://api.github.com/repos/wlanderson0/orbit"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:42:37.098639+00:00", "role": "primary", "sha256": "c47d046b129999da5e63b7723a3f3e2e0cc31c27ce74b45f71f8c6d0df1fc0d8", "supports": ["summary", "environments", "scope", "notes"], "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/orbit/scenarios/browser/browserart/tools.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T14:10:32.910533Z", "role": "primary", "sha256": "2cce76b22252a415d64c356b2062771e528b554dd83b6f88fca769f48fa8a826", "supports": ["summary", "environments", "scope", "notes"], "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/orbit/scenarios/desktop/osworld/tools.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T14:09:35.060985Z", "role": "primary", "sha256": "66515d7ee87eab5a2f42c26132d0bf3e2a3b9c006156d2378181cf91966a2903", "supports": ["notes"], "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/docs/benchmark-qualification.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T14:10:29.541892Z", "role": "primary", "sha256": "94c781028dc80b2a748d759fae2456f6a986da7487651a6bcbb179e8ad4fe023", "supports": ["notes"], "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/orbit/scenarios/browser/browserart/scorer.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:41:29.036076+00:00", "role": "primary", "sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T14:09:31.550310Z", "role": "primary", "sha256": "f68a64f775b1f329416f2fe294bb07b5e475893f77d1cdb5300bdb08acf95410", "supports": ["licenses.code", "licenses.data", "notes"], "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/NOTICE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:28.833839+00:00", "role": "primary", "sha256": "3ea667f4187ba601011e5ff00c4d63676b6d9903d12b80a0d7d5dd9920f00c65", "supports": ["relations"], "url": "https://arxiv.org/html/2609.33102v1"}], "id": "github:wlanderson0/orbit", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-20T13:41:22.467183+00:00", "licenses": {"code": {"basis": "Complete root Apache License 2.0 text for Orbit-original code. NOTICE preserves separate third-party component terms, including MIT-vendored modules; this is not a uniform license for the entire dependency/data bundle.", "declared": "Apache-2.0", "retrieved_at": "2026-10-20T13:41:29.036076+00:00", "url": "https://raw.githubusercontent.com/wlanderson0/orbit/588b3035f92450ec4f0496bb8b629daa205e84c7/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/wlanderson0/orbit"}], "name": "ORBIT framework", "notes": "Reviewed version 1.0.3 source at commit 588b3035, not an asserted reproduction of the paper. BrowserART tools return accessibility trees, including the screenshot-named refresh. Qualification is scenario-specific: the project documents browser-judge and desktop-scoring limits; its opt-in exact-send oracle covers one derived local task, not SMTP delivery. Runtime profiles and scoring policies are separate. Apache-2.0 is the top-level Orbit code declaration; third-party components and external data/environments retain separate terms. No installation, release archive or live evaluation was tested.", "published_at": null, "relations": [{"evidence_url": "https://arxiv.org/html/2609.33102v1", "target_id": "arxiv:2609.33102", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Inspect-based multi-agent evaluation framework with role, scheduling and communication controls, BrowserART accessibility-tree actions, and desktop mouse, keyboard and screenshot tools.", "topics": ["evaluation", "safety-security", "web-navigation", "observability"], "updated_at": null, "url": "https://github.com/wlanderson0/orbit"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["efficiency scores"], "counts": [], "evaluation": "The scorer consumes OSWorld results and a specified maximum step budget; the README distinguishes single-action and grouped-action WES+ and WES-.", "evidence_urls": ["https://raw.githubusercontent.com/WukLab/osworld-human/HEAD/README.md"], "focus": ["action efficiency relative to humans"], "notes": "Reference solutions are published for evaluation and expressly discouraged for training. No task count is inferred from file count.", "observations": ["OSWorld result trajectories", "human reference actions"], "setting": "offline"}}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "35c2f4dbed9d9bfc57f4c5a3f2fd70d313d6b463403b482c2fd4066d27b02cc1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/WukLab/osworld-human/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "21c86601da6e77a3326fbcd143e3c5d9a033cd59e6e13973250966dff6f51ae9", "supports": ["identity"], "url": "https://api.github.com/repos/WukLab/osworld-human"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:55Z", "role": "primary", "sha256": "c7b958d7a075adf5d282be08b68ad2b090568da342d4cb4b501ae808e54d0053", "supports": ["release_status"], "url": "https://api.github.com/repos/WukLab/osworld-human/git/trees/main?recursive=1"}], "id": "github:wuklab/osworld-human", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T03:59:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/WukLab/osworld-human"}], "name": "OSWorld-Human", "notes": "The repository publishes per-application reference JSON files and score.py. It explicitly states that these evaluation solutions should not be used for training. No root code or data-license declaration was found; no license is inherited from OSWorld, and the score script was not executed.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/WukLab/osworld-human/HEAD/README.md", "target_id": "arxiv:2506.16042", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Human reference trajectories and scoring code for measuring the action efficiency of computer-use agents on OSWorld tasks.", "topics": ["evaluation", "efficiency", "workflow"], "updated_at": null, "url": "https://github.com/WukLab/osworld-human"}
{"aliases": [], "authors": [], "details": {"availability": "Public federated-training/evaluation code and ungated annotation partitions; original images are separate dependencies.", "benchmark_protocol": {"actions": ["model predictions on mobile-agent examples"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/wwh0411/FedMABench/raw/main/README.md", "population": "released FedMABench partition families", "qualifier": "exact", "unit": "dataset families", "value": 6}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/wwh0411/FedMABench/HEAD/README.md", "population": "publisher benchmark overview", "qualifier": "exact", "unit": "heterogeneity types", "value": 3}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/wwh0411/FedMABench/HEAD/README.md", "population": "publisher benchmark overview", "qualifier": "exact", "unit": "federated learning algorithms", "value": 8}], "evaluation": "The provided evaluation workflow performs model inference over selected partitions and invokes test_swift_fed.py for accuracy scoring.", "evidence_urls": ["https://raw.githubusercontent.com/wwh0411/FedMABench/HEAD/README.md", "https://huggingface.co/datasets/wwh0411/FedMABench/raw/main/README.md"], "focus": ["mobile-agent federated training", "data heterogeneity"], "notes": "The dataset-family count does not count tasks, apps, screenshots or client instances. Code and data licenses are unknown; external image sources are required.", "observations": ["mobile-agent examples with externally sourced screenshots"], "setting": "offline"}, "repository_revision": "417d8b0006830f423c7d07c213fc57277d75d966"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:17Z", "role": "primary", "sha256": "b49e2b419918078ef9c68f3c9becc6c13d8fe530c165a36bdd61e370746eeea8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/wwh0411/FedMABench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:17Z", "role": "primary", "sha256": "3e1b2b643a263af129b3a4092a84efe47d3643f00509f7b9517da21a4aac0371", "supports": ["identity"], "url": "https://api.github.com/repos/wwh0411/FedMABench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:54Z", "role": "primary", "sha256": "4fb977a469cee3d5083da0b5bf2984fedc6b867146a90b20963985e08f9e9716", "supports": ["release_status", "details.repository_revision", "licenses.code"], "url": "https://api.github.com/repos/wwh0411/FedMABench/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:06Z", "role": "primary", "sha256": "df5379f4a9b350dad0c3d374d2f9a1d6eacd974b6d54750e2a14ecec1faec6e5", "supports": ["identity", "release_status"], "url": "https://huggingface.co/api/datasets/wwh0411/FedMABench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:06Z", "role": "primary", "sha256": "2874269995d8c9f5dabe266433ecf724bf54c27af599d899f3f0090b8ee37cdc", "supports": ["summary", "details.benchmark_protocol", "licenses.data"], "url": "https://huggingface.co/datasets/wwh0411/FedMABench/raw/main/README.md"}], "id": "github:wwh0411/fedmabench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T11:18:17Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/wwh0411/FedMABench"}], "name": "FedMABench", "notes": "The public Hub lists six dataset families of JSON/JSONL partition annotations. The repository explicitly says unpacked screenshots are not provided and must be reconstructed from original AndroidControl and AITW sources. This is offline training/evaluation infrastructure rather than a live Android task environment. No repository license file or dataset-license declaration was found in the inspected sources; neither is inferred from the underlying ms-swift project.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/wwh0411/FedMABench/HEAD/README.md", "target_id": "arxiv:2503.05143", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A federated mobile-agent training and evaluation suite that varies user-data heterogeneity and application distribution, with released partition annotations and inference/scoring scripts.", "topics": ["evaluation", "imitation-learning", "mobile-navigation"], "updated_at": null, "url": "https://github.com/wwh0411/FedMABench"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "fa70fd96c57a163404b77431f89dd16864ed104d7160e347742ff87587229961", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/X-LANCE/Mobile-Env/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:06Z", "role": "primary", "sha256": "58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/X-LANCE/Mobile-Env/HEAD/LICENSE"}], "id": "github:x-lance/mobile-env", "identifiers": {}, "kind": "infrastructure", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:06Z", "url": "https://raw.githubusercontent.com/X-LANCE/Mobile-Env/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/X-LANCE/Mobile-Env"}], "name": "Mobile Env", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/X-LANCE/Mobile-Env/HEAD/README.md", "target_id": "arxiv:2305.08144", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "AndroidEnv-based platform for building GUI agent benchmarks and training agents, where task definition files emit instructions, rewards and episode ends from screen text, view hierarchy or log events.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/X-LANCE/Mobile-Env"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["mouse and keyboard via PyAutoGUI"], "counts": [{"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/evaluation_examples/test_all.json", "population": "test_all.json", "qualifier": "exact", "unit": "task definitions", "value": 369}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/evaluation_examples/test_nogdrive.json", "population": "test_nogdrive.json", "qualifier": "exact", "unit": "task definitions", "value": 361}], "evaluation": "Task-specific execution-based evaluators inspect application or system state.", "evidence_urls": ["https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50", "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/evaluation_examples/test_all.json", "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/evaluation_examples/test_nogdrive.json"], "focus": ["task completion", "cross-application workflows"], "notes": "Counts were computed from unique task identifiers in the two inspected manifests. The subsets differ; record the manifest, repository revision, agent interface, and step budget with a score.", "observations": ["screenshots"], "setting": "interactive"}, "leaderboard_url": "https://os-world.github.io/"}, "environments": ["desktop", "linux", "windows", "macos"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "83b8684c2f4ac4f31996b8b4752195fcb6cc5429f89d92398177b9b721d787fc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:34:59Z", "role": "primary", "sha256": "9ebc5187cbd727ef26c24626820076b102fff812863c640a67c467fea9542ab5", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/evaluation_examples/test_all.json"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:34:59Z", "role": "primary", "sha256": "fcb9497e93a8986407345d3012c872c9c2fed253420730fbea64ccfcace67dbb", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/evaluation_examples/test_nogdrive.json"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "a8d1d88a0f9cce3cdbff73b03a80dc5e5d69e3ba2443946c268815a550c9e0ec", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/LICENSE"}], "id": "github:xlang-ai/osworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/xlang-ai/OSWorld"}], "name": "OSWorld", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/xlang-ai/OSWorld/HEAD/README.md", "target_id": "arxiv:2404.07972", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Execution-based evaluation of agents performing tasks in real desktop environments.", "topics": ["evaluation", "cross-app"], "updated_at": null, "url": "https://github.com/xlang-ai/OSWorld"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["click point or box center", "refusal represented by negative coordinates in the evaluator"], "counts": [], "evaluation": "The inspected Jedi runner maps predictions from resized-image coordinates back to the original image, then checks the predicted center against rectangle/polygon targets or the refusal rule. It reports overall and classification-group accuracy. This is offline grounding, not OSWorld task-execution success.", "evidence_urls": ["https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/README.md", "https://api.github.com/repos/xlang-ai/OSWorld-G/git/trees/main?recursive=1", "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/qwen25_vllm_osworld_g_jedi.py", "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/eval.py"], "focus": ["desktop GUI grounding", "infeasible-target refusal"], "notes": "The repository retains original OSWorld-G.json and refined OSWorld-G_refined.json instructions; treat them as alternative benchmark versions, not additional disjoint tasks. Refined instructions reduce requirements beyond grounding. The same repository also contains Jedi training/rendering material, which is excluded from benchmark counts. The standalone generic eval.py entry point is a stub; the named model runner supplies image_size to its geometry helper. No task count or successful execution is inferred from repository file totals. Apache-2.0 is recorded for code only.", "observations": ["screenshot", "instruction"], "setting": "offline"}}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:57Z", "role": "primary", "sha256": "e814851bc3016a2abf59358f99b80afd9c2bd5894fc34baddfb7e4d0aa03bd45", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:05Z", "role": "primary", "sha256": "48308bba4f26b02cb7b8fbbc433bb31ec81c494dabd3af437ddbb10b8d9ce667", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/xlang-ai/OSWorld-G/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:04Z", "role": "primary", "sha256": "64f64e0b847e0fa200aa6051e20b8035a436df00e2ef0d769688ab13fa4d3046", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/qwen25_vllm_osworld_g_jedi.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:09Z", "role": "primary", "sha256": "b85f5685ce6c200ea648471666177859693d7f40cf50c12d08765ae68580de4a", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/eval.py"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:57Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/LICENSE"}], "id": "github:xlang-ai/osworld-g", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:57Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:57Z", "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/xlang-ai/OSWorld-G"}], "name": "OSWorld-G", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/README.md", "target_id": "arxiv:2505.13227", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Computer-use grounding benchmark with original and refined instructions, released with the Jedi synthetic data pipeline, a 4-million-example Jedi dataset and Jedi-3B and Jedi-7B models.", "topics": ["evaluation", "grounding"], "updated_at": null, "url": "https://github.com/xlang-ai/OSWorld-G"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["PyAutoGUI mouse and keyboard actions"], "counts": [{"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/evaluation_examples/test_all.json", "population": "evaluation_examples/test_all.json", "qualifier": "exact", "unit": "task definitions", "value": 494}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/README.md", "population": "README task-suite coverage", "qualifier": "exact", "unit": "applications", "value": 20}], "evaluation": "Task-specific evaluation functions are invoked after agent execution.", "evidence_urls": ["https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/README.md", "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/evaluation_examples/test_all.json"], "focus": ["data-science workflows", "multi-application task completion"], "notes": "Counted unique task IDs in the full manifest. Verbose versus abstract instructions and account versus non-account tasks form different subsets. The Ubuntu VM and external-account requirements are part of the protocol; no accounts or environments were provisioned.", "observations": ["screenshots", "accessibility trees", "set-of-marks views"], "setting": "interactive"}}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "e3edf30ab45da16382b46d8c542607c1bbbfcf077b6df57a2decf938fc84ff0f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "e3edf30ab45da16382b46d8c542607c1bbbfcf077b6df57a2decf938fc84ff0f", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:06:30Z", "role": "primary", "sha256": "338dd38ba7047b0cf6dbc68f3f7b9374f962708fa2dadcc8d126c232dfc31746", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/evaluation_examples/test_all.json"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "a8d1d88a0f9cce3cdbff73b03a80dc5e5d69e3ba2443946c268815a550c9e0ec", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/LICENSE"}], "id": "github:xlang-ai/spider2-v", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/xlang-ai/Spider2-V"}], "name": "Spider2-V", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/xlang-ai/Spider2-V/HEAD/README.md", "target_id": "arxiv:2407.10956", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Visual agent evaluation on data-science and engineering workflows.", "topics": ["evaluation", "workflow"], "updated_at": null, "url": "https://github.com/xlang-ai/Spider2-V"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "c984f114df33af953a97d22e14e14246f8a98e0866edcaf283900b309b71d8d1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/yjyddq/RiOSWorld/HEAD/README.md"}], "id": "github:yjyddq/riosworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/yjyddq/RiOSWorld"}], "name": "RiOSWorld", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/yjyddq/RiOSWorld/HEAD/README.md", "target_id": "arxiv:2506.00618", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Measures risky behavior of multimodal computer-use agents in an Ubuntu virtual machine, with risk scenarios such as phishing emails and released evaluation trajectories.", "topics": ["evaluation"], "updated_at": null, "url": "https://github.com/yjyddq/RiOSWorld"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T13:17:23Z", "role": "primary", "sha256": "fc3009b355e46549c5128c933bb333aa37b1db5ac8ea8f2d0f300848d8a62200", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ZJU-ACES-ISE/UIDefectMocker/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T13:29:25Z", "role": "primary", "sha256": "2b75b184f28b204d8e2c7b45a341e4583202cf4c9da8964f108770929d1193ba", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ZJU-ACES-ISE/UIDefectMocker/HEAD/LICENSE"}], "id": "github:zju-aces-ise/uidefectmocker", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-16T13:17:23Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T13:29:25Z", "url": "https://raw.githubusercontent.com/ZJU-ACES-ISE/UIDefectMocker/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ZJU-ACES-ISE/UIDefectMocker"}], "name": "UI Defect Mocker", "notes": "This entry describes the released screenshot-defect tool, separately from the under-construction ChatUITest project hub.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Synthesizes visual UI defects from mobile screenshots and Android layout metadata to support GUI-testing experiments.", "topics": ["data-generation", "evaluation", "screen-understanding"], "updated_at": null, "url": "https://github.com/ZJU-ACES-ISE/UIDefectMocker"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "3f0aaefce32210dfba428fcf9e609e04e75d4d1520d24a3a6377611a9b83fa6d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "af69329b4667040c498cc87dfcc165db1f5f9452fe2ad3024ad7eef63d467666", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/LICENSE"}], "id": "github:zju-real/clawgui", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ZJU-REAL/ClawGUI"}], "name": "ClawGUI", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/README.md", "target_id": "arxiv:2604.11784", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Framework for GUI agent training, evaluation, and deployment.", "topics": ["reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://github.com/ZJU-REAL/ClawGUI"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["target localization", "identification of impossible requests"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/likaixin/ScreenSpot-v2-variants/raw/main/README.md", "population": "instruction-variant design", "qualifier": "exact", "unit": "instruction types", "value": 4}], "evaluation": "Publisher supplies original, action, description and negative instruction types; no common end-to-end success metric is asserted.", "evidence_urls": ["https://huggingface.co/datasets/likaixin/ScreenSpot-v2-variants/raw/main/README.md", "https://huggingface.co/api/datasets/likaixin/ScreenSpot-v2-variants"], "focus": ["instruction-form robustness", "negative grounding requests"], "notes": "Reuses ScreenSpot-v2 imagery. Image-file count is not treated as a count of independent screenshots or tasks.", "observations": ["screenshots", "instructions"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/likaixin/ScreenSpot-v2-variants/tree/main", "gated": false, "hub_id": "likaixin/ScreenSpot-v2-variants"}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "7fb2bc9499d8a2b37b663c116af49d11bba493e6767d76a5305fad93257bebf2", "supports": ["name", "kind", "summary", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-v2-variants/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "dea98efa75e063eee2d6bfb5d46da94b41df3d779e2fb0bdaea7866534b5c469", "supports": ["identity", "release_status", "details", "environments"], "url": "https://huggingface.co/api/datasets/likaixin/ScreenSpot-v2-variants"}], "id": "url:17729e1c532ba7043404", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T04:07:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-v2-variants"}], "name": "ScreenSpot-v2 instruction variants", "notes": "Public image files and desktop/mobile/web annotation JSON files are listed. This is a changed instruction set over an existing screenshot collection, not an independent new screenshot corpus. No data license or precise example count is declared in the card; no paper relationship is inferred from its generic project/PDF links.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Instruction variants over ScreenSpot-v2 screenshots that distinguish original commands, explicit actions, target descriptions and impossible operations.", "topics": ["grounding", "evaluation", "robustness"], "updated_at": null, "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-v2-variants"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["mobile UI actions", "clarification responses"], "counts": [], "evaluation": "Paper defines multi-path and noise/ambiguity evaluations; the public card identifies which files are full common-task sets and which are samples.", "evidence_urls": ["https://huggingface.co/datasets/xwk123/MobileBench-v2/raw/main/README.md", "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"], "focus": ["offline multi-path evaluation", "noisy interfaces", "ambiguous instructions"], "notes": "Release completeness differs by split. No full-benchmark sample total is inferred from the archive manifest.", "observations": ["mobile screenshots", "XML UI state", "instructions"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/xwk123/MobileBench-v2/tree/main", "gated": false, "hub_id": "xwk123/MobileBench-v2", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/xwk123/MobileBench-v2/raw/main/README.md"], "reason": "The same first-party card has an Apache-2.0 metadata header but explicitly assigns Apache-2.0 to code and CC-BY-NC-SA-4.0 to the dataset in its License section. Preserve both declarations and leave a single data license unresolved instead of displaying the permissive header as the data permission."}}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:33Z", "role": "primary", "sha256": "13424be795c7b75b8c2195469c1252a7b3ec633e8bb31f1f6d7e44476611cbe9", "supports": ["name", "kind", "summary", "topics", "scope", "relations", "details", "licenses.data"], "url": "https://huggingface.co/datasets/xwk123/MobileBench-v2/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:33Z", "role": "primary", "sha256": "247b1766b8cbdb2943c4602d2667f9f06dec8b84f0d24457080d499177475f22", "supports": ["identity", "release_status", "details"], "url": "https://huggingface.co/api/datasets/xwk123/MobileBench-v2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["relations", "summary"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:26Z", "role": "primary", "sha256": "356a8dca28fa3dc46bc05b13eaf04a3db273e63c06be4fd15d0e87e0343c0824", "supports": ["environments", "notes"], "url": "https://huggingface.co/api/datasets/xwk123/Mobile3M"}], "id": "url:1d5b85321b98765141fc", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T04:01:33Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/xwk123/MobileBench-v2"}], "name": "Mobile-Bench-v2 released data", "notes": "The card distinguishes complete common-simple/common-complex JSON files from sampled noisy and ambiguous subsets; common tasks also require the separately cataloged Mobile3M dataset. Card metadata says Apache-2.0, while the prose explicitly assigns Apache-2.0 to code and CC-BY-NC-SA-4.0 to data. Do not apply the permissive header to dataset payloads, and do not describe every announced split as fully released.", "published_at": null, "relations": [{"evidence_url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50", "target_id": "arxiv:2505.11891", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Offline mobile-agent task data spanning common tasks and released noisy, contaminated and ambiguous-instruction subsets.", "topics": ["evaluation", "mobile-navigation", "robustness"], "updated_at": null, "url": "https://huggingface.co/datasets/xwk123/MobileBench-v2"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": [], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md", "population": "Hub test split at revision b8eca829", "qualifier": "exact", "unit": "task-definition records", "value": 25}], "evaluation": "The published task schema contains rubrics and human_reference_operations. No automated scoring protocol or single numeric agent metric is established by the inspected card; execution uses separately configured Android tooling.", "evidence_urls": ["https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md", "https://huggingface.co/api/datasets/mikewang/mobile_eval_e", "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/Mobile-Agent-E/README.md", "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"], "focus": ["long-horizon multi-app Android task execution"], "notes": "Features are task_id, instruction, type, apps, scenario, rubrics and human-reference operation strings. A task record is not an action count or recorded screenshot trajectory. Public Parquet inventory was checked but no sample rows were downloaded; companion README limits the tested agent platform to Android.", "observations": ["task instructions", "app/scenario metadata"], "setting": "interactive"}, "disabled": false, "download_url": "https://huggingface.co/datasets/mikewang/mobile_eval_e/tree/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83", "gated": false, "hub_id": "mikewang/mobile_eval_e", "hub_revision": "b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md"], "reason": "No data-license declaration is present in the inspected task card; companion code MIT does not establish a dataset grant."}}, "repository_url": "https://github.com/X-PLUG/MobileAgent/tree/main/Mobile-Agent-E"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:15Z", "role": "primary", "sha256": "f8515fbce73ea0ef3d1d558d4567bb949244e6384e61c956ba3ddbc27294f0ae", "supports": ["name", "kind", "environments", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/Mobile-Agent-E/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:29Z", "role": "primary", "sha256": "f9b10eada9e0a64e26f1a7d477768f218b807103199855d893b3ab07feddc755", "supports": ["identity", "release_status", "details.hub_id", "details.hub_revision", "details.gated", "details.disabled", "details.download_url", "details.benchmark_protocol"], "url": "https://huggingface.co/api/datasets/mikewang/mobile_eval_e"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:15Z", "role": "primary", "sha256": "9fdf43753fbfac9868d8bec57fa5601db3e03ed5c30fb981f9da88f6b75ffcc5", "supports": ["summary", "details.benchmark_protocol", "licenses.data"], "url": "https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["summary", "topics", "details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "url:342b081c3668a8bbe155", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T04:27:15Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/mikewang/mobile_eval_e"}], "name": "Mobile-Eval-E task definitions", "notes": "The public Hub manifest contains one test Parquet file. This resource is the benchmark’s task-definition package, not a bundled phone environment or agent. Card metadata supplies rubrics/reference-operation fields but no data-license declaration or verified automated scorer; the code repository’s MIT license is not transferred to these records.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/Mobile-Agent-E/README.md", "target_id": "arxiv:2501.11733", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Task definitions for Mobile-Agent-E’s long-horizon, multi-app Android evaluation, packaged as a 25-record test split with rubrics and human-reference operation strings.", "topics": ["evaluation", "long-horizon", "cross-app"], "updated_at": null, "url": "https://huggingface.co/datasets/mikewang/mobile_eval_e"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["browser interaction"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md", "population": "main V1/V2 suite (shared identities)", "qualifier": "exact", "unit": "task identities", "value": 609}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md", "population": "refusal suite", "qualifier": "exact", "unit": "tasks", "value": 111}], "evaluation": "Task-specific precomputed rubrics; refusal tasks are a separate evaluation population.", "evidence_urls": ["https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"], "focus": ["live web task completion", "unsafe-request refusal"], "notes": "V2 refreshes task strings and rubrics while preserving V1 task IDs. Counts are task definitions, not independently verified runnable website states.", "observations": ["web pages"], "setting": "interactive"}, "disabled": false, "download_url": "https://huggingface.co/datasets/microsoft/WebTailBench/tree/main", "gated": false, "hub_id": "microsoft/WebTailBench", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"], "reason": "The benchmark's Hub dataset card declares MIT in its metadata header (license: mit)."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "c4cfc361d1d47d7855269af4110483363964e23f9c65244a1182830eccccbb7f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.data"], "url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "93469a20d333219dde89a225891a54ee716120d559eb8ac9f69a1a7ee84f57a3", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/microsoft/WebTailBench"}], "id": "url:3b6dfd52f98bfe76bc7f", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T02:42:32Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:42:32Z", "url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/microsoft/WebTailBench"}], "name": "WebTailBench", "notes": "Public TSV files contain V1, V2 and refusal task definitions. The card declares MIT for the dataset. V1 and V2 share the same 609 task identities and must not be summed as distinct tasks. The May 2026 refresh supersedes calendar-bound task wording; live-site availability is not established by this metadata review.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md", "target_id": "arxiv:2511.19663", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Live-web task definitions and rubrics for evaluating computer-use agents, with preserved original tasks, a refreshed V2 task suite and a separate refusal set.", "topics": ["evaluation", "web-navigation", "safety-security"], "updated_at": null, "url": "https://huggingface.co/datasets/microsoft/WebTailBench"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data/tree/main", "gated": false, "hub_id": "KDEGroup/UI-AGILE-Data"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:43:19Z", "role": "primary", "sha256": "ce35413bc1ef28338bc6fd9f132fa271a4f5e0bf23cecd7346f54adf0a964fd7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:43:19Z", "role": "primary", "sha256": "7782422de88830a497d9eb530bd2f0fb6283904cbb25dafcf2ff4d03358a377a", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/KDEGroup/UI-AGILE-Data"}], "id": "url:41abb1b7206689cf5057", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:43:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data"}], "name": "UI-AGILE-Data", "notes": "Public JSON, image archives and Parquet evaluation files are present. The card has no data-license declaration; the model’s MIT declaration is not inherited. The publisher describes fixes to AndroidControl action parameters, but those corrections were not independently reproduced or applied to upstream resources.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data/raw/main/README.md", "target_id": "arxiv:2507.22025", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "UI-AGILE training annotations and screenshots together with converted GUI-grounding and AndroidControl evaluation files.", "topics": ["grounding", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [], "evidence_urls": ["https://huggingface.co/api/datasets/OS-Copilot/ScienceBoard-Traj", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md"], "formats": ["ZIP"], "modalities": ["screenshots", "accessibility trees", "text"], "notes": "Observation modes follow publisher archive labels rather than inspected record schemas. No count or training-use permission is inferred.", "splits": [], "supervision": ["evaluation rollout logs"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/OS-Copilot/ScienceBoard-Traj/tree/main", "gated": false, "hub_id": "OS-Copilot/ScienceBoard-Traj", "languages": [], "modalities": ["image", "text"], "size_category_reported": null}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "0a97233bd9132fdddc6ddb26fd3cf9f2af06ae086429f367a61cfab0c16c4f44", "supports": ["name", "kind", "identity", "release_status", "details"], "url": "https://huggingface.co/api/datasets/OS-Copilot/ScienceBoard-Traj"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "4c3fe93aa740dbf4a279a246ddd836f2590a2a36f39b820abe166e38cb484577", "supports": ["summary", "environments", "relations", "topics", "scope"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "8ee436f67e708bb917992764bc60a017c9db2edb21bdb99356b35c69f73724d7", "supports": ["environments", "notes"], "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "0a97233bd9132fdddc6ddb26fd3cf9f2af06ae086429f367a61cfab0c16c4f44", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/OS-Copilot/ScienceBoard-Traj"}], "id": "url:4784f63cd7117f64a2a3", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T04:07:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/OS-Copilot/ScienceBoard-Traj"}], "name": "ScienceBoard evaluation trajectories", "notes": "The manifest exposes six ZIP archives but no README card or data-license declaration. The official project announces evaluation-result trajectories; this record does not label them as a training corpus or infer a trajectory count from archive names. Internal schemas and complete payload contents were not inspected.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md", "target_id": "arxiv:2505.19897", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Archives of ScienceBoard agent evaluation runs, separated by agent configuration and screenshot, accessibility-tree or set-of-marks observation mode.", "topics": ["evaluation", "science", "workflow"], "updated_at": null, "url": "https://huggingface.co/datasets/OS-Copilot/ScienceBoard-Traj"}
{"aliases": [], "authors": ["Tianlin Shi", "Andrej Karpathy", "Linxi Fan", "Jonathan Hernandez", "Percy Liang"], "details": {"publication": {"authors": ["Tianlin Shi", "Andrej Karpathy", "Linxi Fan", "Jonathan Hernandez", "Percy Liang"], "date_basis": "Publisher citation_publication_date metadata explicitly states 2017/07/17; this is separate from the conference interval.", "date_parts": [2017, 7, 17], "doi": null, "event_dates": "06--11 Aug (conference interval reported in the publisher BibTeX month field)", "evidence_urls": ["https://proceedings.mlr.press/v70/shi17a.html", "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"], "identity_basis": "No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.", "issue": null, "notes": "Publisher citation uses Tianlin Shi and Linxi Fan; the PDF also prints the nicknames Tim and Jim. The ordered publisher author names are retained. DOI not supplied in the retained publisher citation. Publisher metadata supplies the publication date separately from the conference interval.", "pages": "3135-3144", "pdf_url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf", "publisher": "PMLR", "series": "Proceedings of Machine Learning Research", "title": "World of Bits: An Open-Domain Platform for Web-Based Agents", "type": "conference-paper", "url": "https://proceedings.mlr.press/v70/shi17a.html", "venue": "Proceedings of the 34th International Conference on Machine Learning", "volume": "70"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:21Z", "role": "primary", "sha256": "749d2a150e4c2bcde1b219a0ecb86c95546134a057892544e8495b5e45323c44", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.event_dates", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.publisher", "details.publication.series", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.mlr.press/v70/shi17a.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:54:39Z", "role": "primary", "sha256": "087008efb9d056df5fd9fbc724413d6a3f20cb14c7f1597d00bacac53d8342d6", "supports": ["details.publication.notes", "details.publication.pdf_url"], "url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:21Z", "role": "primary", "sha256": "749d2a150e4c2bcde1b219a0ecb86c95546134a057892544e8495b5e45323c44", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.series", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume", "kind", "name", "scope", "summary", "topics"], "url": "https://proceedings.mlr.press/v70/shi17a.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:54:39Z", "role": "primary", "sha256": "087008efb9d056df5fd9fbc724413d6a3f20cb14c7f1597d00bacac53d8342d6", "supports": ["details.publication.notes", "details.publication.pdf_url", "environments"], "url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"}], "id": "url:52a3261eb6a19b402e88", "identifiers": {}, "kind": "paper", "last_verified_at": "2026-10-17T09:53:21Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://proceedings.mlr.press/v70/shi17a.html"}, {"rel": "publication_pdf", "url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"}], "name": "World of Bits: An Open-Domain Platform for Web-Based Agents", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. Publisher citation uses Tianlin Shi and Linxi Fan; the PDF also prints the nicknames Tim and Jim. The ordered publisher author names are retained.", "published_at": "2017-07-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces World of Bits, a web-task platform combining human demonstrations, keyboard and mouse actions, and cached HTTP traffic for repeatable task environments. It evaluates agents trained through behavioral cloning and reinforcement learning.", "topics": ["web-navigation", "reinforcement-learning", "evaluation", "data-generation"], "updated_at": null, "url": "https://proceedings.mlr.press/v70/shi17a.html"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["click target or bounding box"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/Hcompany/WebClick/raw/main/README.md", "population": "test split", "qualifier": "exact", "unit": "examples", "value": 1639}], "evaluation": "Assess localization against annotated targets; agentbrowse, humanbrowse and calendars are distinct source buckets.", "evidence_urls": ["https://huggingface.co/datasets/Hcompany/WebClick/raw/main/README.md"], "focus": ["instruction-conditioned web localization"], "notes": "The card reports more than 100 websites but does not provide a precise website total. Grounding accuracy does not establish task completion.", "observations": ["web screenshots", "instructions"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/Hcompany/WebClick/tree/main", "gated": false, "hub_id": "Hcompany/WebClick"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:33Z", "role": "primary", "sha256": "958bbf0eb9f61afd6c576673fbe6f7d84e6c4b8a9d4d8c4fc0e1224b5d8da30b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/Hcompany/WebClick/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:33Z", "role": "primary", "sha256": "c9576957a3cb14785bf274691b50f90b04567a442ed60ee8c39f0f73108359f4", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/Hcompany/WebClick"}], "id": "url:5e2bd1b3398974b79c9c", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T04:01:33Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/Hcompany/WebClick"}], "name": "WebClick", "notes": "Four public test Parquet shards are present. The dataset card declares Apache-2.0. This is an offline localization test, not an end-to-end web-task success benchmark; reported screenshots and samples are not summed as different populations.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/Hcompany/WebClick/raw/main/README.md", "target_id": "arxiv:2506.02865", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A web-interface localization benchmark with screenshot instructions and target boxes, including agent browsing, human browsing and calendar cases.", "topics": ["grounding", "evaluation", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/datasets/Hcompany/WebClick"}
{"aliases": [], "authors": ["Tsung-Hsiang Chang", "Tom Yeh", "Robert C. Miller"], "details": {"publication": {"authors": ["Tsung-Hsiang Chang", "Tom Yeh", "Robert C. Miller"], "date_basis": "Supplied month precision: 2010-04. Date Issued: April 2010", "date_parts": [2010, 4], "doi": "10.1145/1753326.1753555", "event_dates": null, "evidence_urls": ["https://dspace.mit.edu/entities/publication/6a6097ba-1d97-43cf-bc70-04b5af02efd5"], "identity_basis": "No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.", "issue": null, "notes": "Distinct visual-testing contribution from the 2009 Sikuli paper. The institutional abstract and published-version citation supply the retained evidence. Exact publication day and operating-system coverage are not established in the accessible institutional abstract.", "pages": "1535-1544", "pdf_url": null, "publisher": null, "series": null, "title": "GUI Testing Using Computer Vision", "type": "conference-paper", "url": "https://doi.org/10.1145/1753326.1753555", "venue": "Proceedings of the 28th International Conference on Human Factors in Computing Systems (CHI '10)", "volume": null}}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:21Z", "role": "primary", "sha256": "dbace77c54d1b4ed5dcf825dcb0f4525c32e33ac885e70312e8bb4a731a94e80", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://dspace.mit.edu/entities/publication/6a6097ba-1d97-43cf-bc70-04b5af02efd5"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:21Z", "role": "primary", "sha256": "dbace77c54d1b4ed5dcf825dcb0f4525c32e33ac885e70312e8bb4a731a94e80", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "environments", "kind", "name", "scope", "summary", "topics"], "url": "https://dspace.mit.edu/entities/publication/6a6097ba-1d97-43cf-bc70-04b5af02efd5"}], "id": "url:6262c757080f8838e62c", "identifiers": {"doi": "10.1145/1753326.1753555"}, "kind": "paper", "last_verified_at": "2026-10-17T09:53:21Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://doi.org/10.1145/1753326.1753555"}], "name": "GUI Testing Using Computer Vision", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. This is a separate visual-testing contribution from the 2009 Sikuli paper, not another version of the same article. MIT PDF download returned HTTP 429; its complete institutional abstract and citation metadata remain the retained review source.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents visual GUI test scripts that use images to specify actions and expected screen feedback. Recording both input events and screenshots can generate tests from demonstrations, supporting unit, regression, and test-driven workflows.", "topics": ["evaluation", "grounding", "automation"], "updated_at": null, "url": "https://doi.org/10.1145/1753326.1753555"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/wanlilll/WeaveBench/tree/main", "gated": false, "hub_id": "wanlilll/WeaveBench", "languages": ["en"], "modalities": [], "size_category_reported": ["n<1K"]}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "e65e5793fe5b251ef7793a0618f4e8139fe5c92c2cf8093d0cc6142e8ece9d02", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://huggingface.co/api/datasets/wanlilll/WeaveBench"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "e65e5793fe5b251ef7793a0618f4e8139fe5c92c2cf8093d0cc6142e8ece9d02", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/wanlilll/WeaveBench"}], "id": "url:6a212bc12f82c824f6b9", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:38:27Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "mit", "retrieved_at": "2026-10-16T06:38:27Z", "url": "https://huggingface.co/datasets/wanlilll/WeaveBench"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/wanlilll/WeaveBench"}], "name": "WeaveBench", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Computer-use benchmark mixing graphical interaction and code execution.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/datasets/wanlilll/WeaveBench"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/xlangai/CUA-Gym/tree/main", "gated": false, "hub_id": "xlangai/CUA-Gym", "languages": ["en"], "modalities": ["tabular", "text"], "size_category_reported": ["10K<n<100K"]}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:26Z", "role": "primary", "sha256": "43b8ebde3b971e1b8b61d789c786ead59b45a2490bc6ed19e204bafc03a31f44", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://huggingface.co/api/datasets/xlangai/CUA-Gym"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T16:10:25Z", "role": "primary", "sha256": "962cd9162bea83f7adfedb3da50129bcf4203ae65176bdf53f0af6788b9603ef", "supports": ["relations"], "url": "https://huggingface.co/datasets/xlangai/CUA-Gym/raw/main/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:38:26Z", "role": "primary", "sha256": "43b8ebde3b971e1b8b61d789c786ead59b45a2490bc6ed19e204bafc03a31f44", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/xlangai/CUA-Gym"}], "id": "url:6aa1aa58a7a888395189", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-16T06:38:26Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "cc-by-4.0", "retrieved_at": "2026-10-16T06:38:26Z", "url": "https://huggingface.co/datasets/xlangai/CUA-Gym"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/xlangai/CUA-Gym"}], "name": "CUA-Gym", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/xlangai/CUA-Gym/raw/main/README.md", "target_id": "arxiv:2605.25624", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Computer-use environment and task resources for agent research.", "topics": ["evaluation", "data-generation"], "updated_at": null, "url": "https://huggingface.co/datasets/xlangai/CUA-Gym"}
{"aliases": [], "authors": [], "details": {"artifact_type": "Ubuntu VMware image", "disabled": false, "download_url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/tree/main", "gated": false, "hub_id": "OS-Copilot/ScienceBoard-Env", "license_review": {"weights": {"evidence_urls": ["https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md", "https://huggingface.co/api/models/OS-Copilot/ScienceBoard-Env"], "reason": "This resource is a prebuilt Ubuntu VMware environment archive (VM.zip), not neural model weights. MIT Hub metadata in a model namespace cannot establish a weights license or relicense the guest OS and bundled applications."}}}, "environments": ["linux", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "8ee436f67e708bb917992764bc60a017c9db2edb21bdb99356b35c69f73724d7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.weights"], "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:18Z", "role": "primary", "sha256": "9ffaf62b54a610afe186b3576b55fdd8287289e0cdfb88371bd9849a616f8460", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/OS-Copilot/ScienceBoard-Env"}], "id": "url:75917f2c2b2802aafd3e", "identifiers": {}, "kind": "infrastructure", "last_verified_at": "2026-10-17T04:07:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env"}], "name": "ScienceBoard VMware environment", "notes": "The public Hub manifest contains VM.zip. This is an environment image, not neural model weights despite its Hub namespace. The card declares MIT metadata for the hosted release, but that does not relicense Ubuntu or included applications. The archive was not downloaded or booted; VMware’s own terms are separate.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md", "target_id": "arxiv:2505.19897", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A prebuilt Ubuntu virtual-machine archive containing the software environment used for ScienceBoard scientific computer-use tasks.", "topics": ["sandbox", "science", "evaluation"], "updated_at": null, "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/osunlp/WebJudge-7B/tree/main", "gated": false, "hub_id": "osunlp/WebJudge-7B"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:41Z", "role": "primary", "sha256": "425f11651cdae09997c1b038cdc9c091778c35a51b8de88140fc6dc935b8bb64", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/osunlp/WebJudge-7B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:41Z", "role": "primary", "sha256": "2eebc1f909263be946057306b41d733e29f0f32337dae02da4e989b5477cf31f", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/osunlp/WebJudge-7B"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T04:03:41Z", "role": "primary", "sha256": "425f11651cdae09997c1b038cdc9c091778c35a51b8de88140fc6dc935b8bb64", "supports": ["licenses.weights"], "url": "https://huggingface.co/osunlp/WebJudge-7B/raw/main/README.md"}], "id": "url:772161f06c2b3d4978b5", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T04:03:41Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T04:03:41Z", "url": "https://huggingface.co/osunlp/WebJudge-7B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/osunlp/WebJudge-7B"}], "name": "WebJudge 7B", "notes": "Four public safetensors shards are present. Apache-2.0 is declared for the weights. This is an evaluator checkpoint rather than a navigation policy; human-agreement rates are not treated as task-success scores.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/osunlp/WebJudge-7B/raw/main/README.md", "target_id": "arxiv:2504.01382", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A visual trajectory judge that retains selected intermediate screenshots to evaluate whether a web agent completed its task.", "topics": ["evaluation", "reward-models", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/osunlp/WebJudge-7B"}
{"aliases": ["url:f4bcd696942ee04dcc42", "bevaya/ScreenSpot", "https://huggingface.co/datasets/bevaya/ScreenSpot"], "authors": [], "details": {"benchmark_protocol": {"actions": ["predicted click point; a predicted box is reduced to its center"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/bevaya/ScreenSpot/raw/0be08781e2e188582f6131625ae1598d443b4d5d/README.md", "population": "Hub test split at 0be08781; not unique screenshots", "qualifier": "exact", "unit": "instruction–target examples", "value": 1272}], "evaluation": "The SeeClick evaluator scores whether the predicted point falls inside the target box, with platform and text/icon breakdowns. The Hub packaging uses xyxy boxes, whereas the original evaluator reads xywh and converts it; adapt the loader to the chosen release.", "evidence_urls": ["https://huggingface.co/datasets/bevaya/ScreenSpot/raw/0be08781e2e188582f6131625ae1598d443b4d5d/README.md", "https://huggingface.co/api/datasets/rootsautomation/ScreenSpot", "https://raw.githubusercontent.com/njucckevin/SeeClick/HEAD/pretrain/screenspot_test.py"], "focus": ["single-step GUI element grounding"], "notes": "Offline, zero-shot evaluation across mobile, desktop, and web. The card explicitly excludes training use in its intended-use description. Public API reports ungated Parquet files and canonical Hub id bevaya/ScreenSpot; the catalog identity is preserved. Apache-2.0 is the Hub data declaration. No runtime or underlying screenshot rights audit was performed.", "observations": ["screenshot", "English instruction"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/rootsautomation/ScreenSpot/tree/main", "gated": false, "hub_id": "bevaya/ScreenSpot", "languages": ["en"], "modalities": ["image", "text"], "size_category_reported": null}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:26Z", "role": "primary", "sha256": "9e2d2a2c076032225911ae1032b548eafa6331ea81a2779632610f0f1c82dfd9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://huggingface.co/api/datasets/rootsautomation/ScreenSpot"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:10:00Z", "role": "primary", "sha256": "5ab2240c064502f75a6726311e10e7724ab7055347697f883b15a6ee8e7d00a6", "supports": ["details.benchmark_protocol"], "url": "https://huggingface.co/datasets/bevaya/ScreenSpot/raw/0be08781e2e188582f6131625ae1598d443b4d5d/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:04Z", "role": "primary", "sha256": "eb72113eefd4868114f8a8105eba2d3385da9034a5b3a4860661e03ae7dcae15", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/njucckevin/SeeClick/HEAD/pretrain/screenspot_test.py"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:38:26Z", "role": "primary", "sha256": "9e2d2a2c076032225911ae1032b548eafa6331ea81a2779632610f0f1c82dfd9", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/rootsautomation/ScreenSpot"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:14:56Z", "role": "primary", "sha256": "af0c4babeff84158ff196b54bf47242923bbf502d2805a1dd4eb52dba44be89c", "supports": ["aliases"], "url": "https://huggingface.co/api/datasets/bevaya/ScreenSpot"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:14:56Z", "role": "primary", "sha256": "af0c4babeff84158ff196b54bf47242923bbf502d2805a1dd4eb52dba44be89c", "supports": ["aliases"], "url": "https://huggingface.co/api/datasets/rootsautomation/ScreenSpot"}], "id": "url:7abb418096432f56c685", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:38:26Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "apache-2.0", "retrieved_at": "2026-10-16T06:38:26Z", "url": "https://huggingface.co/datasets/rootsautomation/ScreenSpot"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/rootsautomation/ScreenSpot"}], "name": "ScreenSpot", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI grounding test set of 1,272 screenshot instructions from iOS, Android, macOS, Windows and web interfaces, each labeled with the target element's bounding box.", "topics": ["grounding", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/rootsautomation/ScreenSpot"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["PyAutoGUI script: pointer movement/click/drag, key presses, hotkeys, and text entry"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/Writer/omniact/raw/5e4bcb26745afc8d58ba9519ea9c9fb832386d26/README.md", "population": "train split, HF revision 5e4bcb26", "qualifier": "exact", "unit": "task examples", "value": 6788}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/Writer/omniact/raw/5e4bcb26745afc8d58ba9519ea9c9fb832386d26/README.md", "population": "test split, HF revision 5e4bcb26", "qualifier": "exact", "unit": "task examples", "value": 2020}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/Writer/omniact/raw/5e4bcb26745afc8d58ba9519ea9c9fb832386d26/README.md", "population": "val split, HF revision 5e4bcb26", "qualifier": "exact", "unit": "task examples", "value": 991}], "evaluation": "The bundled evaluator reports weighted sequence match and action match, comparing action types and applying coordinate, key, and writing penalties. Target element-box metadata is evaluation-only. The retained scorer has a click-penalty defect: predicted coordinates are read from gold_line, so score reproduction requires explicit scorer-version review.", "evidence_urls": ["https://huggingface.co/datasets/Writer/omniact/raw/5e4bcb26745afc8d58ba9519ea9c9fb832386d26/README.md", "https://huggingface.co/api/datasets/Writer/omniact", "https://huggingface.co/datasets/Writer/omniact/raw/5e4bcb26745afc8d58ba9519ea9c9fb832386d26/eval.py"], "focus": ["offline desktop/web action-script generation"], "notes": "This released split table totals 9,799 examples; it is not silently substituted for any paper-level collection size. Public files include train/test/val JSON indexes, task/image archives, and eval.py. Evaluation is offline script comparison, not execution success on a live desktop. MIT is the Hub data declaration; original app/content rights are not resolved by it.", "observations": ["screenshot", "natural-language task"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/Writer/omniact/tree/main", "gated": false, "hub_id": "Writer/omniact", "languages": ["en"], "modalities": [], "size_category_reported": null}, "environments": ["desktop", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "2db0a5dd2b983dded3b90a8ecd66cadbb39074f03778a339822418531661b3e3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://huggingface.co/api/datasets/Writer/omniact"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:10:00Z", "role": "primary", "sha256": "7b54ad557f89d621fe39c3d2fe9f02164a2bfcb7fef0db610d964586d3673693", "supports": ["details.benchmark_protocol"], "url": "https://huggingface.co/datasets/Writer/omniact/raw/5e4bcb26745afc8d58ba9519ea9c9fb832386d26/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:06Z", "role": "primary", "sha256": "a786de678af46df5e2b6b1a2a69d2ee239fe8fd5b63663765d84d39e21fc2365", "supports": ["details.benchmark_protocol"], "url": "https://huggingface.co/datasets/Writer/omniact/raw/5e4bcb26745afc8d58ba9519ea9c9fb832386d26/eval.py"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "2db0a5dd2b983dded3b90a8ecd66cadbb39074f03778a339822418531661b3e3", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/Writer/omniact"}], "id": "url:849a86beaee006c6b147", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:38:27Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "mit", "retrieved_at": "2026-10-16T06:38:27Z", "url": "https://huggingface.co/datasets/Writer/omniact"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/Writer/omniact"}], "name": "OmniAct", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/api/datasets/Writer/omniact", "target_id": "arxiv:2402.17553", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Dataset and benchmark for generating computer interaction programs.", "topics": ["evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/Writer/omniact"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_preference_pair/raw/main/README.md", "population": "Hub split named test", "qualifier": "exact", "unit": "rows", "value": 9460}], "evidence_urls": ["https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_preference_pair/raw/main/README.md", "https://raw.githubusercontent.com/kyle8581/Web-Shepherd/HEAD/README.md"], "formats": ["Parquet"], "modalities": ["screenshots", "text", "action history"], "notes": "Preserves the publisher’s split label despite the collection’s training purpose. Rows are not individual preference pairs because rejected is a list. Data license is unknown.", "splits": ["test"], "supervision": ["preferred and rejected continuations", "task checklists"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_preference_pair/tree/main", "gated": false, "hub_id": "LangAGI-Lab/WebPRMCollection_preference_pair"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "9c091c5fc625ce53f53f8262f5cee8d08194e4fd1b8eeed543ff1c3723c823fd", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_preference_pair/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "28a1d5d12e84fc38515bd3c5031ae0cf315c5f8772b90d982d926b98f2857563", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/LangAGI-Lab/WebPRMCollection_preference_pair"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "07431058d8b9a71b80e75b7a74a40a8346c1503eec240628459cc4d4e2f0802f", "supports": ["summary", "environments"], "url": "https://raw.githubusercontent.com/kyle8581/Web-Shepherd/HEAD/README.md"}], "id": "url:8dfc84eb4016204fe70a", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:44:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_preference_pair"}], "name": "WebPRM Collection preference data", "notes": "Ten public Parquet shards are present. The card names its split test and reports 9,460 rows; the project’s 40K-plus preference-annotation claim is a different grain and is not substituted for this row count. A row can contain multiple rejected alternatives. No data-license declaration is present.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_preference_pair/raw/main/README.md", "target_id": "arxiv:2505.15277", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Checklist-annotated web decision records with page observations, prior actions, one preferred continuation and rejected alternatives for process-reward learning.", "topics": ["web-navigation", "reward-models", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_preference_pair"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["candidate-decision preference"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/LangAGI-Lab/WebRewardBench/raw/main/README.md", "population": "Hub test split", "qualifier": "exact", "unit": "rows", "value": 776}], "evaluation": "Compare reward-model preferences with the provided chosen/rejected decision labels; no independent score was reproduced.", "evidence_urls": ["https://huggingface.co/datasets/LangAGI-Lab/WebRewardBench/raw/main/README.md", "https://raw.githubusercontent.com/kyle8581/Web-Shepherd/HEAD/README.md"], "focus": ["web process-reward evaluation"], "notes": "The current Hub row count and project case-count claim use unresolved counting units. Candidate lists mean a row must not be assumed to equal one task or one pair.", "observations": ["page screenshots", "text observations", "action history", "task checklists"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/LangAGI-Lab/WebRewardBench/tree/main", "gated": false, "hub_id": "LangAGI-Lab/WebRewardBench"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "0968f60eea10b95f79cd16ad3746b0e0b1acaeb0e97c45856f709d53a14dd73a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/LangAGI-Lab/WebRewardBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "db06c66b8cf5687008b0ad0d9ae269fcc27936887d238cfe2bdeb52151e1683e", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/LangAGI-Lab/WebRewardBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "07431058d8b9a71b80e75b7a74a40a8346c1503eec240628459cc4d4e2f0802f", "supports": ["summary", "environments"], "url": "https://raw.githubusercontent.com/kyle8581/Web-Shepherd/HEAD/README.md"}], "id": "url:ac6d0f80765f5a32255a", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T02:44:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/LangAGI-Lab/WebRewardBench"}], "name": "WebRewardBench", "notes": "Two public Parquet shards are present. Card metadata reports 776 test rows, whereas the project describes 1,000-plus test cases without reconciling that count to rows. Both observations are retained as a grain/coverage caveat rather than treating either as a confirmed number of independent tasks. No data license is declared.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/LangAGI-Lab/WebRewardBench/raw/main/README.md", "target_id": "arxiv:2505.15277", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "An offline preference-evaluation set for web process reward models, pairing page and trajectory context with preferred and rejected candidate decisions.", "topics": ["web-navigation", "reward-models", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/LangAGI-Lab/WebRewardBench"}
{"aliases": [], "authors": [], "details": {"availability": "Ungated v1.0 JSON task-definition file.", "benchmark_protocol": {"actions": ["mobile GUI actions through a separately configured agent"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/OpenDFM/MobA-MobBench/raw/main/README.md", "population": "MobBench tasks used in the paper", "qualifier": "exact", "unit": "tasks", "value": 50}], "evaluation": "The publisher reports a fifty-task real-device evaluation; the inspected release does not establish an automatic correctness judge.", "evidence_urls": ["https://huggingface.co/datasets/OpenDFM/MobA-MobBench/raw/main/README.md"], "focus": ["mobile task completion"], "notes": "Task definitions are distinct from device images, agent execution and automated scoring.", "observations": ["Android device interface"], "setting": "interactive"}, "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/OpenDFM/MobA-MobBench/raw/main/README.md"], "reason": "Dataset-card declaration for the MobBench task-definition release only."}}, "release_revision": "28941d10c49a16efb991d4157e512e14204bf3d6"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:07Z", "role": "primary", "sha256": "231a8f483256c774840c99383ab6536520a0aed0f04d62b4d64f1f59477b906e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol", "licenses.data"], "url": "https://huggingface.co/datasets/OpenDFM/MobA-MobBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:07Z", "role": "primary", "sha256": "4e7225f956d7cee98fe671e7bfd91fc407d87dd805d8f0c11c3b0c839714ccf3", "supports": ["identity", "release_status", "details.release_revision"], "url": "https://huggingface.co/api/datasets/OpenDFM/MobA-MobBench"}], "id": "url:b53e03a6e4ca4226d126", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T12:32:07Z", "licenses": {"code": null, "data": {"basis": "Dataset-card declaration for the MobBench task-definition release only.", "declared": "cc-by-nc-sa-4.0", "retrieved_at": "2026-10-17T12:32:07Z", "url": "https://huggingface.co/datasets/OpenDFM/MobA-MobBench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/OpenDFM/MobA-MobBench"}], "name": "MobBench task definitions", "notes": "The ungated manifest lists MobBench_tasks_release_v1.0.json and bilingual cards. The card reports fifty tasks; no sample records were downloaded to recount them. This release supplies task definitions and agent setup guidance, not an independently verified automatic judge or a bundled Android environment. The data card declares CC-BY-NC-SA-4.0; code and model licenses are separate.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/OpenDFM/MobA-MobBench/raw/main/README.md", "target_id": "arxiv:2410.13757", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A public set of real-device mobile task definitions used to evaluate MobA, with Chinese and English task material packaged in a versioned JSON release.", "topics": ["evaluation", "mobile-navigation"], "updated_at": null, "url": "https://huggingface.co/datasets/OpenDFM/MobA-MobBench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["click coordinates"], "counts": [], "evaluation": "The inspected Jedi evaluation adapter reads the v2 mobile/desktop/web annotation files, scores a point against the target box, and reports overall, platform, and text/icon accuracy. This is a named compatible evaluator implementation; the OS-Atlas dataset card itself supplies no evaluator specification.", "evidence_urls": ["https://huggingface.co/api/datasets/OS-Copilot/ScreenSpot-v2", "https://osatlas.github.io/", "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/qwen25_vllm_screenspot_v2_jedi.py", "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/eval.py"], "focus": ["single-step GUI grounding on the ScreenSpot-v2 release"], "notes": "The public Hub revision 5efbb1f1 contains three platform JSON annotation files and an image ZIP; its card only declares Apache-2.0. No sample total or equivalence to the original ScreenSpot test count is asserted. Keep v2 separate from v1 and retain evaluator/version and coordinate conversion when comparing results. Files were listed, not downloaded as data.", "observations": ["screenshot", "instruction"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/OS-Copilot/ScreenSpot-v2/tree/main", "gated": false, "hub_id": "OS-Copilot/ScreenSpot-v2", "languages": [], "modalities": ["image"], "size_category_reported": null}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:26Z", "role": "primary", "sha256": "7feddadc306e384f8df6a3d9c8afc0c776cb3734c4c7fe3eb842eab85537f408", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://huggingface.co/api/datasets/OS-Copilot/ScreenSpot-v2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:22:05Z", "role": "primary", "sha256": "94547cb88756cf6a36bfb86d5fe9bdca1e40d7edb9e452c33cb462bbe760fed0", "supports": ["details.benchmark_protocol"], "url": "https://osatlas.github.io/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:09Z", "role": "primary", "sha256": "0e8772bc8fd5a25c50d3c2f6b0caee05ed2ab4659d46d99d60adf2b5ab29e997", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/qwen25_vllm_screenspot_v2_jedi.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:09Z", "role": "primary", "sha256": "b85f5685ce6c200ea648471666177859693d7f40cf50c12d08765ae68580de4a", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/eval.py"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:38:26Z", "role": "primary", "sha256": "7feddadc306e384f8df6a3d9c8afc0c776cb3734c4c7fe3eb842eab85537f408", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/OS-Copilot/ScreenSpot-v2"}], "id": "url:bd7214de6d98917c2350", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:38:26Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "apache-2.0", "retrieved_at": "2026-10-16T06:38:26Z", "url": "https://huggingface.co/datasets/OS-Copilot/ScreenSpot-v2"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/OS-Copilot/ScreenSpot-v2"}], "name": "ScreenSpot v2", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Second version of the ScreenSpot grounding test set, published by OS-Copilot as separate desktop, mobile and web annotation files with a shared screenshot archive.", "topics": ["grounding", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/OS-Copilot/ScreenSpot-v2"}
{"aliases": [], "authors": ["Fei Tang", "Zhangxuan Gu", "Zhengxi Lu", "Shangzhan Zhang", "Zhengwen Zeng", "Shuheng Shen", "Changhua Meng", "Yuchen Yan", "Wenqi Zhang", "Yongliang Shen", "Weiming Lu", "Yueting Zhuang"], "details": {"publication": {"authors": ["Fei Tang", "Zhangxuan Gu", "Zhengxi Lu", "Shangzhan Zhang", "Zhengwen Zeng", "Shuheng Shen", "Changhua Meng", "Yuchen Yan", "Wenqi Zhang", "Yongliang Shen", "Weiming Lu", "Yueting Zhuang"], "date_basis": "Explicit publisher metadata: month = {June}, year = {2026}; precision=month.", "date_parts": [2026, 6], "doi": null, "event_dates": null, "evidence_urls": ["https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html", "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"], "identity_basis": "No title/DOI/canonical URL match in the current paper catalog or raw archive; closest substantive same-team works were reread and have different contributions.", "issue": null, "notes": "CVF open-access accepted version; PDF states that the final proceedings version is on IEEE Xplore.", "pages": "13007-13016", "pdf_url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf", "publisher": null, "series": null, "title": "GUI-SAGE: Enhancing GUI Automation with Self-Explanatory Learning", "type": "conference-paper", "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html", "venue": "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)", "volume": null}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:48+00:00", "role": "primary", "sha256": "1d5ed4e24cf9763cc43f0a8ce14f0e98c01504c73578e5632cea573dfb7da647", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:56:04+00:00", "role": "primary", "sha256": "244f40d77a40dfd184e78cf8e1e2dd9e469d3e594a8ce214279c2ce17b8f1012", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.title"], "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:48+00:00", "role": "primary", "sha256": "1d5ed4e24cf9763cc43f0a8ce14f0e98c01504c73578e5632cea573dfb7da647", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.type", "details.publication.url", "details.publication.venue", "environments", "kind", "scope", "summary", "topics"], "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:56:04+00:00", "role": "primary", "sha256": "244f40d77a40dfd184e78cf8e1e2dd9e469d3e594a8ce214279c2ce17b8f1012", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.title", "environments", "name", "scope", "summary", "topics"], "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"}], "id": "url:beb8f1fd6b945ffc9d93", "identifiers": {}, "kind": "paper", "last_verified_at": "2026-10-17T09:53:48+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}, {"rel": "publication_pdf", "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"}], "name": "GUI-SAGE: Enhancing GUI Automation with Self-Explanatory Learning", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. New work proposal from the official CVF accepted open-access version. No invented day, DOI, or arXiv identifier; original annotations describe the reviewed sources and not independent reproduction.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-SAGE generates reasoning conditioned on known actions and weights policy updates using confidence and reward signals. Its experiments use AndroidControl and GUIOdyssey data and separately evaluate interactive AndroidWorld tasks.", "topics": ["reinforcement-learning", "data-generation", "evaluation", "robustness"], "updated_at": null, "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}
{"aliases": [], "authors": [], "details": {"availability": "Ungated benchmark JSONL/image manifest and public evaluation scripts; training pipeline/checkpoints are pending.", "benchmark_protocol": {"actions": ["critic-model responses"], "counts": [{"basis": "computed", "evidence_url": "https://huggingface.co/datasets/OS-Copilot/OS-Critic-Bench/raw/main/README.md", "population": "mobile/desktop/web configurations in the Hub card", "qualifier": "exact", "unit": "platform JSONL splits", "value": 3}], "evaluation": "The official repository runs model inference and computes accuracy with cal_acc.py.", "evidence_urls": ["https://raw.githubusercontent.com/numbmelon/OS-Oracle/HEAD/README.md", "https://huggingface.co/datasets/OS-Copilot/OS-Critic-Bench/raw/main/README.md"], "focus": ["GUI critic prediction"], "notes": "No sample count is inferred from the n<1K Hub size band. Evaluation code availability does not establish the release of OS-Oracle training data or weights.", "observations": ["benchmark JSONL", "UI screenshots"], "setting": "offline"}, "gated": false, "hub_id": "OS-Copilot/OS-Critic-Bench", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/OS-Copilot/OS-Critic-Bench/raw/main/README.md"], "reason": "Publisher dataset-card license declaration for OS-Critic-Bench; not an evaluation-code or model-weight license."}}, "release_revision": "8fe774ef6a422c0e555f90205eb392f9132e272b"}, "environments": ["cross-platform", "desktop", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:18Z", "role": "primary", "sha256": "58305b8ebe1af5dfbddd636484d63722d4e21d28ed3e22bb8c0068e654f33479", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/numbmelon/OS-Oracle/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:17Z", "role": "primary", "sha256": "17bcf851f39d06e7e357871df1dc1ddf697bcbea31842c16348227bbf4fbb224", "supports": ["identity"], "url": "https://api.github.com/repos/numbmelon/OS-Oracle"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:55Z", "role": "primary", "sha256": "516beea465382b54548e4b8bfd12ae510080a49451b30262ad1c769ed5a3398e", "supports": ["release_status", "licenses.code"], "url": "https://api.github.com/repos/numbmelon/OS-Oracle/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:06Z", "role": "primary", "sha256": "ef311b49437e246ffef6ad13c090277a0bdacc4a124e5b3cfe8d2643ced8f843", "supports": ["identity", "release_status", "details.hub_id", "details.gated", "details.release_revision"], "url": "https://huggingface.co/api/datasets/OS-Copilot/OS-Critic-Bench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:07Z", "role": "primary", "sha256": "4139e25e8911ca8450f6db9acb8dc3a98c126c3c43514c72bcefa4ed33b8ffa8", "supports": ["licenses.data", "details.benchmark_protocol"], "url": "https://huggingface.co/datasets/OS-Copilot/OS-Critic-Bench/raw/main/README.md"}], "id": "url:c73ba0618d1b3c505887", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T11:18:18Z", "licenses": {"code": null, "data": {"basis": "Publisher dataset-card license declaration for OS-Critic-Bench; not an evaluation-code or model-weight license.", "declared": "apache-2.0", "retrieved_at": "2026-10-17T11:23:07Z", "url": "https://huggingface.co/datasets/OS-Copilot/OS-Critic-Bench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/OS-Copilot/OS-Critic-Bench"}], "name": "OS-Critic Bench", "notes": "The Hub card provides split/format/license metadata, while the official OS-Oracle README establishes critic-evaluation purpose and platform scope. The repository currently releases only benchmark inference/scoring components; data-synthesis code, training datasets and model checkpoints remain unchecked TODOs and are not counted as released assets. Apache-2.0 is recorded for benchmark data only; the evaluation code has no inspected license declaration.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/numbmelon/OS-Oracle/HEAD/README.md", "target_id": "arxiv:2512.16295", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A cross-platform GUI-critic benchmark with mobile, desktop and web JSONL splits and screenshot assets, accompanied by public inference and accuracy-scoring scripts.", "topics": ["evaluation", "reward-models", "screen-understanding"], "updated_at": null, "url": "https://huggingface.co/datasets/OS-Copilot/OS-Critic-Bench"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "computed", "evidence_url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md", "population": "sum of icon/text counts in the published card table", "qualifier": "exact", "unit": "target annotations", "value": 1581}], "evidence_urls": ["https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md", "https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/eval_screenspot_pro.py"], "formats": ["JSON annotations", "screen images"], "modalities": ["high-resolution screenshots", "instructions", "bounding boxes"], "notes": "Counts describe target annotations, not unique screenshots or expanded language/style variants. The evaluation code can reuse a target across configurations.", "splits": [], "supervision": ["GUI grounding"]}, "download_url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/tree/main"}, "environments": ["desktop", "windows", "macos", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:35:55Z", "role": "primary", "sha256": "76ae2231564fe84af029f8625c02d9cd7fb6cf85c5f8af1b2a2593e2107719d9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:38:05Z", "role": "primary", "sha256": "a77c859775b33591b5619e8520c94702a53a1f9a29ada93e66744dbe23afe9dc", "supports": ["details.dataset_profile"], "url": "https://raw.githubusercontent.com/likaixin2000/ScreenSpot-Pro-GUI-Grounding/HEAD/eval_screenspot_pro.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:10Z", "role": "primary", "sha256": "2f7f635dee3c593b40efc61246c082ab1d76f7ce78f84a1d4b6391d30fa4e8fd", "supports": ["relations"], "url": "https://gui-agent.github.io/grounding-leaderboard/"}, {"method": "huggingface_card", "retrieved_at": "2026-10-16T11:35:55Z", "role": "primary", "sha256": "76ae2231564fe84af029f8625c02d9cd7fb6cf85c5f8af1b2a2593e2107719d9", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md"}], "id": "url:c79e152e6f7af1a9efed", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-16T11:35:55Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-16T11:35:55Z", "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro"}], "name": "ScreenSpot-Pro data", "notes": "This is the benchmark data asset, distinct from its evaluation-code repository. Target counts are not unique image counts.", "published_at": null, "relations": [{"evidence_url": "https://gui-agent.github.io/grounding-leaderboard/", "target_id": "arxiv:2504.07981", "type": "documented_by"}, {"evidence_url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro/raw/main/README.md", "target_id": "github:likaixin2000/screenspot-pro-gui-grounding", "type": "component_of"}], "review_status": "curated", "scope": "core", "summary": "High-resolution professional interface screenshots and target annotations for evaluating instruction-based GUI localization.", "topics": ["grounding", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/likaixin/ScreenSpot-Pro"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["success verdict", "process score"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md", "population": "fara7b_om2w_browserbase split", "qualifier": "exact", "unit": "trajectories", "value": 106}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md", "population": "internal split", "qualifier": "exact", "unit": "trajectories", "value": 154}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md", "population": "fara7b_om2w_browserbase split", "qualifier": "exact", "unit": "annotation rows", "value": 215}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md", "population": "internal split", "qualifier": "exact", "unit": "annotation rows", "value": 154}], "evaluation": "Agreement metrics including Cohen’s kappa, accuracy and F1 against human annotations; review stages must be kept distinct.", "evidence_urls": ["https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md"], "focus": ["verifier-human agreement", "process quality"], "notes": "Trajectory rows and reviewer rows have different grains. Internal task labels do not make the released Parquet files private. No payload data was downloaded.", "observations": ["screenshots", "action logs", "final answers"], "setting": "offline"}, "disabled": false, "download_url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/tree/main", "gated": false, "hub_id": "microsoft/CUAVerifierBench", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md"], "reason": "The benchmark's Hub dataset card declares MIT in its metadata header (license: mit)."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "44d7190b72ae489297f59d6fadff635ced6dd067c845c77b04f187a7b2d7bec7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.data"], "url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "fb1fe8a9fd6e45fc82157c65c08d5c682d08bead5196308776200ef3c39c1e41", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/microsoft/CUAVerifierBench"}], "id": "url:d5140d438466c1e1348a", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T02:42:32Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:42:32Z", "url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench"}], "name": "CUAVerifierBench", "notes": "Public Parquet configs separate trajectories from per-reviewer annotations. The card declares MIT. The two splits have different review stages: Online-Mind2Web includes blind and verifier-informed annotations, whereas the internal split is blind-only. This measures verifier agreement rather than agent task success.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench/raw/main/README.md", "target_id": "arxiv:2604.06240", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Human-annotated browser-agent trajectories for comparing automated task-success and process-quality verifiers against reviewer judgments.", "topics": ["evaluation", "reward-models", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/datasets/microsoft/CUAVerifierBench"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["mobile interface operations", "clarification requests"], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench/raw/main/README.md", "population": "MobileGUI-VBench release", "qualifier": "exact", "unit": "tasks", "value": 150}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench/raw/main/README.md", "population": "MobileGUI-VBench release", "qualifier": "exact", "unit": "mobile apps", "value": 40}], "evaluation": "Published task definitions include preconditions and clarification requirements; the metadata review does not independently validate the reported agent scores.", "evidence_urls": ["https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench/raw/main/README.md"], "focus": ["mobile task completion", "clarification behavior", "cross-app tasks"], "notes": "The card describes 140 single-app and 10 cross-app tasks. Task and app counts are different units.", "observations": ["mobile app interfaces"], "setting": "interactive"}, "disabled": false, "download_url": "https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench/tree/main", "gated": false, "hub_id": "vivo-ai/MobileGUI-VBench"}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "422ff6a557a1aabd8f6c6ef26e5fd05b992f4acf0e5128ab8bc437021fcc5ea7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "5850fc58f54c877cffe328dba2a0b5e3e98767637aae6d6efd9a037cf06eabce", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/vivo-ai/MobileGUI-VBench"}], "id": "url:d7a4f34a6a5202af5f64", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T02:42:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench"}], "name": "MobileGUI-VBench", "notes": "Public task files are available as JSONL and XLSX with documentation. The card explicitly declares Apache-2.0 for the dataset, despite having no license field in its YAML metadata. It identifies mobile apps without fixing an operating-system taxonomy; model scores are not imported.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench/raw/main/README.md", "target_id": "arxiv:2609.12394", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Mobile-app task definitions covering everyday scenarios, with annotated preconditions and clarification requirements for evaluating agent behavior.", "topics": ["evaluation", "mobile-navigation", "cross-app"], "updated_at": null, "url": "https://huggingface.co/datasets/vivo-ai/MobileGUI-VBench"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://recreation-bench.cc/#leaderboard"}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "79deaa899d085ece77a127d53b12ddb9edea8a796943f35b75e9acec67bcb976", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/Qwen/RecreationBench/raw/main/README.md"}], "id": "url:d9d8e68f19ef0956d189", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:50:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/Qwen/RecreationBench"}], "name": "RecreationBench", "notes": "Dataset card declares multiple upstream licenses; do not treat the package as having one blanket permissive license.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Tasks for recreating reference applications with hybrid computer interaction.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/datasets/Qwen/RecreationBench"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/ritzzai/GUI-R1/tree/main", "gated": false, "hub_id": "ritzzai/GUI-R1"}, "environments": ["desktop", "windows", "linux", "macos", "android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "7b14996d1a99675990cb2f3aeada2b2f9898f9daee46d3680988e16f6365abe3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "f31ae9547d94967dd91f5cde5e5c66430d7d59ee3a525c0914bc600a79999af7", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/ritzzai/GUI-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "94c7c78bede11dde727940f56592a04594c1c59438812dff47540bf53c1bed7d", "supports": ["summary", "environments", "release_status"], "url": "https://raw.githubusercontent.com/ritzz-ai/GUI-R1/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "7b14996d1a99675990cb2f3aeada2b2f9898f9daee46d3680988e16f6365abe3", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md"}], "id": "url:df680385f18b01a825be", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:42:32Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:42:32Z", "url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/ritzzai/GUI-R1"}], "name": "GUI-R1 data bundle", "notes": "The Hub is a dataset repository with Parquet payloads, although the short card mistakenly calls it a model. The official project identifies it as training/evaluation data and describes subsequent expanded mixtures. MIT is declared in this dataset card; upstream dataset terms remain separate. No 3K/800K aggregate count is inferred for the mutable bundle.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md", "target_id": "arxiv:2504.10458", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Training and evaluation Parquet bundles accompanying GUI-R1, including GUI grounding and mobile/web action-prediction data.", "topics": ["grounding", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/ritzzai/GUI-R1"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [], "evidence_urls": ["https://huggingface.co/datasets/h-siyuan/ScreenDrag/raw/main/README.md"], "formats": ["Parquet", "episode files"], "modalities": ["screen images", "instructions", "previous state", "continuous actions"], "notes": "Raw and Parquet folders are alternative views of the same data. Each sample describes a dragging task over multiple timesteps; action vectors encode button state and coordinates.", "splits": ["train", "test"], "supervision": ["continuous GUI dragging"]}, "download_url": "https://huggingface.co/datasets/h-siyuan/ScreenDrag/tree/main", "modalities": ["image", "text", "continuous-actions"]}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:12:56Z", "role": "primary", "sha256": "39c7c028f7ef5ca5c75f79b59a107c7e25444776095a2a08bb7535c4f523f7fa", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.modalities", "details.download_url", "details.dataset_profile"], "url": "https://huggingface.co/datasets/h-siyuan/ScreenDrag/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:09:28Z", "role": "primary", "sha256": "88df552ccbe80efd7020905d77d526f4f9efe2d09a11fed2714b083988f9dcef", "supports": ["relations"], "url": "https://raw.githubusercontent.com/showlab/showui-pi/HEAD/README.md"}], "id": "url:fe43af1801032be708cc", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-16T11:12:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/h-siyuan/ScreenDrag"}], "name": "ScreenDrag", "notes": "Raw and Parquet folders are alternative views of the same data and should not be counted as independent samples. The inspected card did not state a license.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/showlab/showui-pi/HEAD/README.md", "target_id": "github:showlab/showui-pi", "type": "component_of"}], "review_status": "curated", "scope": "core", "summary": "GUI dragging data with screen observations, prior state, and continuous pointer actions, provided as trajectory files and Parquet representations.", "topics": ["grounding", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/h-siyuan/ScreenDrag"}