{"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": ["Nancy Xu", "Sam Masling", "Michael Du", "Giovanni Campagna", "Larry Heck", "James Landay", "Monica S Lam"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e00b8f1b34097455ece2cf97efabbb64f398e3fffdec13394f731096672dc654", "submitted_at": "2021-03-30T04:02:34Z", "updated_at": "2021-04-04T23:51:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2103.16057v2"}}, "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:2103.16057", "identifiers": {"arxiv": "2103.16057"}, "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/2103.16057"}], "name": "Grounding Open-Domain Instructions to Automate Web Support 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": "2021-03-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces RUSS, which translates support instructions into a domain-specific language and grounds them in actionable web elements.", "topics": ["grounding", "web-navigation"], "updated_at": "2021-04-04", "url": "https://arxiv.org/abs/2103.16057"}
{"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": ["Izzeddin Gur", "Ofir Nachum", "Yingjie Miao", "Mustafa Safdari", "Austin Huang", "Aakanksha Chowdhery", "Sharan Narang", "Noah Fiedel", "Aleksandra Faust"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b9089164b5c784735d8915596fe902840e01c8215f507caed0e10d4361d13ec", "submitted_at": "2022-10-08T07:27:17Z", "updated_at": "2023-05-19T17:42:56Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2210.03945v2"}}, "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:2210.03945", "identifiers": {"arxiv": "2210.03945"}, "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/2210.03945"}], "name": "Understanding HTML with Large 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.", "published_at": "2022-10-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies language models for HTML element classification, page description and autonomous web navigation. Fine-tuned models are evaluated on MiniWoB, and a distilled, automatically labeled CommonCrawl HTML corpus supports the training approach.", "topics": ["screen-understanding", "web-navigation", "data-generation"], "updated_at": "2023-05-19", "url": "https://arxiv.org/abs/2210.03945"}
{"aliases": [], "authors": ["Geunwoo Kim", "Pierre Baldi", "Stephen McAleer"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8f36bd6088175b337b750d91d0b7c04d2c156742e5c4090a74f00a482157e386", "submitted_at": "2023-03-30T16:01:52Z", "updated_at": "2023-11-16T20:15:14Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2303.17491v3"}}, "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:2303.17491", "identifiers": {"arxiv": "2303.17491"}, "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/2303.17491"}], "name": "Language Models can Solve Computer Tasks", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2023-03-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces recursive criticism and improvement prompting for language-model agents performing computer tasks, evaluated on MiniWoB++ with a small number of demonstrations per task.", "topics": ["web-navigation", "self-reflection", "training-free"], "updated_at": "2023-11-16", "url": "https://arxiv.org/abs/2303.17491"}
{"aliases": [], "authors": ["Hiroki Furuta", "Kuang-Huei Lee", "Ofir Nachum", "Yutaka Matsuo", "Aleksandra Faust", "Shixiang Shane Gu", "Izzeddin Gur"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ac0ea5a462890317ea29ade32c879d016181afbe5962e2d28cc175e251df41e6", "submitted_at": "2023-05-19T17:44:34Z", "updated_at": "2024-02-25T16:21:00Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2305.11854v4"}, "publication": {"authors": ["Hiroki Furuta", "Kuang-Huei Lee", "Ofir Nachum", "Yutaka Matsuo", "Aleksandra Faust", "Shixiang Shane Gu", "Izzeddin Gur"], "date_basis": "Explicit publisher metadata: 2024-05-31; precision=day.", "date_parts": [2024, 5, 31], "doi": null, "event_dates": null, "evidence_urls": ["https://proceedings.iclr.cc/paper_files/paper/2024/hash/7ef7d8359036afd8c2378d82c21058a4-Abstract-Conference.html", "https://proceedings.iclr.cc/paper_files/paper/2024/file/7ef7d8359036afd8c2378d82c21058a4-Paper-Conference.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 ICLR landing shortens Shixiang Shane Gu to Shixiang Gu. Publication authors use the full PDF title-page form. Volume 2024 and pages 29528–29558 are explicit publisher citation metadata, not values inferred from the event year or PDF page count. Publisher-hosted proceedings record; distinct from the retained arXiv preprint snapshot.", "pages": "29528-29558", "pdf_url": "https://proceedings.iclr.cc/paper_files/paper/2024/file/7ef7d8359036afd8c2378d82c21058a4-Paper-Conference.pdf", "publisher": null, "series": null, "title": "Multimodal Web Navigation with Instruction-Finetuned Foundation Models", "type": "conference-paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/7ef7d8359036afd8c2378d82c21058a4-Abstract-Conference.html", "venue": "International Conference on Learning Representations", "volume": "2024"}}, "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": "2026-10-17T09:53:48+00:00", "role": "primary", "sha256": "6669190c7fc2a37f9cd68d0c49431b8f17b626c16bb3ef4093fc83a6ac62acd3", "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.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/7ef7d8359036afd8c2378d82c21058a4-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:59:46+00:00", "role": "primary", "sha256": "da26ea085252edfda4c084058aa81ea023f885ff97c67f2ffae0698bab6f568b", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes"], "url": "https://proceedings.iclr.cc/paper_files/paper/2024/file/7ef7d8359036afd8c2378d82c21058a4-Paper-Conference.pdf"}, {"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:2305.11854", "identifiers": {"arxiv": "2305.11854"}, "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/2305.11854"}, {"rel": "publication", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/7ef7d8359036afd8c2378d82c21058a4-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/file/7ef7d8359036afd8c2378d82c21058a4-Paper-Conference.pdf"}], "name": "Multimodal Web Navigation with Instruction-Finetuned Foundation Models", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2023-05-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains WebGUM offline from demonstrations by jointly tuning an instruction-following language model and a vision encoder over webpage screenshots and HTML. The agent predicts browser actions and is evaluated on MiniWoB, WebShop, and Mind2Web.", "topics": ["web-navigation", "imitation-learning", "action-prediction"], "updated_at": "2024-02-25", "url": "https://arxiv.org/abs/2305.11854"}
{"aliases": [], "authors": ["Abishek Sridhar", "Robert Lo", "Frank F. Xu", "Hao Zhu", "Shuyan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "896d5f5e9248e2e49da254fb383f84e5b6ce4286b6d30a37ed7944fcfa1f4958", "submitted_at": "2023-05-23T17:10:39Z", "updated_at": "2023-10-30T00:55:53Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2305.14257v3"}}, "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:2305.14257", "identifiers": {"arxiv": "2305.14257"}, "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/2305.14257"}], "name": "Hierarchical Prompting Assists Large Language Model on Web Navigation", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2023-05-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Separates web navigation prompting into a SUMMARIZER that extracts action-relevant observations and an ACTOR that chooses the next action from the condensed context.", "topics": ["web-navigation", "planning", "efficiency"], "updated_at": "2023-10-30", "url": "https://arxiv.org/abs/2305.14257"}
{"aliases": [], "authors": ["Peter Shaw", "Mandar Joshi", "James Cohan", "Jonathan Berant", "Panupong Pasupat", "Hexiang Hu", "Urvashi Khandelwal", "Kenton Lee", "Kristina Toutanova"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d06f7480695f45546b62b5b8c3d8c174ca52936e158d46e626689d0918711417", "submitted_at": "2023-05-31T23:39:18Z", "updated_at": "2023-12-06T23:46:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2306.00245v2"}, "publication": {"authors": ["Peter Shaw", "Mandar Joshi", "James Cohan", "Jonathan Berant", "Panupong Pasupat", "Hexiang Hu", "Urvashi Khandelwal", "Kenton Lee", "Kristina Toutanova"], "date_basis": "Explicit publisher metadata: 2023-12-15; precision=day.", "date_parts": [2023, 12, 15], "doi": "10.52202/075280-1490", "event_dates": null, "evidence_urls": ["https://papers.nips.cc/paper_files/paper/2023/hash/6c52a8a4fadc9129c6e1d1745f2dfd0f-Abstract-Conference.html", "https://proceedings.neurips.cc/paper_files/paper/2023/file/6c52a8a4fadc9129c6e1d1745f2dfd0f-Paper-Conference.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 NeurIPS landing metadata adds the initial N to Kristina Toutanova; the formal paper title page does not. Publication authors use the PDF. Publisher-hosted proceedings record; distinct from the retained arXiv preprint snapshot.", "pages": "34354-34370", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2023/file/6c52a8a4fadc9129c6e1d1745f2dfd0f-Paper-Conference.pdf", "publisher": null, "series": null, "title": "From Pixels to UI Actions: Learning to Follow Instructions via Graphical User Interfaces", "type": "conference-paper", "url": "https://papers.nips.cc/paper_files/paper/2023/hash/6c52a8a4fadc9129c6e1d1745f2dfd0f-Abstract-Conference.html", "venue": "Advances in Neural Information Processing Systems", "volume": "36"}}, "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": "2026-10-17T09:53:48+00:00", "role": "primary", "sha256": "938e0a31512bb9a7793b568858f81eb3ceb650a71aae2d91b5292990e890201b", "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.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://papers.nips.cc/paper_files/paper/2023/hash/6c52a8a4fadc9129c6e1d1745f2dfd0f-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:59:47+00:00", "role": "primary", "sha256": "3827298f44d2f24f1333b4d46ecb5eb88c8b028d192b11424580655b38e024f9", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes"], "url": "https://proceedings.neurips.cc/paper_files/paper/2023/file/6c52a8a4fadc9129c6e1d1745f2dfd0f-Paper-Conference.pdf"}, {"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:2306.00245", "identifiers": {"arxiv": "2306.00245"}, "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/2306.00245"}, {"rel": "publication", "url": "https://papers.nips.cc/paper_files/paper/2023/hash/6c52a8a4fadc9129c6e1d1745f2dfd0f-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2023/file/6c52a8a4fadc9129c6e1d1745f2dfd0f-Paper-Conference.pdf"}], "name": "From Pixels to UI Actions: Learning to Follow Instructions via Graphical User Interfaces", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2023-05-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies instruction-following agents that observe screenshots and act through a generic mouse-and-keyboard action space, evaluating pixel-based pretraining for GUI tasks on MiniWoB++.", "topics": ["grounding", "action-prediction", "web-navigation"], "updated_at": "2023-12-06", "url": "https://arxiv.org/abs/2306.00245"}
{"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": ["Longtao Zheng", "Rundong Wang", "Xinrun Wang", "Bo An"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8fdb5fed2b2fa0ecbf5910dd6813d8de7910af286a1e4e00cdebb734a82e78b0", "submitted_at": "2023-06-13T15:49:41Z", "updated_at": "2024-01-19T06:59:26Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2306.07863v3"}}, "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:2306.07863", "identifiers": {"arxiv": "2306.07863"}, "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/2306.07863"}], "name": "Synapse: Trajectory-as-Exemplar Prompting with Memory for Computer 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": "2023-06-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Synapse filters task-irrelevant page state and places complete abstracted state-action trajectories into the prompt. An embedding-based exemplar memory retrieves relevant demonstrations for new tasks, evaluated on MiniWoB++ and Mind2Web.", "topics": ["memory", "web-navigation", "planning", "efficiency"], "updated_at": "2024-01-19", "url": "https://arxiv.org/abs/2306.07863"}
{"aliases": [], "authors": ["Izzeddin Gur", "Hiroki Furuta", "Austin Huang", "Mustafa Safdari", "Yutaka Matsuo", "Douglas Eck", "Aleksandra Faust"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "65328d525a6458ef8c401fb0ac4a9b6a5142bad462bd15a212d124c83a56209f", "submitted_at": "2023-07-24T14:56:30Z", "updated_at": "2024-02-25T16:17:43Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2307.12856v4"}}, "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:2307.12856", "identifiers": {"arxiv": "2307.12856"}, "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/2307.12856"}], "name": "A Real-World WebAgent with Planning, Long Context Understanding, and Program 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": "2023-07-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines task decomposition, HTML summarization, and generated Python programs to carry out instructions on websites.", "topics": ["web-navigation", "planning"], "updated_at": "2024-02-25", "url": "https://arxiv.org/abs/2307.12856"}
{"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": ["Kaixin Ma", "Hongming Zhang", "Hongwei Wang", "Xiaoman Pan", "Wenhao Yu", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "399c97382cac325a9d60a073a14221dee6ee6254cb0b8aa88a5341a7bb758c97", "submitted_at": "2023-09-15T05:44:08Z", "updated_at": "2024-02-21T17:42:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2309.08172v2"}}, "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:2309.08172", "identifiers": {"arxiv": "2309.08172"}, "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/2309.08172"}], "name": "LASER: LLM Agent with State-Space Exploration for Web 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": "2023-09-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "LASER represents web navigation as exploration over a defined set of states, allowing an agent to revisit earlier states after mistakes. The study evaluates this backtracking approach on WebShop and Amazon tasks.", "topics": ["planning", "robustness", "web-navigation"], "updated_at": "2024-02-21", "url": "https://arxiv.org/abs/2309.08172"}
{"aliases": [], "authors": ["Paloma Sodhi", "S. R. K. Branavan", "Yoav Artzi", "Ryan McDonald"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d843f3c65e5961da773ec0265c2cfbf51d7c8214d9232ecad1638cc0dbb9ef64", "submitted_at": "2023-10-05T17:40:09Z", "updated_at": "2024-08-08T18:00:48Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2310.03720v4"}}, "environments": ["web"], "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:2310.03720", "identifiers": {"arxiv": "2310.03720"}, "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/2310.03720"}], "name": "SteP: Stacked LLM Policies for Web Actions", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "SteP dynamically composes specialized web-action policies through a stack that records the active chain of policy calls. It studies this control structure across WebArena, MiniWoB++, and CRM workflows.", "topics": ["planning", "web-navigation", "workflow"], "updated_at": "2024-08-08", "url": "https://arxiv.org/abs/2310.03720"}
{"aliases": [], "authors": ["Tao Li", "Gang Li", "Zhiwei Deng", "Bryan Wang", "Yang Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "95092986184a232af62c61fa2aedefb837851dce9e0ff7a1128a67d47edfde40", "submitted_at": "2023-10-12T21:53:37Z", "updated_at": "2023-10-23T17:39:51Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2310.08740v3"}}, "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:2310.08740", "identifiers": {"arxiv": "2310.08740"}, "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/2310.08740"}], "name": "A Zero-Shot Language Agent for Computer Control with Structured Reflection", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-10-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies a zero-shot computer-control agent that plans executable actions and organizes its own reflections after mistakes. MiniWoB++ experiments examine learning to complete tasks without supplied expert trajectories.", "topics": ["self-reflection", "planning", "web-navigation"], "updated_at": "2023-10-23", "url": "https://arxiv.org/abs/2310.08740"}
{"aliases": [], "authors": ["Tianbao Xie", "Fan Zhou", "Zhoujun Cheng", "Peng Shi", "Luoxuan Weng", "Yitao Liu", "Toh Jing Hua", "Junning Zhao", "Qian Liu", "Che Liu", "Leo Z. Liu", "Yiheng Xu", "Hongjin Su", "Dongchan Shin", "Caiming Xiong", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "15d7945fcf1ee4290761d34bbeef9e24bed82bb748bdb7aaa791457c42892fec", "submitted_at": "2023-10-16T17:54:53Z", "updated_at": "2023-10-16T17:54:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2310.10634v1"}}, "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": "2026-10-17T17:11:36Z", "role": "primary", "sha256": "90df698b27a336e2fb7f0ac60a3b8d3e3897c00035ab7968c593a5ab539d177a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "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"}], "id": "arxiv:2310.10634", "identifiers": {"arxiv": "2310.10634"}, "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/2310.10634"}], "name": "OpenAgents: An Open Platform for Language 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. Scope is adjacent: autonomous web browsing is one of three hosted agents, beside data-analysis and API-plugin agents.", "published_at": "2023-10-16", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "OpenAgents hosts three language agents, for data analysis, API plugins and autonomous web browsing, behind a web interface for non-expert users, with local deployment for developers.", "topics": ["web-navigation", "tool-integration"], "updated_at": "2023-10-16", "url": "https://arxiv.org/abs/2310.10634"}
{"aliases": [], "authors": ["Heyi Tao", "Sethuraman T", "Michal Shlapentokh-Rothman", "Derek Hoiem"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7defced3e3beb77a5a6513dbfdff3f102edd922907b523bd062522e54720fa5e", "submitted_at": "2023-10-24T17:57:03Z", "updated_at": "2023-10-25T03:54:11Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2310.16042v2"}}, "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:2310.16042", "identifiers": {"arxiv": "2310.16042"}, "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/2310.16042"}], "name": "WebWISE: Web Interface Control and Sequential Exploration with 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": "2023-10-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebWISE filters DOM elements into observations and generates short programs for click, scroll, and text-entry steps. It evaluates in-context learning from either one supplied example or an example collected during a successful zero-shot MiniWoB++ trial.", "topics": ["web-navigation", "tool-integration", "planning"], "updated_at": "2023-10-25", "url": "https://arxiv.org/abs/2310.16042"}
{"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": ["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": ["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": ["Boyuan Zheng", "Boyu Gou", "Jihyung Kil", "Huan Sun", "Yu Su"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a005fb14b21cf17521f4ea966e555d9963092b709bd9280c6e72a9bf0f5fa93a", "submitted_at": "2024-01-03T08:33:09Z", "updated_at": "2024-03-12T23:14:33Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2401.01614v2"}, "publication": {"authors": ["Boyuan Zheng", "Boyu Gou", "Jihyung Kil", "Huan Sun", "Yu Su"], "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/zheng24e.html"], "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 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": "61349-61385", "pdf_url": "https://raw.githubusercontent.com/mlresearch/v235/main/assets/zheng24e/zheng24e.pdf", "publisher": "PMLR", "series": "Proceedings of Machine Learning Research", "title": "GPT-4V(ision) is a Generalist Web Agent, if Grounded", "type": "conference-paper", "url": "https://proceedings.mlr.press/v235/zheng24e.html", "venue": "Proceedings of the 41st International Conference on Machine Learning", "volume": "235"}}, "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": "2026-10-17T09:53:47+00:00", "role": "primary", "sha256": "eb144b5dfe51e5a9092d273fcffb85343808d011b03d22e059c9cc65e642ef33", "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.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/zheng24e.html"}, {"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:2401.01614", "identifiers": {"arxiv": "2401.01614"}, "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/2401.01614"}, {"rel": "publication", "url": "https://proceedings.mlr.press/v235/zheng24e.html"}, {"rel": "publication_pdf", "url": "https://raw.githubusercontent.com/mlresearch/v235/main/assets/zheng24e/zheng24e.pdf"}], "name": "GPT-4V(ision) is a Generalist Web Agent, if Grounded", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-01-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SeeAct and live-website evaluation, examining how visual planning and element grounding affect web-agent task completion.", "topics": ["web-navigation", "grounding"], "updated_at": "2024-03-12", "url": "https://arxiv.org/abs/2401.01614"}
{"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": ["Jihyung Kil", "Chan Hee Song", "Boyuan Zheng", "Xiang Deng", "Yu Su", "Wei-Lun Chao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3da421fc440f7dcfc53d8ef218f55389e4da4dbac196f54b40402bc877ff0365", "submitted_at": "2024-02-06T23:52:10Z", "updated_at": "2024-03-30T05:18:05Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2402.04476v2"}}, "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:2402.04476", "identifiers": {"arxiv": "2402.04476"}, "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/2402.04476"}], "name": "Dual-View Visual Contextualization for Web 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": "2024-02-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses screenshot geometry and visual content to enrich HTML element representations for web navigation. Nearby elements contribute textual and visual context, with evaluation on cross-task, cross-website, and cross-domain Mind2Web splits.", "topics": ["grounding", "screen-understanding", "web-navigation"], "updated_at": "2024-03-30", "url": "https://arxiv.org/abs/2402.04476"}
{"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": ["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": ["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": ["Hanyu Lai", "Xiao Liu", "Iat Long Iong", "Shuntian Yao", "Yuxuan Chen", "Pengbo Shen", "Hao Yu", "Hanchen Zhang", "Xiaohan Zhang", "Yuxiao Dong", "Jie Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "340206da25df688ff73da617fd66a872586d1e36981c431d18331cee9b9669e0", "submitted_at": "2024-04-04T17:58:40Z", "updated_at": "2024-10-12T05:05:38Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.03648v2"}}, "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:2404.03648", "identifiers": {"arxiv": "2404.03648"}, "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/2404.03648"}], "name": "AutoWebGLM: A Large Language Model-based Web Navigating 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": "2024-04-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds a web agent through simplified HTML observations, human-and-AI browsing data, curriculum training, and reinforcement-learning refinement.", "topics": ["web-navigation", "reinforcement-learning", "data-generation"], "updated_at": "2024-10-12", "url": "https://arxiv.org/abs/2404.03648"}
{"aliases": [], "authors": ["Michael Lutz", "Arth Bohra", "Manvel Saroyan", "Artem Harutyunyan", "Giovanni Campagna"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4c18dbfedcccf6434b3a50a9aa62552372df8b772ccda6a1aa699a8f052476bc", "submitted_at": "2024-04-08T23:10:47Z", "updated_at": "2024-04-08T23:10:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2404.05902v1"}}, "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:2404.05902", "identifiers": {"arxiv": "2404.05902"}, "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/2404.05902"}], "name": "WILBUR: Adaptive In-Context Learning for Robust and Accurate 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": "2024-04-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "WILBUR learns to rank prior task demonstrations for a web agent’s prompt and synthesizes instructions from past runs. A generated task curriculum and a backtracking mechanism support adaptation and recovery, evaluated on WebVoyager.", "topics": ["web-navigation", "memory", "robustness", "data-generation"], "updated_at": "2024-04-08", "url": "https://arxiv.org/abs/2404.05902"}
{"aliases": [], "authors": ["Moghis Fereidouni", "Adib Mosharrof", "A. B. Siddique"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "127d037bb462ae4ed7bd72c46a62e2c4427611b30b67193454e40cc5c83cca3f", "submitted_at": "2024-04-16T20:15:32Z", "updated_at": "2025-01-26T19:07:41Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.10887v2"}}, "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:2404.10887", "identifiers": {"arxiv": "2404.10887"}, "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/2404.10887"}], "name": "Grounded Language Agent for Product Search via Intelligent Web Interactions", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-04-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GLAINTEL, a Flan-T5-based product-search web agent studied under unsupervised learning, demonstration-based training, and unsupervised domain adaptation. It compares ways to combine demonstrations with reinforcement learning.", "topics": ["web-navigation", "reinforcement-learning", "imitation-learning"], "updated_at": "2025-01-26", "url": "https://arxiv.org/abs/2404.10887"}
{"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": ["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": ["Junhee Cho", "Jihoon Kim", "Daseul Bae", "Jinho Choo", "Youngjune Gwon", "Yeong-Dae Kwon"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5f85b2cf7650f41ba9ee6e4ae83cd9041e077a44adcb685e702d644bd5779963", "submitted_at": "2024-06-11T05:21:20Z", "updated_at": "2024-12-26T11:02:28Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2406.06947v3"}}, "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"], "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:2406.06947", "identifiers": {"arxiv": "2406.06947"}, "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/2406.06947"}], "name": "CAAP: Context-Aware Action Planning Prompting to Solve Computer Tasks with Front-End UI Only", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "CAAP converts screenshots into textual observations and plans keyboard-and-mouse actions using a prompt that examines task context from multiple perspectives. The agent is evaluated on MiniWoB++ and WebShop using only front-end visual observations.", "topics": ["planning", "screen-understanding", "web-navigation"], "updated_at": "2024-12-26", "url": "https://arxiv.org/abs/2406.06947"}
{"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": ["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": ["Tamer Abuelsaad", "Deepak Akkil", "Prasenjit Dey", "Ashish Jagmohan", "Aditya Vempaty", "Ravi Kokku"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "18cfd29fcb66d703d90f747c162db1b3fc6a43f400b4eacbddf081bb83ed7947", "submitted_at": "2024-07-17T21:44:28Z", "updated_at": "2024-07-17T21:44:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.13032v1"}}, "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"], "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:2407.13032", "identifiers": {"arxiv": "2407.13032"}, "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/2407.13032"}], "name": "Agent-E: From Autonomous Web Navigation to Foundational Design Principles in Agentic Systems", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Agent-E combines hierarchical web-task control with DOM distillation and observations of interface changes. WebVoyager experiments inform a broader discussion of primitive skills, observation reduction, and self-improvement in agent architecture.", "topics": ["web-navigation", "planning", "efficiency", "state-tracking"], "updated_at": "2024-07-17", "url": "https://arxiv.org/abs/2407.13032"}
{"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": ["Yao Zhang", "Zijian Ma", "Yunpu Ma", "Zhen Han", "Yu Wu", "Volker Tresp"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9e3a34f16101df7a96ffdabbdf86b060e195494f16499f20bdf54110a0394cae", "submitted_at": "2024-08-28T17:49:29Z", "updated_at": "2024-08-28T17:49:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2408.15978v1"}}, "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"], "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:2408.15978", "identifiers": {"arxiv": "2408.15978"}, "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/2408.15978"}], "name": "WebPilot: A Versatile and Autonomous Multi-Agent System for Web Task Execution with Strategic Exploration", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-08-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebPilot separates global subtask planning from local Monte Carlo tree search during web execution. It continually refines the high-level plan and adapts local search to uncertain transitions and incomplete observations, evaluated on WebArena and MiniWoB++.", "topics": ["planning", "web-navigation", "long-horizon"], "updated_at": "2024-08-28", "url": "https://arxiv.org/abs/2408.15978"}
{"aliases": [], "authors": ["Zora Zhiruo Wang", "Jiayuan Mao", "Daniel Fried", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d4c142be3a9c37bbb0d4f401a02de25bd25b4c3968f42ad9a61513abff2b5460", "submitted_at": "2024-09-11T17:21:00Z", "updated_at": "2024-09-11T17:21:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2409.07429v1"}}, "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"], "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:2409.07429", "identifiers": {"arxiv": "2409.07429"}, "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/2409.07429"}], "name": "Agent Workflow Memory", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-09-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Agent Workflow Memory extracts reusable task routines and selectively supplies them to guide later web actions. It studies both offline workflow induction from training examples and online induction from test experience on Mind2Web and WebArena.", "topics": ["memory", "workflow", "web-navigation"], "updated_at": "2024-09-11", "url": "https://arxiv.org/abs/2409.07429"}
{"aliases": [], "authors": ["Mobina Shahbandeh", "Parsa Alian", "Noor Nashid", "Ali Mesbah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5bdbd80e252cfa37dc62d666a202e647f93813549c55dde82daf36afe309c7f2", "submitted_at": "2024-09-16T21:18:39Z", "updated_at": "2024-09-16T21:18:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2409.10741v1"}}, "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:2409.10741", "identifiers": {"arxiv": "2409.10741"}, "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/2409.10741"}], "name": "NaviQAte: Functionality-Guided Web Application 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": "2024-09-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Frames web-application exploration as functionality-oriented question answering, generating navigation actions from multimodal context.", "topics": ["web-navigation", "automation"], "updated_at": "2024-09-16", "url": "https://arxiv.org/abs/2409.10741"}
{"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": ["Brian Tang", "Kang G. Shin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "377d46c738f1a6f3e1c22ea99971a2fcfcf25ded82b2d2a24f728e614e22a1b3", "submitted_at": "2024-09-23T18:06:32Z", "updated_at": "2024-09-23T18:06:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2409.15441v1"}}, "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:2409.15441", "identifiers": {"arxiv": "2409.15441"}, "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/2409.15441"}], "name": "Steward: Natural Language Web Automation", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-09-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Steward, a natural-language browser automation system that repeatedly plans and executes website actions. The paper examines completion detection, state representation, responsiveness, and caching in real-site tasks.", "topics": ["web-navigation", "automation", "efficiency"], "updated_at": "2024-09-23", "url": "https://arxiv.org/abs/2409.15441"}
{"aliases": [], "authors": ["Tianyue Ou", "Frank F. Xu", "Aman Madaan", "Jiarui Liu", "Robert Lo", "Abishek Sridhar", "Sudipta Sengupta", "Dan Roth", "Graham Neubig", "Shuyan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6c19b869b10f0b4c985b88b032f78e0e72f83e7ca8d9f346417054cd2a9002e3", "submitted_at": "2024-09-24T00:51:45Z", "updated_at": "2024-11-27T16:34:52Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2409.15637v2"}, "publication": {"authors": ["Tianyue Ou", "Frank F. Xu", "Aman Madaan", "Jiarui Liu", "Robert Lo", "Abishek Sridhar", "Sudipta Sengupta", "Dan Roth", "Graham Neubig", "Shuyan Zhou"], "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": "10.52202/079017-2908", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/a6a6891cf1dfc64d664f086cf5976e93-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.", "pages": "91618-91652", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/a6a6891cf1dfc64d664f086cf5976e93-Paper-Conference.pdf", "publisher": null, "series": null, "title": "Synatra: Turning Indirect Knowledge into Direct Demonstrations for Digital Agents at Scale", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/a6a6891cf1dfc64d664f086cf5976e93-Abstract-Conference.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": "bf42d07f0959d7ad5ed0f4e00c34ef6cb7e8b1f10f17ac1e7a0ee1951523f821", "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/a6a6891cf1dfc64d664f086cf5976e93-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"], "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:2409.15637", "identifiers": {"arxiv": "2409.15637"}, "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/2409.15637"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/a6a6891cf1dfc64d664f086cf5976e93-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/a6a6891cf1dfc64d664f086cf5976e93-Paper-Conference.pdf"}], "name": "Synatra: Turning Indirect Knowledge into Direct Demonstrations for Digital Agents at Scale", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-09-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Synatra converts indirect procedural knowledge, including human-oriented tutorials, into direct demonstrations for digital agents. It trains a 7B CodeLlama on 100,000 synthesized demonstrations and evaluates the resulting policy on Mind2Web, MiniWoB++, and WebArena.", "topics": ["data-generation", "imitation-learning", "web-navigation"], "updated_at": "2024-11-27", "url": "https://arxiv.org/abs/2409.15637"}
{"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": ["Shikhar Murty", "Hao Zhu", "Dzmitry Bahdanau", "Christopher D. Manning"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5e227f3686b3234745ca1749cb8fb7c2b19383ac9dfebf68976f61f07d1ddf3b", "submitted_at": "2024-10-03T18:56:51Z", "updated_at": "2025-02-05T18:56:51Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2410.02907v2"}}, "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:2410.02907", "identifiers": {"arxiv": "2410.02907"}, "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/2410.02907"}], "name": "NNetNav: Unsupervised Learning of Browser Agents Through Environment Interaction 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": "2024-10-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Generates browser-agent demonstrations through autonomous exploration and retrospective task labeling, pruning trajectories that lack meaningful subtasks.", "topics": ["data-generation", "web-navigation"], "updated_at": "2025-02-05", "url": "https://arxiv.org/abs/2410.02907"}
{"aliases": [], "authors": ["Ke Yang", "Yao Liu", "Sapana Chaudhary", "Rasool Fakoor", "Pratik Chaudhari", "George Karypis", "Huzefa Rangwala"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "09268d51b3d810ffc0024ae332d840afc4501c8f6a2017128fca525be4809a13", "submitted_at": "2024-10-17T17:50:38Z", "updated_at": "2025-05-24T03:55:22Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2410.13825v2"}}, "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:2410.13825", "identifiers": {"arxiv": "2410.13825"}, "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/2410.13825"}], "name": "AgentOccam: A Simple Yet Strong Baseline for 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": "2024-10-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies how simplifying and aligning a web agent's observations and actions with language-model capabilities affects task completion.", "topics": ["web-navigation", "planning"], "updated_at": "2025-05-24", "url": "https://arxiv.org/abs/2410.13825"}
{"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": ["Hongliang He", "Wenlin Yao", "Kaixin Ma", "Wenhao Yu", "Hongming Zhang", "Tianqing Fang", "Zhenzhong Lan", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db2a115281f831ea9333a03fc2a9ebb2f09528eb082060f00a33b125f9ef7762", "submitted_at": "2024-10-25T15:01:27Z", "updated_at": "2024-10-25T15:01:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2410.19609v1"}, "publication": {"authors": ["Hongliang He", "Wenlin Yao", "Kaixin Ma", "Wenhao Yu", "Hongming Zhang", "Tianqing Fang", "Zhenzhong Lan", "Dong Yu"], "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.1336", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.acl-long.1336/"], "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": "27545-27564", "pdf_url": "https://aclanthology.org/2025.acl-long.1336.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "OpenWebVoyager: Building Multimodal Web Agents via Iterative Real-World Exploration, Feedback and Optimization", "type": "conference-paper", "url": "https://aclanthology.org/2025.acl-long.1336/", "venue": "Proceedings of the 63rd 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": "2026-10-18T04:09:55Z", "role": "primary", "sha256": "efd9fe9277a980d42a991abe428ccc3983299c6a8a4d91e18d19cc07a37ed22c", "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.1336/"}, {"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:2410.19609", "identifiers": {"arxiv": "2410.19609"}, "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/2410.19609"}, {"rel": "publication", "url": "https://aclanthology.org/2025.acl-long.1336/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.acl-long.1336.pdf"}], "name": "OpenWebVoyager: Building Multimodal Web Agents via Iterative Real-World Exploration, Feedback and Optimization", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Develops multimodal browser agents through repeated cycles of real-web exploration, trajectory feedback, and policy optimization.", "topics": ["web-navigation", "data-generation"], "updated_at": "2024-10-25", "url": "https://arxiv.org/abs/2410.19609"}
{"aliases": [], "authors": ["Jaekyeom Kim", "Dong-Ki Kim", "Lajanugen Logeswaran", "Sungryull Sohn", "Honglak Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf1fb8aa2e289a72f3c44c99e2d531eaa5eeea08b78ff6f3a21e139d47502b17", "submitted_at": "2024-10-29T21:37:04Z", "updated_at": "2024-10-29T21:37:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2410.22552v1"}}, "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:2410.22552", "identifiers": {"arxiv": "2410.22552"}, "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/2410.22552"}], "name": "Auto-Intent: Automated Intent Discovery and Self-Exploration for Large Language Model 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": "2024-10-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Discovers compact intents from demonstrations and uses intent predictions to guide a language-model web agent's decisions and exploration.", "topics": ["planning", "web-navigation"], "updated_at": "2024-10-29", "url": "https://arxiv.org/abs/2410.22552"}
{"aliases": [], "authors": ["Nalin Tiwary", "Vardhan Dongre", "Sanil Arun Chawla", "Ashwin Lamani", "Dilek Hakkani-Tür"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1c67b1559410152607fc9d54abf25c37ef939c9f755f089bdf87d4f39393c085", "submitted_at": "2024-10-31T01:51:41Z", "updated_at": "2024-10-31T01:51:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2410.23555v1"}}, "environments": ["web"], "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": "fced2234354232b8834eef0cf2557b2932990e74b3530fc95459cd3fa0e11792", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2410.23555v1"}], "id": "arxiv:2410.23555", "identifiers": {"arxiv": "2410.23555"}, "kind": "paper", "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/2410.23555"}], "name": "From Context to Action: Analysis of the Impact of State Representation and Context on the Generalization of Multi-Turn Web Navigation Agents", "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": "2024-10-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Analyzes how interaction history and webpage representation affect multiturn web-agent generalization across unfamiliar websites, task categories, and geographic settings.", "topics": ["web-navigation", "interaction-history", "state-tracking"], "updated_at": "2024-10-31", "url": "https://arxiv.org/abs/2410.23555"}
{"aliases": [], "authors": ["Zehan Qi", "Xiao Liu", "Iat Long Iong", "Hanyu Lai", "Xueqiao Sun", "Wenyi Zhao", "Yu Yang", "Xinyue Yang", "Jiadai Sun", "Shuntian Yao", "Tianjie Zhang", "Wei Xu", "Jie Tang", "Yuxiao Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db039a2592cd79a1f8f7f1c985c2ddf42447c63d1f58d2815ba1d6ef2369da47", "submitted_at": "2024-11-04T17:59:58Z", "updated_at": "2025-01-27T11:56:15Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2411.02337v3"}}, "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:2411.02337", "identifiers": {"arxiv": "2411.02337"}, "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/2411.02337"}], "name": "WebRL: Training LLM Web Agents via Self-Evolving Online Curriculum 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": "2024-11-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains web agents with an evolving task curriculum, outcome supervision, and reinforcement-learning updates intended to manage policy drift.", "topics": ["reinforcement-learning", "web-navigation"], "updated_at": "2025-01-27", "url": "https://arxiv.org/abs/2411.02337"}
{"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": ["Gaurav Verma", "Rachneet Kaur", "Nishan Srishankar", "Zhen Zeng", "Tucker Balch", "Manuela Veloso"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "94d908dd89ff27434805fe7915c5a93d5c2b1d873449babc1c566c5582cca734", "submitted_at": "2024-11-20T16:54:15Z", "updated_at": "2024-11-20T16:54:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2411.13451v1"}, "publication": {"authors": ["Gaurav Verma", "Rachneet Kaur", "Nishan Srishankar", "Zhen Zeng", "Tucker Balch", "Manuela Veloso"], "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.1008", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.acl-long.1008/"], "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": "20635-20651", "pdf_url": "https://aclanthology.org/2025.acl-long.1008.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "AdaptAgent: Adapting Multimodal Web Agents with Few-Shot Learning from Human Demonstrations", "type": "conference-paper", "url": "https://aclanthology.org/2025.acl-long.1008/", "venue": "Proceedings of the 63rd 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": "2026-10-18T04:09:56Z", "role": "primary", "sha256": "e8930e90d2d78f0d6ef35acc487ba355d3798cdafbcc1d2637258bad81407d41", "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.1008/"}, {"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:2411.13451", "identifiers": {"arxiv": "2411.13451"}, "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/2411.13451"}, {"rel": "publication", "url": "https://aclanthology.org/2025.acl-long.1008/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.acl-long.1008.pdf"}], "name": "AdaptAgent: Adapting Multimodal Web Agents with Few-Shot Learning from Human Demonstrations", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-11-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Adapts multimodal web agents to unfamiliar websites using a small number of human demonstrations through prompting or meta-learning.", "topics": ["web-navigation", "data-generation"], "updated_at": "2024-11-20", "url": "https://arxiv.org/abs/2411.13451"}
{"aliases": [], "authors": ["Junhong Shen", "Atishay Jain", "Zedian Xiao", "Ishan Amlekar", "Mouad Hadji", "Aaron Podolny", "Ameet Talwalkar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ef5e3848c3395a5620c5af422e6ebf198e405cfde1314cfac50d347daf68a429", "submitted_at": "2024-11-22T15:26:23Z", "updated_at": "2024-12-05T02:00:07Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2411.15004v2"}}, "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:2411.15004", "identifiers": {"arxiv": "2411.15004"}, "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/2411.15004"}], "name": "ScribeAgent: Towards Specialized Web Agents Using Production-Scale Workflow 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": "2024-11-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies fine-tuning open language models on production web-workflow traces, including training choices for specialized web agents.", "topics": ["workflow", "web-navigation"], "updated_at": "2024-12-05", "url": "https://arxiv.org/abs/2411.15004"}
{"aliases": [], "authors": ["Yiheng Xu", "Dunjie Lu", "Zhennan Shen", "Junli Wang", "Zekun Wang", "Yuchen Mao", "Caiming Xiong", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8ee630d434cef63f63c368752aeae14b0a60d49c3fde33b3a79ecbfc8fc4d06e", "submitted_at": "2024-12-12T18:59:27Z", "updated_at": "2025-03-03T18:59:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2412.09605v2"}}, "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:2412.09605", "identifiers": {"arxiv": "2412.09605"}, "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/2412.09605"}], "name": "AgentTrek: Agent Trajectory Synthesis via Guiding Replay with Web Tutorials", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-12-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Converts public tutorials into task specifications and executes them with an agent to synthesize multimodal interaction trajectories.", "topics": ["data-generation", "web-navigation"], "updated_at": "2025-03-03", "url": "https://arxiv.org/abs/2412.09605"}
{"aliases": [], "authors": ["Jiarun Liu", "Jia Hao", "Chunhong Zhang", "Zheng Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e2abbae042d20381aa218e6a2b14a1fc28074543e430ad09031237d22587be7a", "submitted_at": "2024-12-14T08:25:28Z", "updated_at": "2024-12-14T08:25:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2412.10742v1"}}, "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:2412.10742", "identifiers": {"arxiv": "2412.10742"}, "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/2412.10742"}], "name": "WEPO: Web Element Preference Optimization for LLM-based Web 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": "2024-12-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses web-element preference optimization with sampled negative HTML elements to align high-level user intent with navigation actions.", "topics": ["grounding", "web-navigation"], "updated_at": "2024-12-14", "url": "https://arxiv.org/abs/2412.10742"}
{"aliases": [], "authors": ["Yifei Zhou", "Qianlan Yang", "Kaixiang Lin", "Min Bai", "Xiong Zhou", "Yu-Xiong Wang", "Sergey Levine", "Erran Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "362d2a470603bece43d2205e56dd633e7310cfad2ba2d91b0b765abcbcc41b2f", "submitted_at": "2024-12-17T18:59:50Z", "updated_at": "2024-12-17T18:59:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2412.13194v1"}}, "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:2412.13194", "identifiers": {"arxiv": "2412.13194"}, "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/2412.13194"}], "name": "Proposer-Agent-Evaluator(PAE): Autonomous Skill Discovery For Foundation Model 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": "2024-12-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines autonomous task proposal, web interaction, and visual success evaluation to discover skills and improve browser-agent policies.", "topics": ["data-generation", "reinforcement-learning", "web-navigation"], "updated_at": "2024-12-17", "url": "https://arxiv.org/abs/2412.13194"}
{"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": ["Ruichen Zhang", "Mufan Qiu", "Zhen Tan", "Mohan Zhang", "Vincent Lu", "Jie Peng", "Kaidi Xu", "Leandro Z. Agudelo", "Peter Qian", "Tianlong Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0535de4778b6a31adb64f833733ebf94acf54ba872d98afd40b3e1012fb737fd", "submitted_at": "2025-02-11T20:41:49Z", "updated_at": "2025-03-06T19:40:57Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2502.07942v2"}}, "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:2502.07942", "identifiers": {"arxiv": "2502.07942"}, "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/2502.07942"}], "name": "Symbiotic Cooperation for Web Agents: Harnessing Complementary Strengths of Large and Small LLMs", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Couples large- and small-model web-agent exploration with trajectory synthesis and distillation to improve both model types iteratively.", "topics": ["data-generation", "web-navigation", "efficiency"], "updated_at": "2025-03-06", "url": "https://arxiv.org/abs/2502.07942"}
{"aliases": [], "authors": ["Vardaan Pahuja", "Yadong Lu", "Corby Rosset", "Boyu Gou", "Arindam Mitra", "Spencer Whitehead", "Yu Su", "Ahmed Awadallah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d9f459a02b7a464d9e1e49579943d1b93d46308c1b00e795c903d2fa1502a2b2", "submitted_at": "2025-02-17T02:13:48Z", "updated_at": "2025-05-30T02:30:57Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2502.11357v4"}, "publication": {"authors": ["Vardaan Pahuja", "Yadong Lu", "Corby Rosset", "Boyu Gou", "Arindam Mitra", "Spencer Whitehead", "Yu Su", "Ahmed Awadallah"], "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.326", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.findings-acl.326/", "https://aclanthology.org/2025.findings-acl.326.pdf"], "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. The publisher PDF title and complete first-page author list were visually checked.", "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. Authors follow the publisher PDF, page 1. Landing-page citation metadata gives Ahmed Hassan for the final author, while the supplied BibTeX expands the name to Ahmed Hassan Awadallah; the PDF uses Ahmed Awadallah.", "pages": "6300-6323", "pdf_url": "https://aclanthology.org/2025.findings-acl.326.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "Explorer: Scaling Exploration-driven Web Trajectory Synthesis for Multimodal Web Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.findings-acl.326/", "venue": "Findings of the Association for Computational Linguistics: ACL 2025", "volume": null}}, "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": "2026-10-18T04:09:58Z", "role": "primary", "sha256": "8ec40afe322b7c72c9f6d6db617fa2d4fae559cf6cd4e2fcc0ff5a9cc9485e05", "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.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.findings-acl.326/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:14:02.000475+00:00", "role": "primary", "sha256": "c1b232e161839514713789c408ae64f66684b49770033c7f3b04bf1d1a232157", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes"], "url": "https://aclanthology.org/2025.findings-acl.326.pdf"}, {"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:2502.11357", "identifiers": {"arxiv": "2502.11357"}, "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/2502.11357"}, {"rel": "publication", "url": "https://aclanthology.org/2025.findings-acl.326/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.findings-acl.326.pdf"}], "name": "Explorer: Scaling Exploration-driven Web Trajectory Synthesis for Multimodal 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": "2025-02-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Synthesizes multimodal web trajectories through exploration and refinement, then trains a browser agent to study the effect of scaling interaction data.", "topics": ["data-generation", "web-navigation"], "updated_at": "2025-05-30", "url": "https://arxiv.org/abs/2502.11357"}
{"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": ["Danqing Zhang", "Balaji Rama", "Jingyi Ni", "Shiying He", "Fu Zhao", "Kunyu Chen", "Arnold Chen", "Junyu Cao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "df682f883bed5e0e9917074dfa130839fabaa318c55714f26943d944774854fe", "submitted_at": "2025-03-04T19:13:10Z", "updated_at": "2025-05-06T06:42:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2503.02950v2"}, "publication": {"authors": ["Danqing Zhang", "Balaji Rama", "Jingyi Ni", "Shiying He", "Fu Zhao", "Kunyu Chen", "Arnold Chen", "Junyu Cao"], "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-demo.36", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.naacl-demo.36/"], "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": "449-455", "pdf_url": "https://aclanthology.org/2025.naacl-demo.36.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "LiteWebAgent: The Open-Source Suite for VLM-Based Web-Agent Applications", "type": "conference-paper", "url": "https://aclanthology.org/2025.naacl-demo.36/", "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": "2026-10-16T06:36:54Z", "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": "2026-10-18T04:09:58Z", "role": "primary", "sha256": "9b130228c049f324632f571dd9c9e74c9a293ee240351c7a6fd4fef3359afc12", "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-demo.36/"}, {"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:2503.02950", "identifiers": {"arxiv": "2503.02950"}, "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/2503.02950"}, {"rel": "publication", "url": "https://aclanthology.org/2025.naacl-demo.36/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.naacl-demo.36.pdf"}], "name": "LiteWebAgent: The Open-Source Suite for VLM-Based Web-Agent 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": "2025-03-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a modular visual web-agent stack with planning and memory components, exposed through a remote-browser application and Chrome extension.", "topics": ["web-navigation", "tool-integration", "planning"], "updated_at": "2025-05-06", "url": "https://arxiv.org/abs/2503.02950"}
{"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": ["Liangbo Ning", "Ziran Liang", "Zhuohang Jiang", "Haohao Qu", "Yujuan Ding", "Wenqi Fan", "Xiao-yong Wei", "Shanru Lin", "Hui Liu", "Philip S. Yu", "Qing Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4b214bd8e5a34c0e8ca16a3be4fcf638a59a17ca524276dbef3c6b75f08398bb", "submitted_at": "2025-03-30T08:15:44Z", "updated_at": "2025-08-05T11:47:41Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2503.23350v4"}}, "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:2503.23350", "identifiers": {"arxiv": "2503.23350"}, "kind": "survey", "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/2503.23350"}], "name": "A Survey of WebAgents: Towards Next-Generation AI Agents for Web Automation with Large Foundation Models", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-03-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Surveys foundation-model web agents through their architectures, training methods and trustworthiness. It organizes research on completing browser tasks from user instructions and identifies directions for further development.", "topics": ["survey", "web-navigation", "safety-security"], "updated_at": "2025-08-05", "url": "https://arxiv.org/abs/2503.23350"}
{"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": ["Boyuan Zheng", "Michael Y. Fatemi", "Xiaolong Jin", "Zora Zhiruo Wang", "Apurva Gandhi", "Yueqi Song", "Yu Gu", "Jayanth Srinivasa", "Gaowen Liu", "Graham Neubig", "Yu Su"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5f0489433a41b49ab8ab42eea57503685ba60beadabea7822134e96d865a7435", "submitted_at": "2025-04-09T17:51:50Z", "updated_at": "2025-04-09T17:51:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.07079v1"}}, "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:2504.07079", "identifiers": {"arxiv": "2504.07079"}, "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/2504.07079"}], "name": "SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skills", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "SkillWeaver explores websites, practices interactions, and converts the resulting experience into a reusable library of callable skill APIs. Its WebArena and live-website experiments examine skill refinement, composition, and transfer from stronger agents to weaker ones.", "topics": ["web-navigation", "tool-integration", "planning"], "updated_at": "2025-04-09", "url": "https://arxiv.org/abs/2504.07079"}
{"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": ["Zhisong Zhang", "Tianqing Fang", "Kaixin Ma", "Wenhao Yu", "Hongming Zhang", "Haitao Mi", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "76f84d5375ab7321512a1f4f8724c011cf180bf5a7a0d532ff3218cc9d7a6e75", "submitted_at": "2025-04-16T05:41:20Z", "updated_at": "2026-01-15T02:00:45Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2504.11788v3"}}, "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:2504.11788", "identifiers": {"arxiv": "2504.11788"}, "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/2504.11788"}], "name": "WebRollback: Enhancing Web Agents with Explicit Rollback Mechanisms", "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": "WebRollback gives a web agent an explicit action for returning to an earlier state in its navigation trajectory. It studies this recovery mechanism on two live web-navigation benchmarks under both zero-shot and fine-tuned settings.", "topics": ["web-navigation", "planning", "robustness"], "updated_at": "2026-01-15", "url": "https://arxiv.org/abs/2504.11788"}
{"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": ["Tianqing Fang", "Hongming Zhang", "Zhisong Zhang", "Kaixin Ma", "Wenhao Yu", "Haitao Mi", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "86ca58c92ffc2cbbcb0297d8c7618dfab966e36a07b141c068998b0f2934d11e", "submitted_at": "2025-04-23T02:54:31Z", "updated_at": "2025-08-21T05:55:43Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.21024v2"}}, "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:2504.21024", "identifiers": {"arxiv": "2504.21024"}, "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/2504.21024"}], "name": "WebEvolver: Enhancing Web Agent Self-Improvement with Coevolving World Model", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-04-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebEvolver trains a co-evolving world model to predict web observations and generate additional agent-training trajectories. The same model supports look-ahead action selection at inference, with evaluation on live web-task benchmarks.", "topics": ["world-models", "data-generation", "web-navigation", "planning"], "updated_at": "2025-08-21", "url": "https://arxiv.org/abs/2504.21024"}
{"aliases": [], "authors": ["Xiaoxi Li", "Jiajie Jin", "Guanting Dong", "Hongjin Qian", "Yongkang Wu", "Ji-Rong Wen", "Yutao Zhu", "Zhicheng Dou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "351b6d7fbe4e217a5211884832c6ad07d76a0463e6a8c996815999ff3a5f1921", "submitted_at": "2025-04-30T16:25:25Z", "updated_at": "2025-10-13T12:40:15Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.21776v2"}}, "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"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T13:00:06.851520+00:00", "role": "primary", "sha256": "d06baa20c7775347cbb4492473418d11cbe56941ab90baa16381353ab6c39e07", "supports": ["summary", "scope", "notes"], "url": "https://arxiv.org/html/2504.21776v2"}], "id": "arxiv:2504.21776", "identifiers": {"arxiv": "2504.21776"}, "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/2504.21776"}], "name": "WebThinker: Empowering Large Reasoning Models with Deep Research Capability", "notes": "Adjacent web-research method. The retained abstract and full-text Section 3 and Appendix A describe search APIs and Crawl4AI URL retrieval; no visual GUI execution is inferred. Results were not independently reproduced.", "published_at": "2025-04-30", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Combines iterative web search and link-following with interleaved reasoning and report drafting for deep research. Its reported web explorer retrieves page content through a crawler rather than controlling a visual browser interface.", "topics": ["web-navigation", "planning", "tool-integration"], "updated_at": "2025-10-13", "url": "https://arxiv.org/abs/2504.21776"}
{"aliases": [], "authors": ["Tiange Luo", "Lajanugen Logeswaran", "Justin Johnson", "Honglak Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3d2dc8ca397adce1ecf6cf53afa82a8d4f1825bf00277aeb117966f965abfe48", "submitted_at": "2025-05-01T17:45:59Z", "updated_at": "2025-07-14T17:45:50Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.00684v2"}}, "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:2505.00684", "identifiers": {"arxiv": "2505.00684"}, "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/2505.00684"}], "name": "Visual Test-time Scaling for GUI Agent 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": "2025-05-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "RegionFocus adds visual test-time scaling by zooming into selected screenshot regions and maintaining an image-based map of landmarks and actions. It evaluates this grounding procedure with UI-TARS and Qwen2.5-VL on ScreenSpot-Pro and WebVoyager.", "topics": ["grounding", "screen-understanding", "web-navigation"], "updated_at": "2025-07-14", "url": "https://arxiv.org/abs/2505.00684"}
{"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": ["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": ["Zhepei Wei", "Wenlin Yao", "Yao Liu", "Weizhi Zhang", "Qin Lu", "Liang Qiu", "Changlong Yu", "Puyang Xu", "Chao Zhang", "Bing Yin", "Hyokun Yun", "Lihong Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f311fb50b36c234787e652b5689246460153a8f442e83b312048cdf6a4fb1502", "submitted_at": "2025-05-22T09:07:43Z", "updated_at": "2025-10-08T21:16:43Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.16421v2"}}, "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:2505.16421", "identifiers": {"arxiv": "2505.16421"}, "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/2505.16421"}], "name": "WebAgent-R1: Training Web Agents via End-to-End Multi-Turn 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": "2025-05-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains web agents through asynchronous multi-turn environment interaction, using task-success rewards to optimize complete trajectories.", "topics": ["reinforcement-learning", "web-navigation"], "updated_at": "2025-10-08", "url": "https://arxiv.org/abs/2505.16421"}
{"aliases": [], "authors": ["Minda Hu", "Tianqing Fang", "Jianshu Zhang", "Junyu Ma", "Zhisong Zhang", "Jingyan Zhou", "Hongming Zhang", "Haitao Mi", "Dong Yu", "Irwin King"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "171071f26a6dc8b9c8851d2f11d60b113983191c905631e431f352711e0d0e26", "submitted_at": "2025-05-26T14:03:37Z", "updated_at": "2025-09-18T11:32:15Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.20013v2"}}, "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:2505.20013", "identifiers": {"arxiv": "2505.20013"}, "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/2505.20013"}], "name": "WebCoT: Enhancing Web Agent Reasoning by Reconstructing Chain-of-Thought in Reflection, Branching, and Rollback", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-05-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebCoT turns reflection, look-ahead, branching and rollback procedures into chain-of-thought training rationales. Fine-tuning on these reconstructed trajectories is evaluated for web-agent self-improvement and transfer across navigation tasks.", "topics": ["data-generation", "planning", "self-reflection", "web-navigation"], "updated_at": "2025-09-18", "url": "https://arxiv.org/abs/2505.20013"}
{"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": ["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": ["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": ["Apurva Gandhi", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2bccee69e390b5170ff264e689c209d13317e7d3be9b37e6d3fbed0e0d089943", "submitted_at": "2025-06-04T03:27:56Z", "updated_at": "2026-03-02T22:31:00Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.03533v2"}}, "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:2506.03533", "identifiers": {"arxiv": "2506.03533"}, "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/2506.03533"}], "name": "Go-Browse: Training Web Agents with Structured Exploration", "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": "Go-Browse frames web-environment exploration as graph search so that information can be reused across data-collection episodes. It collects 10,000 successful trajectories and 40,000 interaction steps on WebArena and uses them to fine-tune a web agent.", "topics": ["data-generation", "web-navigation"], "updated_at": "2026-03-02", "url": "https://arxiv.org/abs/2506.03533"}
{"aliases": [], "authors": ["Junhong Shen", "Hao Bai", "Lunjun Zhang", "Yifei Zhou", "Amrith Setlur", "Shengbang Tong", "Diego Caples", "Nan Jiang", "Tong Zhang", "Ameet Talwalkar", "Aviral Kumar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e4b6c8b297e88dae689ba2a48724a441698cd5309a3bbe1198f38343e88aceb1", "submitted_at": "2025-06-09T17:50:02Z", "updated_at": "2025-06-10T12:50:18Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.07976v2"}}, "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:2506.07976", "identifiers": {"arxiv": "2506.07976"}, "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/2506.07976"}], "name": "Thinking vs. Doing: Agents that Reason by Scaling Test-Time Interaction", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Test-Time Interaction scales the interaction horizon so web agents can explore, backtrack and revise plans within a rollout. Its online reinforcement-learning curriculum adjusts rollout lengths and is evaluated on WebVoyager and WebArena.", "topics": ["reinforcement-learning", "web-navigation", "planning", "long-horizon"], "updated_at": "2025-06-10", "url": "https://arxiv.org/abs/2506.07976"}
{"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": ["Yuquan Xie", "Zaijing Li", "Rui Shao", "Gongwei Chen", "Kaiwen Zhou", "Yinchuan Li", "Dongmei Jiang", "Liqiang Nie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b353606ce29a5bc60d263e86827a825af46883e25da5bd290b48affa9dd3838", "submitted_at": "2025-06-12T06:21:19Z", "updated_at": "2025-06-12T06:21:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.10387v1"}}, "environments": ["android", "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:2506.10387", "identifiers": {"arxiv": "2506.10387"}, "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/2506.10387"}], "name": "Mirage-1: Augmenting and Updating GUI Agent with Hierarchical Multimodal Skills", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Mirage-1 organizes interaction trajectories into execution, core, and meta-skills, then uses them to guide Monte Carlo tree search during online GUI tasks. It introduces AndroidLH for long-horizon evaluation and tests the approach on Android and web-navigation benchmarks.", "topics": ["planning", "memory", "long-horizon", "mobile-navigation", "web-navigation"], "updated_at": "2025-06-12", "url": "https://arxiv.org/abs/2506.10387"}
{"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": ["Imene Kerboua", "Sahar Omidi Shayegan", "Megh Thakkar", "Xing Han Lù", "Massimo Caccia", "Véronique Eglin", "Alexandre Aussem", "Jérémy Espinas", "Alexandre Lacoste"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b8972688fb95908c024a85e3ffc258be61280ee8d22e6ab97dfb0c4997b83445", "submitted_at": "2025-06-30T19:24:45Z", "updated_at": "2025-06-30T19:24:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.00210v1"}}, "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:2507.00210", "identifiers": {"arxiv": "2507.00210"}, "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/2507.00210"}], "name": "LineRetriever: Planning-Aware Observation Reduction 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": "2025-06-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "LineRetriever uses a language model to select observation lines that matter for upcoming web-navigation actions. It reduces DOM or accessibility-tree context by considering the planning horizon instead of relying solely on semantic similarity.", "topics": ["web-navigation", "efficiency", "planning"], "updated_at": "2025-06-30", "url": "https://arxiv.org/abs/2507.00210"}
{"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": ["Gongwei Chen", "Xurui Zhou", "Rui Shao", "Yibo Lyu", "Kaiwen Zhou", "Shuai Wang", "Wentao Li", "Yinchuan Li", "Zhongang Qi", "Liqiang Nie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "774e2515b563f75d4561daaa6a4b672423d917a3f50e98308ab448e2ae0731e5", "submitted_at": "2025-07-04T17:37:15Z", "updated_at": "2025-07-04T17:37:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.03730v1"}}, "environments": ["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:2507.03730", "identifiers": {"arxiv": "2507.03730"}, "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/2507.03730"}], "name": "Less is More: Empowering GUI Agent with Context-Aware Simplification", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "SimpAgent prunes distracting interface elements with visual masks and compresses interaction history using consistency-based guidance. The study evaluates these two forms of context reduction during GUI navigation in web and mobile environments.", "topics": ["efficiency", "interaction-history", "web-navigation", "mobile-navigation"], "updated_at": "2025-07-04", "url": "https://arxiv.org/abs/2507.03730"}
{"aliases": [], "authors": ["Joonhyung Park", "Peng Tang", "Sagnik Das", "Srikar Appalaraju", "Kunwar Yashraj Singh", "R. Manmatha", "Shabnam Ghadar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aff9b208c446bd57f24258734557e0cd5caaf9146af68e594ec6a7e575919a6f", "submitted_at": "2025-07-08T04:56:57Z", "updated_at": "2025-07-08T04:56:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.05673v1"}}, "environments": ["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:2507.05673", "identifiers": {"arxiv": "2507.05673"}, "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/2507.05673"}], "name": "R-VLM: Region-Aware Vision Language Model for Precise 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": "2025-07-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "R-VLM localizes GUI elements through zoomed-in region proposals and an intersection-over-union-aware training objective. It evaluates the grounding model on ScreenSpot and AgentStudio and studies its effect on navigation in AITW and Mind2Web.", "topics": ["grounding", "screen-understanding", "web-navigation"], "updated_at": "2025-07-08", "url": "https://arxiv.org/abs/2507.05673"}
{"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": ["Hussein Mozannar", "Gagan Bansal", "Cheng Tan", "Adam Fourney", "Victor Dibia", "Jingya Chen", "Jack Gerrits", "Tyler Payne", "Matheus Kunzler Maldaner", "Madeleine Grunde-McLaughlin", "Eric Zhu", "Griffin Bassman", "Jacob Alber", "Peter Chang", "Ricky Loynd", "Friederike Niedtner", "Ece Kamar", "Maya Murad", "Rafah Hosn", "Saleema Amershi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c9416ede89036d640412496b5f0def218b69271c5f2449689d4e4afacb43c556", "submitted_at": "2025-07-30T03:49:14Z", "updated_at": "2025-07-30T03:49:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.22358v1"}}, "environments": ["web"], "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:2507.22358", "identifiers": {"arxiv": "2507.22358"}, "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/2507.22358"}], "name": "Magentic-UI: Towards Human-in-the-loop Agentic Systems", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Magentic-UI combines browser interaction, code execution, file operations, and MCP tools in an interface for human-agent collaboration. It studies collaborative planning and task execution, action guards, and persistent memory through benchmark tasks, simulated users, user studies, and safety assessments.", "topics": ["human-in-the-loop", "tool-integration", "web-navigation", "safety-security"], "updated_at": "2025-07-30", "url": "https://arxiv.org/abs/2507.22358"}
{"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": ["Thassilo M. Schiepanski", "Nicholas Piël"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "789cfb56a83974c38e6f5e359dac5b6b3075cb70ac8afc178e36a901ce742041", "submitted_at": "2025-08-06T12:56:54Z", "updated_at": "2026-08-31T14:39:02Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2508.04412v4"}}, "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:2508.04412", "identifiers": {"arxiv": "2508.04412"}, "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/2508.04412"}], "name": "Beyond Pixels: Exploring DOM Downsampling for 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": "2025-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "D2Snap reduces serialized DOM observations while preserving actionable elements and features that distinguish their uses. The study compares reduced DOM inputs with screenshot-based observations for a web agent on tasks sampled from Online-Mind2Web.", "topics": ["web-navigation", "efficiency", "screen-understanding"], "updated_at": "2026-08-31", "url": "https://arxiv.org/abs/2508.04412"}
{"aliases": [], "authors": ["Jiarun Liu", "Chunhong Zhang", "Zheng Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2d574d4157b775639c08ed12813892584b011dd3c0271ca2586f9d6a61d00c14", "submitted_at": "2025-08-07T07:05:22Z", "updated_at": "2025-08-07T07:05:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.05081v1"}}, "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:2508.05081", "identifiers": {"arxiv": "2508.05081"}, "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/2508.05081"}], "name": "Cognitive Duality for Adaptive 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": "2025-08-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CogniWeb, which switches between fast reactive processing and slower deliberative reasoning according to task complexity. WebArena experiments examine this combination of learned behaviors and online planning.", "topics": ["planning", "efficiency", "web-navigation"], "updated_at": "2025-08-07", "url": "https://arxiv.org/abs/2508.05081"}
{"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": ["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": ["Tanvir Bhathal", "Asanshay Gupta"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "67da98d98fec647de9b482fabd8c7861063c0fdb91f1ec303fe8df759ee0f69d", "submitted_at": "2025-08-23T11:02:59Z", "updated_at": "2025-08-23T11:02:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.16987v1"}}, "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:2508.16987", "identifiers": {"arxiv": "2508.16987"}, "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/2508.16987"}], "name": "WebSight: A Vision-First Architecture for Robust 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": "2025-08-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebSight uses a vision-language model trained on web-interface examples within a modular agent with planning, action, verification, and episodic memory components. It evaluates element clicking on Showdown Clicks and full browsing tasks on WebVoyager using visual observations.", "topics": ["web-navigation", "grounding", "planning", "memory"], "updated_at": "2025-08-23", "url": "https://arxiv.org/abs/2508.16987"}
{"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": ["Shuning Zhang", "Yutong Jiang", "Rongjun Ma", "Yuting Yang", "Mingyao Xu", "Zhixin Huang", "Xin Yi", "Hewu Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f37ebc10535f96d17d28990e608a1f2eb6cc004f40d58b3288c4a81d781160ea", "submitted_at": "2025-09-15T13:58:52Z", "updated_at": "2025-09-15T13:58:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.11939v1"}}, "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:2509.11939", "identifiers": {"arxiv": "2509.11939"}, "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/2509.11939"}], "name": "PrivWeb: Unobtrusive and Content-aware Privacy Protection For Web Agents", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-09-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces PrivWeb, which uses a local language model to anonymize interface information according to user preferences. Adaptive notifications and selective pauses give users control over sensitive information access during web-agent tasks.", "topics": ["safety-security", "human-in-the-loop", "web-navigation"], "updated_at": "2025-09-15", "url": "https://arxiv.org/abs/2509.11939"}
{"aliases": [], "authors": ["Ziyuan Chen", "Zhenghui Zhao", "Zhangye Han", "Miancan Liu", "Xianhang Ye", "Yiqing Li", "Hongbo Min", "Jinkui Ren", "Xiantao Zhang", "Guitao Cao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cc1e4f78f1d3ffdb1b67f7c0bb645561a76db19dfa5f442fa9aa0a0d4d1ae0de", "submitted_at": "2025-09-17T16:58:44Z", "updated_at": "2025-09-19T02:13:09Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.14172v2"}}, "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:2509.14172", "identifiers": {"arxiv": "2509.14172"}, "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/2509.14172"}], "name": "TGPO: Tree-Guided Preference Optimization for Robust Web Agent 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": "2025-09-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "TGPO merges semantically equivalent web states into tree-structured trajectories for offline preference optimization. A process reward model scores subgoal progress, redundant behavior, and action validity, while dynamic weighting emphasizes consequential decisions.", "topics": ["reinforcement-learning", "reward-models", "web-navigation"], "updated_at": "2025-09-19", "url": "https://arxiv.org/abs/2509.14172"}
{"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": ["Ran Xu", "Kaixin Ma", "Wenhao Yu", "Hongming Zhang", "Joyce C. Ho", "Carl Yang", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e1c6fd3e84212bac4c4c8341ffb3478c97becce08f92dac89e5de77ad42cd07d", "submitted_at": "2025-09-29T02:04:20Z", "updated_at": "2025-09-29T02:04:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.24183v1"}}, "environments": ["web", "android"], "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": "2026-10-17T09:20:58+00:00", "role": "primary", "sha256": "fe9f70994c119a85cb8075461174e9699c423d79094e611f1ac95a51666600c1", "supports": ["summary", "environments", "scope"], "url": "https://arxiv.org/html/2509.24183v1"}], "id": "arxiv:2509.24183", "identifiers": {"arxiv": "2509.24183"}, "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/2509.24183"}], "name": "Retrieval-augmented GUI Agents with Generative Guidelines", "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": "RAG-GUI trains a small guidance adapter to filter and summarize retrieved web tutorials while keeping the acting agent fixed. Supervised and rejection-sampling stages prepare the adapter for inference-time assistance, tested on offline web/mobile tasks and interactive Android tasks.", "topics": ["memory", "planning", "imitation-learning", "web-navigation", "mobile-navigation"], "updated_at": "2025-09-29", "url": "https://arxiv.org/abs/2509.24183"}
{"aliases": [], "authors": ["Siru Ouyang", "Jun Yan", "I-Hung Hsu", "Yanfei Chen", "Ke Jiang", "Zifeng Wang", "Rujun Han", "Long T. Le", "Samira Daruki", "Xiangru Tang", "Vishy Tirumalashetty", "George Lee", "Mahsan Rofouei", "Hangfei Lin", "Jiawei Han", "Chen-Yu Lee", "Tomas Pfister"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d984883b65895d02debd54edbd198f0aadf8bd295da869680c3ebbe08d2e9492", "submitted_at": "2025-09-29T17:51:03Z", "updated_at": "2026-03-16T20:49:28Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.25140v2"}}, "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:2509.25140", "identifiers": {"arxiv": "2509.25140"}, "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/2509.25140"}], "name": "ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Distills generalizable reasoning strategies from both successful and failed agent trajectories into a persistent memory bank for test-time retrieval.", "topics": ["memory", "self-reflection", "web-navigation"], "updated_at": "2026-03-16", "url": "https://arxiv.org/abs/2509.25140"}
{"aliases": [], "authors": ["Viraj Prabhu", "Yutong Dai", "Matthew Fernandez", "Jing Gu", "Krithika Ramakrishnan", "Yanqi Luo", "Silvio Savarese", "Caiming Xiong", "Junnan Li", "Zeyuan Chen", "Ran Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "23794419031fe219fc342c041d666045080b274926dc22ab101a65581b4e4efd", "submitted_at": "2025-10-01T23:41:47Z", "updated_at": "2025-10-01T23:41:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.01524v1"}}, "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"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:27:51+00:00", "role": "primary", "sha256": "8466eed10f7f630a68fa742ff9de6b5ba052c497ff421acd4fbee529093d3a58", "supports": ["summary", "scope", "environments"], "url": "https://arxiv.org/html/2510.01524v1"}], "id": "arxiv:2510.01524", "identifiers": {"arxiv": "2510.01524"}, "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/2510.01524"}], "name": "WALT: Web Agents that Learn Tools", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "WALT explores website functionality and converts demonstrated interactions into validated, reusable tools combining DOM actions, URL navigation, and agentic fallbacks. It evaluates this expanded browser action space on WebArena and VisualWebArena.", "topics": ["web-navigation", "tool-integration", "efficiency", "automation"], "updated_at": "2025-10-01", "url": "https://arxiv.org/abs/2510.01524"}
{"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": ["Hadi Nekoei", "Aman Jaiswal", "Patrice Bechard", "Oleh Shliazhko", "Orlando Marquez Ayala", "Mathieu Reymond", "Massimo Caccia", "Alexandre Drouin", "Sarath Chandar", "Alexandre Lacoste"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0d4bd517ada9303e765827b3badfa2ee432bc5511ff4ee0cc42d937735e1c8de", "submitted_at": "2025-10-05T21:34:42Z", "updated_at": "2026-02-20T19:34:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.04373v2"}}, "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:2510.04373", "identifiers": {"arxiv": "2510.04373"}, "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/2510.04373"}], "name": "JEF-Hinter: Leveraging Offline Knowledge for Improving Web Agents Adaptation", "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": "2025-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces JEF-Hinter, which distills successful and failed trajectories into compact episodic hints and retrieves them for the current agent state. Web benchmark experiments examine guidance from decisive steps rather than full demonstration replay.", "topics": ["memory", "interaction-history", "web-navigation"], "updated_at": "2026-02-20", "url": "https://arxiv.org/abs/2510.04373"}
{"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": ["Tao Yu", "Zhengbo Zhang", "Zhiheng Lyu", "Junhao Gong", "Hongzhu Yi", "Xinming Wang", "Yuxuan Zhou", "Jiabing Yang", "Ping Nie", "Yan Huang", "Wenhu Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fe982af6e7285c21e13dcd4565255b501b401b188da64069d9cfe2fab7891f88", "submitted_at": "2025-10-12T15:43:37Z", "updated_at": "2025-10-14T08:54:57Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.10666v2"}}, "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:2510.10666", "identifiers": {"arxiv": "2510.10666"}, "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/2510.10666"}], "name": "BrowserAgent: Building Web Agents with Human-Inspired Web Browsing Actions", "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": "2025-10-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains BrowserAgent to use Playwright-based scrolling, clicking and typing directly on webpages for information-seeking tasks. Supervised and rejection fine-tuning are paired with explicit memory for conclusions carried across steps.", "topics": ["web-navigation", "imitation-learning", "memory"], "updated_at": "2025-10-14", "url": "https://arxiv.org/abs/2510.10666"}
{"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": ["Yuxuan Lu", "Ziyi Wang", "Jing Huang", "Hui Liu", "Jiri Gesi", "Yan Han", "Shihan Fu", "Tianqi Zheng", "Xianfeng Tang", "Chen Luo", "Yisi Sang", "Jin Lai", "Dakuo Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a9a1f8ebdf70ff2fa7b0c807160cd583c0f017a536457ff5915e589fca5fa751", "submitted_at": "2025-10-17T22:54:33Z", "updated_at": "2026-05-17T17:26:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.16252v2"}}, "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:2510.16252", "identifiers": {"arxiv": "2510.16252"}, "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/2510.16252"}], "name": "WEBSERV: A Full-Stack and RL-Ready Web Environment for Training Web Agents at Scale", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebServ combines copy-on-write container environments with a compact DOM-derived browser interface for training web agents. Network-aware waiting supports action execution in single-page applications, with evaluations and reinforcement-learning experiments on WebArena-Lite.", "topics": ["sandbox", "reinforcement-learning", "efficiency", "web-navigation"], "updated_at": "2026-05-17", "url": "https://arxiv.org/abs/2510.16252"}
{"aliases": [], "authors": ["Zhaoyang Wang", "Yiming Liang", "Xuchao Zhang", "Qianhui Wu", "Siwei Han", "Anson Bastos", "Rujia Wang", "Chetan Bansal", "Baolin Peng", "Jianfeng Gao", "Saravan Rajmohan", "Huaxiu Yao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7420d51203de06ed3afe6f2695b8612ec36dc072765a4f2394f366f9e6bf8838", "submitted_at": "2025-11-08T18:45:33Z", "updated_at": "2026-04-13T06:03:02Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2511.06101v3"}}, "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:2511.06101", "identifiers": {"arxiv": "2511.06101"}, "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/2511.06101"}], "name": "SynthAgent: Adapting Web Agents with Synthetic Supervision", "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": "2025-11-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SynthAgent, which generates tasks by exploring website elements and refines them when observations reveal conflicts. A later trajectory-refinement stage removes misaligned behavior before fine-tuning agents for the target websites.", "topics": ["data-generation", "imitation-learning", "web-navigation"], "updated_at": "2026-04-13", "url": "https://arxiv.org/abs/2511.06101"}
{"aliases": [], "authors": ["Genglin Liu", "Shijie Geng", "Sha Li", "Hejie Cui", "Sarah Zhang", "Xin Liu", "Tianyi Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8a105701d9265bacd3971f71f34a0e7ed315b6e54e9a7e675588c50713fcc4de", "submitted_at": "2025-11-17T05:38:50Z", "updated_at": "2026-07-22T21:19:00Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2511.12997v2"}}, "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:2511.12997", "identifiers": {"arxiv": "2511.12997"}, "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/2511.12997"}], "name": "WebCoach: Self-Evolving Web Agents with Cross-Session Memory Guidance", "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": "2025-11-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebCoach, which condenses navigation logs into persistent episodic memory and retrieves experiences by similarity and recency. Runtime hooks selectively supply advice to browser agents, evaluated on WebVoyager without retraining the underlying models.", "topics": ["memory", "interaction-history", "web-navigation"], "updated_at": "2026-07-22", "url": "https://arxiv.org/abs/2511.12997"}
{"aliases": [], "authors": ["Aram Vardanyan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "17debfa3b59e5a2fcae950c2659e5851ae101cc58181e0cc4e0320befe9b4cb1", "submitted_at": "2025-11-22T12:18:35Z", "updated_at": "2025-11-22T12:18:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.19477v1"}}, "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:2511.19477", "identifiers": {"arxiv": "2511.19477"}, "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/2511.19477"}], "name": "Building Browser Agents: Architecture, Security, and Practical Solutions", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. Broader claims about the safety of general-purpose browsing are the paper's position, not a catalog conclusion.", "published_at": "2025-11-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Reports architectural and security lessons from building a browser agent, including hybrid accessibility-tree and visual context, browser tooling and programmatic constraints. A WebGames evaluation accompanies the discussion of reliability and specialization.", "topics": ["web-navigation", "safety-security", "tool-integration"], "updated_at": "2025-11-22", "url": "https://arxiv.org/abs/2511.19477"}
{"aliases": [], "authors": ["Ahmed Awadallah", "Yash Lara", "Raghav Magazine", "Hussein Mozannar", "Akshay Nambi", "Yash Pandya", "Aravind Rajeswaran", "Corby Rosset", "Alexey Taymanov", "Vibhav Vineet", "Spencer Whitehead", "Andrew Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f0e23ba9b31a01956836e67329652882f84fe8ebef1075385a6af5e68cdae936", "submitted_at": "2025-11-24T19:56:28Z", "updated_at": "2025-11-24T19:56:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.19663v1"}}, "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", "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:2511.19663", "identifiers": {"arxiv": "2511.19663"}, "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/2511.19663"}], "name": "Fara-7B: An Efficient Agentic Model 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": "2025-11-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces FaraGen synthetic web trajectories, a compact screenshot-based action model, and WebTailBench for underrepresented browser tasks.", "topics": ["data-generation", "efficiency", "web-navigation"], "updated_at": "2025-11-24", "url": "https://arxiv.org/abs/2511.19663"}
{"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": ["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": ["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": ["Baixuan Li", "Jialong Wu", "Wenbiao Yin", "Kuan Li", "Zhongwang Zhang", "Huifeng Yin", "Zhengwei Tao", "Liwen Zhang", "Pengjun Xie", "Jingren Zhou", "Yong Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fde71dd935acb953bc5d01bcd48f0d6c3351db6a66126761baf134c114e7893c", "submitted_at": "2025-12-29T17:59:14Z", "updated_at": "2025-12-29T17:59:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.23647v1"}}, "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:2512.23647", "identifiers": {"arxiv": "2512.23647"}, "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/2512.23647"}], "name": "Nested Browser-Use Learning for Agentic Information Seeking", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2025-12-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "NestBrowse organizes browser actions in a nested framework that separates interaction control from page exploration. It evaluates this design for deep information seeking beyond snippet retrieval and direct URL fetching.", "topics": ["web-navigation", "planning", "efficiency", "tool-integration"], "updated_at": "2025-12-29", "url": "https://arxiv.org/abs/2512.23647"}
{"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": ["Jihong Wang", "Jiamu Zhou", "Weiming Zhang", "Teng Wang", "Weiwen Liu", "Zhuosheng Zhang", "Xingyu Lou", "Weinan Zhang", "Huarong Deng", "Jun Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a2afe8d0d8d83f91003b7885635e098e19f69024306c43c3edb23c110e553f63", "submitted_at": "2026-01-12T07:08:42Z", "updated_at": "2026-04-23T09:34:51Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.07262v3"}}, "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:2601.07262", "identifiers": {"arxiv": "2601.07262"}, "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/2601.07262"}], "name": "ColorBrowserAgent: Complex Long-Horizon Browser Agent with Adaptive 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": "2026-01-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "ColorBrowserAgent converts human feedback into reusable website knowledge and uses that knowledge to guide progressive context summaries. It evaluates the resulting memory and adaptation mechanisms on long browser workflows.", "topics": ["memory", "human-in-the-loop", "long-horizon", "web-navigation"], "updated_at": "2026-04-23", "url": "https://arxiv.org/abs/2601.07262"}
{"aliases": [], "authors": ["Xinyi Wu", "Geng Hong", "Yueyue Chen", "MingXuan Liu", "Feier Jin", "Xudong Pan", "Jiarun Dai", "Baojun Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4af00e0632ee21f70409998fc22f67e9a4ed1de415b488eda122c6ef94cab7aa", "submitted_at": "2026-01-12T07:10:08Z", "updated_at": "2026-01-12T07:10:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.07263v1"}}, "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:2601.07263", "identifiers": {"arxiv": "2601.07263"}, "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/2601.07263"}], "name": "When Bots Take the Bait: Exposing and Mitigating the Emerging Social Engineering Attack in Web Automation 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": "2026-01-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies deceptive social-engineering web patterns targeting browser automation agents and proposes defenses against bait-based manipulation.", "topics": ["safety-security", "web-navigation", "robustness"], "updated_at": "2026-01-12", "url": "https://arxiv.org/abs/2601.07263"}
{"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": ["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": ["Hang Ding", "Peidong Liu", "Junqiao Wang", "Ziwei Ji", "Meng Cao", "Rongzhao Zhang", "Lynn Ai", "Eric Yang", "Tianyu Shi", "Lei Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aedb5ed90c91d72f62f5ab60be5374b6f431f887425bc760883ce3fb85a46d43", "submitted_at": "2026-01-29T18:59:07Z", "updated_at": "2026-04-17T18:27:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.22149v2"}}, "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:2601.22149", "identifiers": {"arxiv": "2601.22149"}, "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/2601.22149"}], "name": "DynaWeb: Model-Based Reinforcement Learning of 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": "DynaWeb learns a world model that predicts web-page representations after actions and uses it to generate policy-training rollouts. It interleaves these rollouts with expert trajectories and evaluates the trained web agents on WebArena and WebVoyager.", "topics": ["world-models", "reinforcement-learning", "data-generation", "web-navigation"], "updated_at": "2026-04-17", "url": "https://arxiv.org/abs/2601.22149"}
{"aliases": [], "authors": ["Aiden Yiliu Li", "Xinyue Hao", "Shilong Liu", "Mengdi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a5b730fa81936d6a1b2b0a8b74a522a22de39b977929165c44a5050e490c301f", "submitted_at": "2026-02-02T18:50:07Z", "updated_at": "2026-02-02T18:50:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.02468v1"}}, "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:2602.02468", "identifiers": {"arxiv": "2602.02468"}, "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/2602.02468"}], "name": "Avenir-Web: Human-Experience-Imitating Multimodal Web Agents with Mixture of Grounding Experts", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Avenir-Web combines multiple grounding specialists with procedural experience, a task checklist and adaptive memory. It evaluates this planning and execution framework on live Online-Mind2Web tasks.", "topics": ["grounding", "planning", "memory", "web-navigation"], "updated_at": "2026-02-02", "url": "https://arxiv.org/abs/2602.02468"}
{"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": ["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": ["Yuyu Guo", "Wenjie Yang", "Siyuan Yang", "Ziyang Liu", "Cheng Chen", "Yuan Wei", "Yun Hu", "Yang Huang", "Guoliang Hao", "Dongsheng Yuan", "Jianming Wang", "Xin Chen", "Hang Yu", "Lei Lei", "Peng Di"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "04298ed8a4bc0f2811f52eeee6da0a13faa265bb3c5cbc5ea6be573a77c21f27", "submitted_at": "2026-02-14T02:33:55Z", "updated_at": "2026-04-30T02:05:10Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.13559v2"}}, "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", "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:2602.13559", "identifiers": {"arxiv": "2602.13559"}, "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/2602.13559"}], "name": "OpAgent: Operator Agent for Web Navigation", "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": "2026-02-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "OpAgent combines multi-task GUI fine-tuning with online web reinforcement learning using outcome judgments and rule-based progress rewards. A modular planner, grounder, reflector and summarizer support navigation and recovery, evaluated separately from the underlying trained policy.", "topics": ["web-navigation", "reinforcement-learning", "planning", "self-reflection"], "updated_at": "2026-04-30", "url": "https://arxiv.org/abs/2602.13559"}
{"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": ["Chenyang Ma", "Clyde Fare", "Matthew Wilson", "Dave Braines"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "353036db8a5609d64792ea49523ca424cf265d981ff34d1e7bc8fed35ffb21c0", "submitted_at": "2026-02-16T15:59:56Z", "updated_at": "2026-02-16T15:59:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.14865v1"}}, "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:2602.14865", "identifiers": {"arxiv": "2602.14865"}, "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/2602.14865"}], "name": "EmbeWebAgent: Embedding Web Agents into Any Customized UI", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-02-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "EmbeWebAgent embeds agent access into existing web applications through frontend observations and a page-specific function registry. Its workflow combines GUI primitives, composite actions and MCP tools, demonstrated in a live interface.", "topics": ["web-navigation", "tool-integration", "automation", "workflow"], "updated_at": "2026-02-16", "url": "https://arxiv.org/abs/2602.14865"}
{"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": ["Faria Huq", "Zora Zhiruo Wang", "Zhanqiu Guo", "Venu Arvind Arangarajan", "Tianyue Ou", "Frank Xu", "Shuyan Zhou", "Graham Neubig", "Jeffrey P. Bigham"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fe83d38ece7cd8f39d932611abaf7a170e7cb9b7c9744bd1176eb42e73be85b5", "submitted_at": "2026-02-19T18:11:28Z", "updated_at": "2026-07-07T23:37:46Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2602.17588v4"}}, "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:2602.17588", "identifiers": {"arxiv": "2602.17588"}, "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/2602.17588"}], "name": "Modeling Distinct Human Interaction 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": "2026-02-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "CowCorpus records interleaved human and agent actions during web navigation and identifies distinct patterns of user intervention. Models trained to anticipate these interventions are integrated into live web agents and assessed in a user study.", "topics": ["human-in-the-loop", "web-navigation", "data-generation", "action-prediction"], "updated_at": "2026-07-07", "url": "https://arxiv.org/abs/2602.17588"}
{"aliases": [], "authors": ["Hongbin Zhong", "Fazle Faisal", "Luis França", "Tanakorn Leesatapornwongsa", "Adriana Szekeres", "Kexin Rong", "Suman Nath"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e179c833d43a88098d576e31e726eb3e4289f728659697364d0b448282df3fc1", "submitted_at": "2026-02-24T03:03:18Z", "updated_at": "2026-09-28T17:40:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.20502v2"}}, "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:2602.20502", "identifiers": {"arxiv": "2602.20502"}, "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/2602.20502"}], "name": "ActionEngine: From Reactive to Programmatic Web Agents via State Machine Memory", "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": "2026-02-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "ActionEngine explores web applications to build an updateable state-machine memory of GUI states, operations and transitions. It generates complete task programs from that memory, using reactive repair to handle changed interfaces or missing knowledge.", "topics": ["memory", "planning", "efficiency", "web-navigation"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2602.20502"}
{"aliases": [], "authors": ["Linxi Jiang", "Zhijie Liu", "Haotian Luo", "Zhiqiang Lin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d3b24ac636d86124a70b44e84657b963db4c5af3367ae70e5c35e0bebb3f600f", "submitted_at": "2026-02-28T05:25:03Z", "updated_at": "2026-02-28T05:25:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.00476v1"}}, "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:2603.00476", "identifiers": {"arxiv": "2603.00476"}, "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/2603.00476"}], "name": "Atomicity for Agents: Exposing, Exploiting, and Mitigating TOCTOU Vulnerabilities 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": "2026-02-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Identifies time-of-check to time-of-use (TOCTOU) race vulnerabilities when web pages mutate between agent planning and action execution.", "topics": ["safety-security", "web-navigation", "state-tracking"], "updated_at": "2026-02-28", "url": "https://arxiv.org/abs/2603.00476"}
{"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", "Qingyun Shi", "Shengze Xu", "Shengbo Cai", "Tieyong Zeng", "Li Ling", "Yanyi Shang", "Dehan Kong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fcca36d9a28ace835ede3fa5a93cb9c54608f164240c18bfe007bc945273b460", "submitted_at": "2026-03-05T10:51:34Z", "updated_at": "2026-04-14T08:14:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.05044v2"}}, "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:2603.05044", "identifiers": {"arxiv": "2603.05044"}, "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/2603.05044"}], "name": "WebFactory: Automated Compression of Foundational Language Intelligence into Grounded Web 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": "2026-03-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebFactory, a GUI-agent training pipeline that synthesizes websites, generates tasks and trajectories, and trains with decomposed rewards. Offline and online transfer evaluations examine agents trained on the generated environments.", "topics": ["data-generation", "reinforcement-learning", "web-navigation"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2603.05044"}
{"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": ["Zijian Lu", "Yiping Zuo", "Yupeng Nie", "Xin He", "Weibei Fan", "Lianyong Qi", "Shi Jin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8e8182e64c7a57a2dd6eb381762bc9a60b08e13ba5a337d0e9125ab6a5518b6c", "submitted_at": "2026-03-20T09:25:50Z", "updated_at": "2026-04-15T07:34:34Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2603.20340v3"}}, "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:2603.20340", "identifiers": {"arxiv": "2603.20340"}, "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/2603.20340"}], "name": "ContractSkill: Repairable Contract-Based Skills for Multimodal Web 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. The abstract limits its portability evidence to transfer within a benchmark family, not generalization across arbitrary benchmarks.", "published_at": "2026-03-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Turns draft web-agent skills into structured executable artifacts that support deterministic checks, fault localization and local repair. VisualWebArena and MiniWoB experiments examine execution and reuse within the same benchmark family.", "topics": ["web-navigation", "planning", "robustness"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2603.20340"}
{"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": ["Qianlong Lan", "Anuj Kaul"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bb8164fa076c55cddf73508e6b68ce24a6594917e65862504ba9f08b4e97ee48", "submitted_at": "2026-03-24T23:49:15Z", "updated_at": "2026-03-24T23:49:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.23791v1"}}, "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:2603.23791", "identifiers": {"arxiv": "2603.23791"}, "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/2603.23791"}], "name": "The Cognitive Firewall:Securing Browser Based AI Agents Against Indirect Prompt Injection Via Hybrid Edge Cloud 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": "2026-03-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Proposes a split-compute browser defense pairing a local visual sentinel, a cloud planner, and deterministic policy enforcement against indirect prompt injection.", "topics": ["safety-security", "web-navigation"], "updated_at": "2026-03-24", "url": "https://arxiv.org/abs/2603.23791"}
{"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": ["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": ["Wei Zou", "Mingwen Dong", "Miguel Romero Calvo", "Shuaichen Chang", "Jiang Guo", "Dongkyu Lee", "Xing Niu", "Xiaofei Ma", "Yanjun Qi", "Jiarong Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a37f4bdc495e10e4ae301ebff6faa28cdfa6de152d4619c051083b5dce41e6ba", "submitted_at": "2026-04-03T01:25:12Z", "updated_at": "2026-04-07T14:45:15Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.02623v2"}}, "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:2604.02623", "identifiers": {"arxiv": "2604.02623"}, "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/2604.02623"}], "name": "Poison Once, Exploit Forever: Environment-Injected Memory Poisoning Attacks on Web 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": "2026-04-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies persistent web-agent memory contamination introduced through environmental observations instead of direct memory access. Experiments on WebArena and VisualWebArena examine cross-session effects and how interaction failures alter susceptibility.", "topics": ["safety-security", "memory", "web-navigation"], "updated_at": "2026-04-07", "url": "https://arxiv.org/abs/2604.02623"}
{"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": ["Tanmay Gupta", "Piper Wolters", "Zixian Ma", "Peter Sushko", "Rock Yuren Pang", "Diego Llanes", "Yue Yang", "Taira Anderson", "Boyuan Zheng", "Zhongzheng Ren", "Harsh Trivedi", "Taylor Blanton", "Caleb Ouellette", "Winson Han", "Ali Farhadi", "Ranjay Krishna"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dbc0d512a11b88a9036d06d0fcc26a8ab0d0bc54fad986d4b6225c8073897bc2", "submitted_at": "2026-04-09T17:54:02Z", "updated_at": "2026-04-09T17:54:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.08516v1"}}, "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:2604.08516", "identifiers": {"arxiv": "2604.08516"}, "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/2604.08516"}], "name": "MolmoWeb: Open Visual Web Agent and Open Data for the Open Web", "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 states that artifacts will be released; this annotation does not independently verify release status or licenses.", "published_at": "2026-04-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents MolmoWebMix, combining synthetic browser trajectories, human demonstrations and GUI perception data, and MolmoWeb screenshot-to-action policies trained from it. Browser benchmarks also test parallel rollouts with best-of-N selection.", "topics": ["web-navigation", "data-generation", "imitation-learning"], "updated_at": "2026-04-09", "url": "https://arxiv.org/abs/2604.08516"}
{"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": ["Zhaoyang Wang", "Qianhui Wu", "Xuchao Zhang", "Chaoyun Zhang", "Wenlin Yao", "Fazle Elahi Faisal", "Baolin Peng", "Si Qin", "Suman Nath", "Qingwei Lin", "Chetan Bansal", "Dongmei Zhang", "Saravan Rajmohan", "Jianfeng Gao", "Huaxiu Yao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf709a3d7096bc07218206edb1accfdda64036a39ea770e637b3e073dd9a76c8", "submitted_at": "2026-04-14T21:48:15Z", "updated_at": "2026-08-31T07:02:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.13318v2"}}, "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:2604.13318", "identifiers": {"arxiv": "2604.13318"}, "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/2604.13318"}], "name": "WebXSkill: Skill Learning for 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": "2026-04-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebXSkill pairs parameterized action programs with step-level language guidance, mining reusable skills from synthetic browser-agent traces. A URL-based graph supports retrieval, and deployment alternates between direct program execution and guided native planning.", "topics": ["memory", "planning", "web-navigation", "automation"], "updated_at": "2026-08-31", "url": "https://arxiv.org/abs/2604.13318"}
{"aliases": [], "authors": ["Lingfeng Zhang", "Yongan Sun", "Jinpeng Hu", "Hui Ma", "Yang Ying", "Kuien Liu", "Zenglin Shi", "Meng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "123577fbb4431f22bc18d2d32187744268f9d63b3149e9e1775fb44673771feb", "submitted_at": "2026-04-20T05:19:49Z", "updated_at": "2026-04-21T03:15:01Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.17821v2"}}, "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:2604.17821", "identifiers": {"arxiv": "2604.17821"}, "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/2604.17821"}], "name": "WebUncertainty: Dual-Level Uncertainty Driven Planning and Reasoning For Autonomous Web Agent", "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": "2026-04-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebUncertainty adapts planning mode to task uncertainty and guides action-level tree search with confidence-derived uncertainty estimates. Its evaluations examine autonomous execution of dynamic, long-horizon web tasks on WebArena and WebVoyager.", "topics": ["planning", "web-navigation", "long-horizon", "robustness"], "updated_at": "2026-04-21", "url": "https://arxiv.org/abs/2604.17821"}
{"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": ["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": ["Donguk Kwon", "Dongha Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6bf74d53b22530951788bd329167f5ba9474827c712590ae9c6d316fac4d62bc", "submitted_at": "2026-05-08T02:11:23Z", "updated_at": "2026-05-08T02:11:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.07134v1"}}, "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:2605.07134", "identifiers": {"arxiv": "2605.07134"}, "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/2605.07134"}], "name": "Region4Web: Rethinking Observation Space Granularity for Web 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": "2026-05-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Reorganizes webpage accessibility trees into functional regions through hierarchical decomposition and semantic abstraction. PageDigest supplies a persistent region-level observation to web agents, evaluated on WebArena.", "topics": ["screen-understanding", "web-navigation", "efficiency"], "updated_at": "2026-05-08", "url": "https://arxiv.org/abs/2605.07134"}
{"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": ["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": ["Fatemeh Pesaran Zadeh", "Seyeon Choi", "Xing Han Lù", "Siva Reddy", "Gunhee Kim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aa6e11bae4d952d79276543ced6f81193b36868a6222fe51ee7b0816aa72fbab", "submitted_at": "2026-05-19T09:19:01Z", "updated_at": "2026-05-26T02:53:26Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.20291v2"}}, "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:2605.20291", "identifiers": {"arxiv": "2605.20291"}, "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/2605.20291"}], "name": "Weasel: Out-of-Domain Generalization for Web Agents via Importance-Diversity Data Selection", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-05-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Weasel selects web-agent training steps by balancing importance and diversity across states, sites and interaction patterns. It also prunes training accessibility trees around known action targets and regenerates rationales to match the model’s reasoning style.", "topics": ["web-navigation", "imitation-learning", "efficiency", "robustness", "data-generation"], "updated_at": "2026-05-26", "url": "https://arxiv.org/abs/2605.20291"}
{"aliases": [], "authors": ["Caleb Winston", "Ron Yifeng Wang", "Azalia Mirhoseini", "Christos Kozyrakis"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b73bf7776e280a9040b1d7d1e29957913723253014c1dfe84f64916b2a08e5e", "submitted_at": "2026-05-20T17:54:27Z", "updated_at": "2026-05-29T17:54:05Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.21470v2"}}, "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:2605.21470", "identifiers": {"arxiv": "2605.21470"}, "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/2605.21470"}], "name": "Agent JIT Compilation for Latency-Optimizing Web Agent Planning and Scheduling", "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": "2026-05-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Agent JIT compilation turns browser task descriptions into validated executable plans containing model calls, tools and parallel work. A cost-based scheduler and pre/postcondition protocol target latency and incorrect tool use in web-agent execution.", "topics": ["planning", "efficiency", "tool-integration", "web-navigation"], "updated_at": "2026-05-29", "url": "https://arxiv.org/abs/2605.21470"}
{"aliases": [], "authors": ["Yubo Li", "Yidi Miao", "Yuntian Shen", "Yuxin Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e64d36e920a33851aacf93138a2081d0470424dfec6a8fd11915928660824851", "submitted_at": "2026-05-24T00:07:25Z", "updated_at": "2026-05-26T04:25:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.24785v2"}}, "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:2605.24785", "identifiers": {"arxiv": "2605.24785"}, "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/2605.24785"}], "name": "PANDO: Efficient Multimodal AI Agents via Online Skill Distillation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-05-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Analyzes inefficiency in VisualWebArena trajectories and distills reusable online skills to reduce repeated actions and token overhead in multimodal web agents.", "topics": ["efficiency", "memory", "web-navigation"], "updated_at": "2026-05-26", "url": "https://arxiv.org/abs/2605.24785"}
{"aliases": [], "authors": ["Jiaxi Li", "Ke Deng", "Yun Wang", "Jingyuan Huang", "Yucheng Shi", "Qiaoyu Tan", "Jin Lu", "Ninghao Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6d891956bde9b8672859fcfcd2d54dfd3e4720964c25d77cc28bc1200ef3a56c", "submitted_at": "2026-06-03T03:11:50Z", "updated_at": "2026-06-03T03:11:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.04391v1"}}, "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:2606.04391", "identifiers": {"arxiv": "2606.04391"}, "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/2606.04391"}], "name": "Online Skill Learning for Web Agents via State-Grounded Dynamic Retrieval", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "State-Grounded Dynamic Retrieval extracts reusable sub-procedures from completed web trajectories and represents them as paired text and code. Skills are retrieved against both the current webpage state and the task goal at intermediate execution steps.", "topics": ["web-navigation", "memory", "interaction-history", "planning", "workflow"], "updated_at": "2026-06-03", "url": "https://arxiv.org/abs/2606.04391"}
{"aliases": [], "authors": ["Hao Bai", "Rui Yang", "Chenlu Ye", "Spencer Whitehead", "Aviral Kumar", "Tong Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8a8afe44ffa73739505070c5f8004d8b4bbe98455fc0f5f3cef4b450bf441c74", "submitted_at": "2026-06-04T02:18:44Z", "updated_at": "2026-08-07T17:58:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2606.05597v3"}}, "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:2606.05597", "identifiers": {"arxiv": "2606.05597"}, "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/2606.05597"}], "name": "AsyncWebRL: Efficient Asynchronous Reinforcement Learning for Multi-Step Visual 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": "2026-06-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "AsyncWebRL overlaps web-agent rollouts, updates and policy refresh while managing screenshot handling and a persistent rollout pool. It also replaces trajectory-length normalization in multi-step GRPO to reduce verbose failed trajectories, with throughput and out-of-distribution evaluation.", "topics": ["reinforcement-learning", "efficiency", "web-navigation"], "updated_at": "2026-08-07", "url": "https://arxiv.org/abs/2606.05597"}
{"aliases": [], "authors": ["Radeen Mostafa", "Sawradip Saha"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d29dec745923e652654521f65a32f919a99d4577992006efc1a5e469bf775d81", "submitted_at": "2026-06-08T12:18:22Z", "updated_at": "2026-06-08T12:18:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.09399v1"}}, "environments": ["web"], "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:2606.09399", "identifiers": {"arxiv": "2606.09399"}, "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/2606.09399"}], "name": "RunAgent SuperBrowser: A Theory of Autonomous Web Navigation Grounded in Human Browsing Behaviour", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "SUPERBROWSER combines screenshot-based target proposals with separate orchestration, planning and action roles. A bounded ledger and context-eviction process retain recent actions, task facts and checkpoints during browser execution.", "topics": ["web-navigation", "grounding", "planning", "memory", "efficiency"], "updated_at": "2026-06-08", "url": "https://arxiv.org/abs/2606.09399"}
{"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": ["Jayoo Hwang", "Xiaowen Zhang", "Vedant Padwal"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "386bd82be97d5cbe64fdb3478e427362e710d78d423c9e52661a8a96ea2fdecf", "submitted_at": "2026-06-09T04:53:19Z", "updated_at": "2026-06-09T04:53:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.10423v1"}}, "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:2606.10423", "identifiers": {"arxiv": "2606.10423"}, "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/2606.10423"}], "name": "WebChallenger: A Reliable and Efficient Generalist Web Agent", "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": "2026-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebChallenger builds a hierarchical DOM-derived page representation for selective observation, reusable site maps and compound actions. The framework combines these mechanisms without model fine-tuning and evaluates generalization across multiple web-agent benchmarks.", "topics": ["web-navigation", "memory", "efficiency", "training-free", "workflow"], "updated_at": "2026-06-09", "url": "https://arxiv.org/abs/2606.10423"}
{"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": ["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": ["Tianyi Men", "Zhuoran Jin", "Pengfei Cao", "Yubo Chen", "Kang Liu", "Jun Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "42166ac3980296a53b56bc4a076634beec36ba885b3260b5d36115b5a5261cd0", "submitted_at": "2026-06-25T17:44:48Z", "updated_at": "2026-06-25T17:44:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.27330v1"}}, "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:2606.27330", "identifiers": {"arxiv": "2606.27330"}, "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/2606.27330"}], "name": "Empowering GUI Agents via Autonomous Experience Exploration and Hindsight Experience Utilization for Task Planning", "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": "2026-06-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "PEEU explores web environments and turns hindsight experience into aligned high-level planning supervision. A hierarchical task-decomposition analysis examines why atomic-skill competence alone does not ensure planning generalization across websites.", "topics": ["planning", "data-generation", "imitation-learning", "web-navigation"], "updated_at": "2026-06-25", "url": "https://arxiv.org/abs/2606.27330"}
{"aliases": [], "authors": ["Kristina Nikolić", "Egor Zverev", "Javier Rando", "Matthew Jagielski", "Edoardo Debenedetti", "Florian Tramèr"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "97d6c625b6ebc08cec7471df543812a91c4252f50d8d35f8d51fbef802dcb5a2", "submitted_at": "2026-07-06T16:22:37Z", "updated_at": "2026-07-06T16:22:37Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.05277v1"}}, "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:2607.05277", "identifiers": {"arxiv": "2607.05277"}, "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/2607.05277"}], "name": "Untrusted Content Masking for Web Agents with Security Guarantees", "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": "2026-07-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Untrusted Content Masking uses DOM structure to redact untrusted webpage regions before agent observation. A sandboxed interaction interface separates privileges so browser actions can proceed under the proposed trust-boundary model without exposing protected content to the planner.", "topics": ["safety-security", "sandbox", "web-navigation"], "updated_at": "2026-07-06", "url": "https://arxiv.org/abs/2607.05277"}
{"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": ["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": ["Yashaswi Malla", "Sandra Siby"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "60c93c90e6d4026452810a866a0c49d57afdec15ad5ced30606c925a8e83c8f7", "submitted_at": "2026-07-31T18:27:12Z", "updated_at": "2026-07-31T18:27:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.00202v1"}}, "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:2608.00202", "identifiers": {"arxiv": "2608.00202"}, "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/2608.00202"}], "name": "From Monoliths to Swarms: A Study of Attack Surface Evolution in the Transition to Multi-Agent Web Systems", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Categorizes structural attack surfaces that emerge when single-agent browser systems are decomposed into multi-agent web workflows.", "topics": ["safety-security", "web-navigation", "planning"], "updated_at": "2026-07-31", "url": "https://arxiv.org/abs/2608.00202"}
{"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": ["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": ["Ruitao Wang", "Yuwen Hao", "Menglin Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf543ef0b77eb6fe430c8cef5382932f13e4e4136aaba9c14c44b5a591d6612a", "submitted_at": "2026-08-12T12:03:20Z", "updated_at": "2026-08-12T12:03:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.12429v1"}}, "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:2608.12429", "identifiers": {"arxiv": "2608.12429"}, "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/2608.12429"}], "name": "SynWeaver: Website-Prior Task and Trajectory Co-Synthesis for 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": "2026-08-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "SynWeaver explores a website to build a map of page states and executable interactions, then learns website priors for task proposals. Joint task/trajectory synthesis revises inconsistent pairs and verifies or repairs the resulting supervision.", "topics": ["data-generation", "web-navigation", "state-tracking", "robustness"], "updated_at": "2026-08-12", "url": "https://arxiv.org/abs/2608.12429"}
{"aliases": [], "authors": ["AIMAE Team", "Tianxiang Chen", "Yan Cheng", "Zhangye Han", "Xiaowei Li", "Chang Liu", "Cheng Liu", "Zhongqiang Ma", "Long Peng", "Xiaobing Tu", "Yinggui Wang", "Hongliang Wei", "Chen Wu", "Daiping Xin", "Kunyu Zhou", "Pengyang Zhou", "Peiyuan Chen", "Ziyuan Chen", "Yutao Deng", "Chunyu Dong", "Xiangyu Fu", "Yicheng Feng", "Ruian He", "Haochen Li", "Miancan Liu", "Zhengqin Liu", "Wei Peng", "Jinkui Ren", "Haoyu Tan", "Dong Xiao", "Rongkun Xue", "Shujian Yang", "Xianhang Ye", "Ziqi Yuan", "Ziyang Yu", "Linghan Zhang", "Xiantao Zhang", "Xuanpu Zhao", "Yinan Zhao", "Zhenghui Zhao", "Bin Zhu", "Likai Zou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "474187aaa79165dc8652eec1e2af6b1733b2ee37b3cd125c89ff0defcaf2e193", "submitted_at": "2026-08-18T03:23:31Z", "updated_at": "2026-08-18T03:23:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.17319v1"}}, "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", "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:2608.17319", "identifiers": {"arxiv": "2608.17319"}, "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/2608.17319"}], "name": "Wuying-Browser-Agent: Real-World Centric Fundamental Long-Horizon 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": "2026-08-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines a browser execution framework, recovery-oriented supervision, and online reinforcement learning, with a bilingual live-web task benchmark.", "topics": ["web-navigation", "long-horizon", "reinforcement-learning"], "updated_at": "2026-08-18", "url": "https://arxiv.org/abs/2608.17319"}
{"aliases": [], "authors": ["Fei Tang", "Huawen Shen", "Zhiqiong Lu", "Zhengxi Lu", "Pengyuan Lyu", "Chengquan Zhang", "Weiming Lu", "Jun Xiao", "Yueting Zhuang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0d2180e9b1b0c3ae20e32b0b597dc209e363551aaa5031c98504ef08314979a6", "submitted_at": "2026-08-25T17:35:42Z", "updated_at": "2026-08-25T17:35:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.24848v1"}}, "environments": ["web"], "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:2608.24848", "identifiers": {"arxiv": "2608.24848"}, "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/2608.24848"}], "name": "BrowserForge: Scaling Web Episode via Parallel Browser Sandboxes", "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 BrowserForge, which combines broad website sourcing, parallel browser sandboxes, and task-proposer/solver interaction to collect verified web trajectories. The pipeline uses page structure during synthesis but trains a policy that acts from screenshots.", "topics": ["data-generation", "sandbox", "web-navigation"], "updated_at": "2026-08-25", "url": "https://arxiv.org/abs/2608.24848"}
{"aliases": [], "authors": ["Lily Bundgaard Wanscher", "Markus Heidemann Lorensen", "Mohammed Ammad Shafiq", "Mahyar Tourchi Moghaddam", "Mina Alipour"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "046d93a43dd5a011809a780c7d998f98ee0abb3e931d733b8ad7807b86050ec7", "submitted_at": "2026-07-26T17:21:24Z", "updated_at": "2026-07-26T17:21:24Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.24913v1"}}, "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:2608.24913", "identifiers": {"arxiv": "2608.24913"}, "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/2608.24913"}], "name": "From Blind Edits to Verified Repair: Building Trustworthy User-Side LLM Agents for Web Accessibility", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds a local browser-extension agent that generates and verifies additive CSS repairs for WCAG web accessibility violations.", "topics": ["ui-design", "web-navigation", "safety-security"], "updated_at": "2026-07-26", "url": "https://arxiv.org/abs/2608.24913"}
{"aliases": [], "authors": ["Junxuan Li", "Zijun Liu", "Ziyi Huang", "Peng Li", "Yuzhou Liu", "Ming Yan", "Yang Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bbf657b2228bad25e668e3c9d7ec20c0ecab403d34b7e0b5608d350c46b8e50f", "submitted_at": "2026-08-29T14:45:09Z", "updated_at": "2026-08-29T14:45:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.29305v1"}}, "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", "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:2608.29305", "identifiers": {"arxiv": "2608.29305"}, "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/2608.29305"}], "name": "Learning Simple Test-Time Environments for 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": "2026-08-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Test-Time Environment Decomposition, which uses trial interaction to break complex web observations into simpler modules. A label-free adaptation procedure learns from these sub-environments and studies whether their gains compose in full web tasks.", "topics": ["planning", "robustness", "web-navigation"], "updated_at": "2026-08-29", "url": "https://arxiv.org/abs/2608.29305"}
{"aliases": [], "authors": ["Bowei He", "Xiaokun Zhang", "Meng Ding", "Xue Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "edbe52f304634dfa06e0f232090a4fe998c7ea111418f57942976d348c4e0997", "submitted_at": "2026-08-31T08:16:56Z", "updated_at": "2026-08-31T08:16:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.05511v1"}}, "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", "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:2609.05511", "identifiers": {"arxiv": "2609.05511"}, "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/2609.05511"}], "name": "SCAFFOLD: Self-Improving Web Agents via Recursive Parametric Skill Abstraction", "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 SCAFFOLD, which extracts executable skills from web trajectories, composes them into a hierarchy, and compresses redundant skills through behavioral checks. It periodically distills skill-guided trajectories into the policy and studies repeated improvement across web benchmarks.", "topics": ["memory", "automation", "web-navigation"], "updated_at": "2026-08-31", "url": "https://arxiv.org/abs/2609.05511"}
{"aliases": [], "authors": ["Jianwei Zhang", "Sihan Cao", "Pengcheng Zheng", "Ya Wen", "Pei Ke", "Kuien Liu", "Shen Gao", "Wei Dong", "Yang Yang", "Chaoning Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aaeb2a57b2bb92ed926c9b65fb40b6bbf272e0ad9bb7415716d24eab0aeb4402", "submitted_at": "2026-08-31T08:40:34Z", "updated_at": "2026-08-31T08:40:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.05513v1"}}, "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", "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:2609.05513", "identifiers": {"arxiv": "2609.05513"}, "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/2609.05513"}], "name": "When and What to Teach: Budget-Aware Online Adaptation 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": "2026-08-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Proposes an online teaching framework that decides which web-agent episodes warrant a stronger teacher and which trajectory turns merit training. A solvability gate and score-guided trimming aim to preserve first-attempt performance while reducing teaching and training costs.", "topics": ["efficiency", "data-generation", "web-navigation"], "updated_at": "2026-08-31", "url": "https://arxiv.org/abs/2609.05513"}
{"aliases": [], "authors": ["Alexandru Ianta", "Eleni Stroulia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "29b615736812b466b0df7133b7070184a1b7c8dacbe4c73fe32b878b654bb78d", "submitted_at": "2026-09-11T19:57:22Z", "updated_at": "2026-09-11T19:57:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.13491v1"}}, "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:2609.13491", "identifiers": {"arxiv": "2609.13491"}, "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/2609.13491"}], "name": "Token Efficient Task Execution via Application Behavior Modeling 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": "2026-09-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "OdoBot builds a model of web-application behavior from successful task demonstrations and uses it to guide task execution. Its Canvas LMS evaluation compares token use and completion with existing web agents.", "topics": ["memory", "efficiency", "web-navigation"], "updated_at": "2026-09-11", "url": "https://arxiv.org/abs/2609.13491"}
{"aliases": [], "authors": ["Xinyun Cao", "Adriana Szekeres", "Fazle Elahi Faisal"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0ce929c347ed5160b1f8455cc9d7e73ac2cc1b678e9d386d00dd4b2ef146b56d", "submitted_at": "2026-09-11T21:24:40Z", "updated_at": "2026-09-11T21:24:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.13548v1"}}, "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:2609.13548", "identifiers": {"arxiv": "2609.13548"}, "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/2609.13548"}], "name": "AutoTailor: Automatic, User-Aligned Capability Selection and Adaptation for 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": "2026-09-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "AutoTailor compiles web trajectories into reusable browser-automation APIs, then filters and dynamically revises the capability inventory. WebArena experiments distinguish API-only operation from an agent that can fall back to ordinary reasoning and browser actions.", "topics": ["tool-integration", "web-navigation", "workflow", "efficiency"], "updated_at": "2026-09-11", "url": "https://arxiv.org/abs/2609.13548"}
{"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": ["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": ["Yangzhe Peng", "Xiaoyang Wang", "Yiyang Zhao", "Lijun Wu", "Kun He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7cc6fd87d2299e2881eab9a2db96e4022663e2d94da5549dcde166c0c8b8be11", "submitted_at": "2026-09-26T08:14:50Z", "updated_at": "2026-09-26T08:14:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32351v1"}}, "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:2609.32351", "identifiers": {"arxiv": "2609.32351"}, "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/2609.32351"}], "name": "From Trajectories to Grounded Preferences: Process Preference Synthesis via Interaction Element Graphs for Web PRMs", "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 SURFPRM, which turns web demonstrations into an interaction-element graph for synthesizing grounded negative actions. Its contrastive preference data trains process reward models to distinguish plausible on-page alternatives and guide downstream trajectory search.", "topics": ["data-generation", "reward-models", "web-navigation"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32351"}
{"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": ["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": ["Keliang Li", "Heng Wang", "Chen Hu", "Daxin Jiang", "Hong Chang", "Shiguang Shan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "88daacf5fda62f59a08b504a633ad984ee1d03791bba24e89c223c0686adf34e", "submitted_at": "2026-09-27T15:11:30Z", "updated_at": "2026-09-27T15:11:30Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.33646v1"}}, "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:2609.33646", "identifiers": {"arxiv": "2609.33646"}, "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/2609.33646"}], "name": "Probe to Act: Elevating Browser-Use Agent via Active Visual Probing", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Probe to Act, which lets browser agents request visual evidence linking DOM handles and screen regions before committing an operation. It retains selected interaction evidence as memory and studies both prompted use and distillation into open models.", "topics": ["grounding", "memory", "web-navigation"], "updated_at": "2026-09-27", "url": "https://arxiv.org/abs/2609.33646"}
{"aliases": [], "authors": ["Ziqi Zhang", "Shaohui Li", "Bing Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7efd2eafa1984d7c8f4bf6ce79a94cd2ca4189da330bdd056b4be9c07f7f9dc1", "submitted_at": "2026-09-28T04:02:07Z", "updated_at": "2026-09-28T04:02:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.35904v1"}}, "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", "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:2609.35904", "identifiers": {"arxiv": "2609.35904"}, "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/2609.35904"}], "name": "Structured Interaction, Visual Localization, and Robust Execution for Complex Web Tasks: A Technical Report on the WebRetriever Challenge", "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": "Describes a WebRetriever agent that combines semantic webpage operations with visual localization when structured representations are insufficient. Hierarchical context management and fault-aware execution support processing tasks across concurrent browsers.", "topics": ["web-navigation", "grounding", "robustness", "memory"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.35904"}
{"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": ["Hongbo Zhang", "Liuyang Song", "Quanquan Li", "Daqian Yang", "Yan Wen", "Zhengtao Yao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bb910e64ebb65bd49e8729e4373c1c81ce4b2d0a32647a5abaf15f52737e578d", "submitted_at": "2026-09-30T02:11:02Z", "updated_at": "2026-09-30T02:11:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.38778v1"}}, "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:2609.38778", "identifiers": {"arxiv": "2609.38778"}, "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/2609.38778"}], "name": "Action Conditioned Bisimulation For GUI Agent Memory", "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": "Defines a GUI-memory merge rule that treats two states as equivalent only when shared actions lead to matching outcomes and successor states. The training-free method replaces observation-similarity merging and is tested in closed-loop MiniWoB++ interaction.", "topics": ["memory", "web-navigation"], "updated_at": "2026-09-30", "url": "https://arxiv.org/abs/2609.38778"}
{"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": ["Jiangang Han"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c25f023b6ff66baa38661362d5730fc338e781db5bd23dd91216d3937ae730b0", "submitted_at": "2026-10-02T09:18:28Z", "updated_at": "2026-10-07T13:46:51Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2610.03036v2"}}, "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:2610.03036", "identifiers": {"arxiv": "2610.03036"}, "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/2610.03036"}], "name": "WebFovea: When the Model Is Right but the Click Is Wrong -- Reliable Round Trips for Vision-Based Web Agents on Live 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": "2026-10-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes WebFovea, a vision-based live-web agent whose execution loop checks action parsing, browser effects, feedback accuracy, and observation delivery. The report analyzes failures such as coordinate mismatches and silent control interactions and documents corresponding harness changes.", "topics": ["web-navigation", "robustness", "observability"], "updated_at": "2026-10-07", "url": "https://arxiv.org/abs/2610.03036"}
{"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": ["Sarthak Choudhary", "Mihai Christodorescu", "Ashish Hooda", "Somesh Jha", "Tongxin Li", "Damien Octeau"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f38c9e37ab7fc8356a0c8c72370de673d1c2b6294e15dfdc72c280269f43439c", "submitted_at": "2026-10-07T05:24:40Z", "updated_at": "2026-10-07T05:24:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.09469v1"}}, "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:2610.09469", "identifiers": {"arxiv": "2610.09469"}, "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/2610.09469"}], "name": "Secure-CUA: Controlling Untrusted Influence 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": "2026-10-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Proposes Secure-CUA, which commits to per-action transactions before reading untrusted page content and separates content queries from GUI target selection. The paper formalizes its protection assumptions and evaluates benign task utility on WebArena.", "topics": ["safety-security", "web-navigation"], "updated_at": "2026-10-07", "url": "https://arxiv.org/abs/2610.09469"}
{"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": ["Sunghwan Kim", "Deogyong Kim", "Sangam Lee", "Donguk Kwon", "Dongha Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cadcf2ab7be51b17d06975eba0e56573c333dca9ec034ade67b84e62d0096903", "submitted_at": "2026-10-09T11:23:40Z", "updated_at": "2026-10-09T11:23:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.13416v1"}}, "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:2610.13416", "identifiers": {"arxiv": "2610.13416"}, "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/2610.13416"}], "name": "Web$hell: A Unified Interface for Agentic Web Access", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Web$hell, an interface that compiles live pages into persistent information and action objects so agents can gather evidence, navigate websites, and reuse acquired content through shared operations.", "topics": ["web-navigation", "tool-integration", "memory", "efficiency"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2610.13416"}
{"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": ["Zijian Cai", "Haowei Liu", "Kailai Chen", "Lin Peng", "Cong Wan", "Ziqi Zhang", "Chunfeng Yuan", "Bing Li", "Weiming Hu", "Haiyang Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4596cc60436c4fa6c2aa0951b4699b804edb73aa1b1a46900b1d14f901d20f3b", "submitted_at": "2026-10-12T07:42:11Z", "updated_at": "2026-10-12T07:42:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.16319v1"}}, "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:2610.16319", "identifiers": {"arxiv": "2610.16319"}, "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/2610.16319"}], "name": "HyWe: Unifying Search and Operation for Real-World Hybrid 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": "2026-10-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces HyWe, which switches between information search and browser operation within one trajectory, alongside a demonstration pipeline and a benchmark of hybrid web tasks.", "topics": ["web-navigation", "planning", "data-generation"], "updated_at": "2026-10-12", "url": "https://arxiv.org/abs/2610.16319"}
{"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": ["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": ["Md Farhan Ishmam", "Kenneth Marino"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6d154a14c5ac85fb352721c203fe7bd709eaba4287ef7376e6a6f5fc395708f1", "submitted_at": "2026-10-14T09:06:49Z", "updated_at": "2026-10-14T09:06:49Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.19834v1"}}, "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:2610.19834", "identifiers": {"arxiv": "2610.19834"}, "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/2610.19834"}], "name": "Teaching Agents in their Own Style: Blackbox Teacher Distillation 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": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebKD, which repairs failed student web-agent trajectories with minimal teacher edits and trains on corrections that preserve the student's reasoning style.", "topics": ["data-generation", "web-navigation", "efficiency"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.19834"}
{"aliases": [], "authors": ["Chengguang Gan", "Yunhao Liang", "Hanjun Wei", "Qinghao Zhang", "Zhixi Cai", "Shiwen Ni"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "47488d3b5450753b4e1a4b0991e50dc92fbab197a9b54bd8d4e3e1a3d762ee96", "submitted_at": "2026-10-14T12:39:51Z", "updated_at": "2026-10-14T12:39:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.20107v1"}}, "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:2610.20107", "identifiers": {"arxiv": "2610.20107"}, "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/2610.20107"}], "name": "Why Generate a Click? Plug-in Decision Heads 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": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Adds decision heads to a frozen vision-language generator for selecting web actions and targets, with a confidence gate that falls back to text generation.", "topics": ["efficiency", "grounding", "web-navigation"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.20107"}
{"aliases": [], "authors": ["Md Farhan Ishmam", "Minh Pham Dinh", "Kenneth Marino"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0f064032a0b0d673273d90a9250c76a6b823b1728761635b8a555a0bd96b888a", "submitted_at": "2026-10-15T18:37:41Z", "updated_at": "2026-10-15T18:37:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22539v1"}}, "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": "2026-10-20T10:34:07.486792+00:00", "role": "primary", "sha256": "0c95ddb42bbfc91aabf25e2dc2d2393a9977b6d979a2c891c5bcf80798dbfbc3", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.22539v1"}], "id": "arxiv:2610.22539", "identifiers": {"arxiv": "2610.22539"}, "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/2610.22539"}], "name": "Don't Retrain, Remix: Adaptation of Web Agents using Mixture Hypernetworks", "notes": "Reviewed the complete retained abstract and selected exact-v1 method, evaluation and analysis sections. Adapter-bank and router training use source demonstrations. Domain scouting accesses target training-task intents and answer-stripped reference procedures; version scouting is task-free. No target-time gradients does not mean no target task information. No independent reproduction or artifact/license verification.", "published_at": "2026-10-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebMix uses a scout-produced interface manual to route mixtures of source-trained LoRA adapters for web agents, adapting to target interfaces without target-time gradient updates.", "topics": ["web-navigation", "imitation-learning", "robustness"], "updated_at": "2026-10-15", "url": "https://arxiv.org/abs/2610.22539"}
{"aliases": [], "authors": ["Umar Iqbal"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "077dd623c980813e83197a38b6545161a3d2fc842d99fe291c53372b9770acd2", "submitted_at": "2026-10-15T22:02:45Z", "updated_at": "2026-10-15T22:02:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22697v1"}}, "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": "2026-10-20T13:38:46.444001Z", "role": "primary", "sha256": "cb6a8a9c0cc47be21b36b3d1831b26c6189d8f2e4af0d7e26d643c9c2a65452f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.22697v1"}], "id": "arxiv:2610.22697", "identifiers": {"arxiv": "2610.22697"}, "kind": "paper", "last_verified_at": "2026-10-20T13:38:46.444001Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.22697"}], "name": "What It Is, Not What It Says: Securing Agentic Browsing Interactions with Web Semantics", "notes": "Evaluation instructs the model to simulate compromise and measures planner outputs and permission appearance. Utility uses plan matching and LLM-judged element suitability; no steps are executed in that assessment. Assumes an honest first party, trusted browser/harness and untampered instrumentation; multi-page and availability attacks are outside scope. Parallel latency reduction is an unimplemented estimate. No independent reproduction.", "published_at": "2026-10-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Intact records webpage element provenance and capabilities, isolates planning per element, and uses a browser harness to constrain actions and request consent for third-party interactions.", "topics": ["safety-security", "web-navigation", "planning", "human-in-the-loop"], "updated_at": "2026-10-15", "url": "https://arxiv.org/abs/2610.22697"}
{"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": ["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": [], "details": {"component_path": "WebEvolver", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/License.txt"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate."}}, "repository_tree_sha": "a82450e1b5e15070bee79ee8b79f8d9a97cb85f7", "repository_url": "https://github.com/Tencent/SelfEvolvingAgent"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "447dfa1e52789f7f3f111e9e0d453fb1f9fd8bf2d7b858f28fbe21a2239b4b2e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/WebEvolver/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T15:37:01Z", "role": "primary", "sha256": "4214af42bcadf1ea52a46c2008f8e741f0aa955a58f346d9880858589be7de1d", "supports": ["identity"], "url": "https://api.github.com/repos/Tencent/SelfEvolvingAgent"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:25:25Z", "role": "primary", "sha256": "ad0f4ae4b58b5a0e84feeb5df717929b97ea37cb945c17c98510ba54509c99e7", "supports": ["identity", "release_status", "details.repository_tree_sha"], "url": "https://api.github.com/repos/Tencent/SelfEvolvingAgent/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:13Z", "role": "primary", "sha256": "873a78a46309fa3892d94b72cf73fd55542bae0e0ba8235afb26a848a64a8255", "supports": ["identity"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:19Z", "role": "primary", "sha256": "1d7a5369af7ec336bc173dba6cfee9ad0898e527f96c5b7a10cee930806c67a3", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/License.txt"}], "id": "component:tencent/selfevolvingagent/webevolver", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-18T04:27:14Z", "licenses": {"code": {"basis": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate.", "declared": "MIT", "retrieved_at": "2026-10-18T04:35:19Z", "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/License.txt"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Tencent/SelfEvolvingAgent/tree/main/WebEvolver"}], "name": "WebEvolver code component", "notes": "Scoped to the WebEvolver directory; the repository also hosts WebCoT and VScan, which are not merged into this resource. Training and model services require separate setup. Root License.txt explicitly declares MIT for SelfEvolvingAgent and separately identifies CogKernel’s Apache-2.0 dependency terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/WebEvolver/README.md", "target_id": "arxiv:2504.21024", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Component for web-agent trajectory collection, supervised-data preparation and world-model-based trajectory synthesis.", "topics": ["web-navigation", "world-models", "data-generation", "self-reflection"], "updated_at": null, "url": "https://github.com/Tencent/SelfEvolvingAgent/tree/main/WebEvolver"}
{"aliases": [], "authors": [], "details": {"availability": "Public Web Agent source and Chrome-extension archive manifest with local setup instructions.", "component_path": "real_agents/web_agent", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/xlang-ai/OpenAgents/HEAD/LICENSE"], "reason": "OpenAgents source license, scoped here to the reviewed Web Agent component; bundled third-party components retain their notices."}}, "repository_revision": "ff2e46440699af1324eb25655b622c4a131265bb"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "fb6ef7fdb133612467e27ffa642ccbe73cfb4bd8fb0c63c1bc629ad686694e1d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/xlang-ai/OpenAgents/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "b843b0519dc74bf6dbf693e0274e8ef8eb1ade94a656d3ab23faac2df4f7c9fa", "supports": ["identity"], "url": "https://api.github.com/repos/xlang-ai/OpenAgents"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:05Z", "role": "primary", "sha256": "cf5a65d5a6b7ba2a9cf950421f6902534d523f7e8d850d1b23bd5395fd3d04c8", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/xlang-ai/OpenAgents/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "bcd2232e1ec4ad9ba0628cddd30031c364d62c6f112d8ecc2c24b9a4d35696af", "supports": ["release_status", "notes", "environments"], "url": "https://raw.githubusercontent.com/xlang-ai/OpenAgents/HEAD/frontend/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:45:23Z", "role": "primary", "sha256": "5b7084cfe6149602740cf4cd85e958142aed3aea0a0a0345a82ec1618ceb7b19", "supports": ["summary", "environments", "scope"], "url": "https://raw.githubusercontent.com/xlang-ai/OpenAgents/HEAD/real_agents/web_agent/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:09Z", "role": "primary", "sha256": "a86eea1027b8d92a68ded4231671379388df757992fa69d3e8fc8f8115d124fc", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/xlang-ai/OpenAgents/HEAD/LICENSE"}], "id": "component:xlang-ai/openagents/web_agent", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-17T12:30:54Z", "licenses": {"code": {"basis": "OpenAgents source license, scoped here to the reviewed Web Agent component; bundled third-party components retain their notices.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T12:34:09Z", "url": "https://raw.githubusercontent.com/xlang-ai/OpenAgents/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/xlang-ai/OpenAgents/tree/main/real_agents/web_agent"}], "name": "OpenAgents Web Agent", "notes": "This component record covers real_agents/web_agent and the documented WeBot Chrome-extension integration. The broader Data Agent and Plugins Agent are not treated as GUI-control assets. The source manifest includes the agent implementation and frontend/webot_extension.zip; the archive was not downloaded or executed. The README describes Docker deployment as under development and recommends source deployment. Apache-2.0 applies to project source; third-party extension/dependency notices remain separate.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/xlang-ai/OpenAgents/HEAD/README.md", "target_id": "arxiv:2310.10634", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "The OpenAgents browser-control component, connecting a Chrome extension to a web agent and local chat platform for navigating sites and completing interface tasks.", "topics": ["web-navigation", "automation", "tool-integration"], "updated_at": null, "url": "https://github.com/xlang-ai/OpenAgents/tree/main/real_agents/web_agent"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:07Z", "role": "primary", "sha256": "b68b2cc26bfc6a654fec35f3c53c1b1334f3d617671673407d3ed3b59f3db973", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ai-nikolai/StateAct/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "38f8c08395c6e67ff9bf52b27e772667f193a172bd6f726507c57776ad3e5e5f", "supports": ["identity"], "url": "https://api.github.com/repos/ai-nikolai/StateAct"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:07Z", "role": "primary", "sha256": "97841e6eab1002bcea949e031f5abd3567482ebbcf03b991bd64643027c468bf", "supports": ["release_status"], "url": "https://api.github.com/repos/ai-nikolai/StateAct/git/trees/main?recursive=1"}], "id": "github:ai-nikolai/stateact", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:02:07Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ai-nikolai/StateAct"}], "name": "StateAct", "notes": "The repository contains a WebShop runner as well as ALFWorld and TextCraft experiments. Inclusion concerns its web-shopping interaction component. No code license or downloadable trained checkpoint is declared in the inspected root metadata.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ai-nikolai/StateAct/HEAD/README.md", "target_id": "arxiv:2410.02810", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "State-tracking prompts and experiment runners for language agents, including a WebShop implementation alongside text-based environments.", "topics": ["planning", "state-tracking", "web-navigation"], "updated_at": null, "url": "https://github.com/ai-nikolai/StateAct"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "3e5f24b3175afeba8635dacc7b66eba1ce89fea49996a20281ff2df5a41f1e24", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "ae4f966004dd672fd5b994add5a5cba1fee9233d4d81b224e94c339728fafd87", "supports": ["identity"], "url": "https://api.github.com/repos/amazon-science/PAE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "b1e24a3f38fae67f71033a3f920164d282d3f8bb072f438349dbe9eb23730a7a", "supports": ["release_status"], "url": "https://api.github.com/repos/amazon-science/PAE/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:15Z", "role": "primary", "sha256": "db2e35513dbadcdc67f5819a3bfee2777786538dd3531620cd5fbd4b6ed6e538", "supports": ["licenses"], "url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/LICENSE.txt"}], "id": "github:amazon-science/pae", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:02:06Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/amazon-science/PAE"}], "name": "Proposer-Agent-Evaluator (PAE)", "notes": "Source code and browser-environment modules are present. The repository LICENSE is CC-BY-NC-4.0, while its README describes that license as applying to a dataset, leaving code scope ambiguous. The separately linked dataset is gated and was not accessed; model and dataset terms must be checked separately.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/README.md", "target_id": "arxiv:2412.13194", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A browser-agent training system that proposes tasks, collects visual interaction trajectories, and uses automated evaluation to drive policy improvement.", "topics": ["reinforcement-learning", "web-navigation", "data-generation"], "updated_at": null, "url": "https://github.com/amazon-science/PAE"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "0d4b28c993ccb977c7910ff3fecb5de29db364a7a7ad55523c8e052c9bae2566", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "b02ec75f69ab8bbebe0d59eeccdcdcacf11a0d6b00bdfe803c692a9d587301e2", "supports": ["identity"], "url": "https://api.github.com/repos/ApGa/Go-Browse"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "cd3a40a596ad9ead856973bb31e5b7f7e88a2f60969799c01a2a8a6d09fe3421", "supports": ["release_status"], "url": "https://api.github.com/repos/ApGa/Go-Browse/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "e7107f1e1b3a0c65cbbae6c496d9e71b570025bdddebb3d402a6080dd916bcba", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "e7107f1e1b3a0c65cbbae6c496d9e71b570025bdddebb3d402a6080dd916bcba", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/LICENSE"}], "id": "github:apga/go-browse", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:44:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T02:45:20Z", "url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ApGa/Go-Browse"}], "name": "Go-Browse", "notes": "The release contains exploration agents, data-processing scripts, fine-tuning and WebArena evaluation entry points under MIT. Checkpoints and processed datasets are separate Hub resources with separate license status. The arXiv badge image has an unrelated number; the linked paper and citation consistently identify 2506.03533.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/README.md", "target_id": "arxiv:2506.03533", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Structured website-exploration and data-processing code that revisits discovered pages, proposes tasks and collects trajectories for web-agent fine-tuning.", "topics": ["web-navigation", "data-generation", "imitation-learning"], "updated_at": null, "url": "https://github.com/ApGa/Go-Browse"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "fcca30a036d122e34c943740c29d8cb91642f0c615050a8d54ec891c0eee56d2", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/browser-use/browser-use/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "135c57671b0ee957669b05a0ca00cc06f645496d351e2e732832c6e276da9a5e", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/browser-use/browser-use/HEAD/LICENSE"}], "id": "github:browser-use/browser-use", "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/browser-use/browser-use/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/browser-use/browser-use"}], "name": "Browser Use", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/browser-use/browser-use/HEAD/README.md", "target_id": "organization:browser-use", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Open-source browser agent library that runs locally with any LLM against a local or cloud browser, and also serves as a CLI for coding agents and as a driver for Claude's browser toolset.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/browser-use/browser-use"}
{"aliases": [], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "release_tag": "v3.0.0"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "5047bcf2f0e92f0d34a57ac7d83166903a16a9909c700c0647b8a48363bd94f9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/browser-use/web-ui/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "7fb0738bc9b837f7b3e645c406e04d5c7c4d0860ea636a96e4e9d84d9510dd82", "supports": ["identity"], "url": "https://api.github.com/repos/browser-use/web-ui"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "fcbf1775e3513625103d389c737e98e939e17140baac9bc8000b582ab2729a46", "supports": ["release_status"], "url": "https://api.github.com/repos/browser-use/web-ui/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:00Z", "role": "primary", "sha256": "f87ee752c38653c447eed4a9abd83e71dedd40298758bcb9aa993b731362c453", "supports": ["release_status"], "url": "https://api.github.com/repos/browser-use/web-ui/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:01Z", "role": "primary", "sha256": "612d1382ec1a426eaf3672aafb13bafcfe91b2c90a0fe339822ffc9c895003ca", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/browser-use/web-ui/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T04:01:01Z", "role": "primary", "sha256": "612d1382ec1a426eaf3672aafb13bafcfe91b2c90a0fe339822ffc9c895003ca", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/browser-use/web-ui/HEAD/LICENSE"}], "id": "github:browser-use/web-ui", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T03:59:54Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T04:01:01Z", "url": "https://raw.githubusercontent.com/browser-use/web-ui/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/browser-use/web-ui"}], "name": "Browser Use WebUI", "notes": "The repository publishes UI source and release v3.0.0 under MIT. It is a distinct user interface built on the existing Browser Use framework, not a second framework identity. The UI supports several external model providers whose terms are separate; no dependencies were installed.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A Gradio interface for configuring and running Browser Use agents, including persistent browser sessions and interaction recordings.", "topics": ["web-navigation", "tool-integration", "recording"], "updated_at": null, "url": "https://github.com/browser-use/web-ui"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "fb2039ee569e37cf1d652ca26c0a1951a318d4ac3240018b958985c0aec91007", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/browserbase/stagehand/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "fac7e7cd1cfc504054ee1fc53c38e0f3099d7357917e85a73792f6511fe0ea96", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/browserbase/stagehand/HEAD/LICENSE"}], "id": "github:browserbase/stagehand", "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/browserbase/stagehand/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/browserbase/stagehand"}], "name": "Stagehand", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/browserbase/stagehand/HEAD/README.md", "target_id": "organization:browserbase", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Browser automation framework combining code and natural-language actions.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/browserbase/stagehand"}
{"aliases": [], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/browseros-ai/BrowserOS/HEAD/LICENSE", "https://raw.githubusercontent.com/browseros-ai/BrowserOS/HEAD/README.md"], "reason": "The repository explicitly identifies its code as AGPL-3.0; this declaration does not cover every incorporated third-party component."}}, "release_tag": "browserclaw/v0.51.1.2"}, "environments": ["web", "desktop", "windows", "macos", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "bfe2979ae0e99985154042090f483e172cca229fe408f8f9b47102662beec983", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.code"], "url": "https://raw.githubusercontent.com/browseros-ai/BrowserOS/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "2b8c90b9fb2edffccf7406952465015df87622331921fe0f25a44fc9f5a9b86b", "supports": ["identity"], "url": "https://api.github.com/repos/browseros-ai/BrowserOS"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "1f43b364f532a8bdd798df5dd86d3b73f8f545ef73a66e9b3b4da59ea9d0faa1", "supports": ["release_status"], "url": "https://api.github.com/repos/browseros-ai/BrowserOS/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:59Z", "role": "primary", "sha256": "87bea16220429a93fefc2b2a57452fd67b2618784025db18eb1f899c0396098f", "supports": ["release_status"], "url": "https://api.github.com/repos/browseros-ai/BrowserOS/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:59Z", "role": "primary", "sha256": "8486a10c4393cee1c25392769ddd3b2d6c242d6ec7928e1414efff7dfb2f07ef", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/browseros-ai/BrowserOS/HEAD/LICENSE"}], "id": "github:browseros-ai/browseros", "identifiers": {}, "kind": "infrastructure", "last_verified_at": "2026-10-17T03:59:54Z", "licenses": {"code": {"basis": "Repository LICENSE and explicit README declaration; incorporated Chromium and other components retain their own terms.", "declared": "AGPL-3.0", "retrieved_at": "2026-10-17T04:00:59Z", "url": "https://raw.githubusercontent.com/browseros-ai/BrowserOS/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/browseros-ai/BrowserOS"}], "name": "BrowserOS / BrowserOS neo", "notes": "The one repository includes BrowserOS neo and BrowserOS; they are retained as one source-bundle resource. Official release metadata lists desktop installers, but none were downloaded. The repository declares AGPL-3.0 and retains separate Chromium-related licensing. Publisher privacy and performance claims were not independently verified.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Chromium-based browser implementations for external agent control and built-in assistant workflows, with session visibility and replay facilities.", "topics": ["web-navigation", "tool-integration", "recording"], "updated_at": null, "url": "https://github.com/browseros-ai/BrowserOS"}
{"aliases": ["github:bytedance/agent-r", "bytedance/Agent-R", "https://github.com/bytedance/Agent-R"], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:20Z", "role": "primary", "sha256": "8471a0761a94b1af84099a80efb6ac53f17e34491a196e82ca068a5113c0b8be", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ByteDance-Seed/Agent-R/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:19Z", "role": "primary", "sha256": "3fff7137bc4f39a3a757f3b820892fb330449044ef2721e9f9d7719d23ff2b98", "supports": ["identity", "availability"], "url": "https://api.github.com/repos/ByteDance-Seed/Agent-R"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "e3c70b04094e8e77e249065b2b500ac197fc6fafee6e3275d78289b3790bf7ed", "supports": ["release_status"], "url": "https://api.github.com/repos/ByteDance-Seed/Agent-R/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "283ea6cc2997a1a70da0049e09adf9317bb60ca1b51279b65196b83a69e1996b", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ByteDance-Seed/Agent-R/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "283ea6cc2997a1a70da0049e09adf9317bb60ca1b51279b65196b83a69e1996b", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ByteDance-Seed/Agent-R/HEAD/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T18:05:46Z", "role": "primary", "sha256": "3fff7137bc4f39a3a757f3b820892fb330449044ef2721e9f9d7719d23ff2b98", "supports": ["aliases"], "url": "https://api.github.com/repos/bytedance/Agent-R"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T18:05:46Z", "role": "primary", "sha256": "3fff7137bc4f39a3a757f3b820892fb330449044ef2721e9f9d7719d23ff2b98", "supports": ["aliases"], "url": "https://api.github.com/repos/ByteDance-Seed/Agent-R"}], "id": "github:bytedance-seed/agent-r", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:03:20Z", "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/ByteDance-Seed/Agent-R/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ByteDance-Seed/Agent-R"}], "name": "Agent-R", "notes": "The canonical repository is ByteDance-Seed/Agent-R and is archived but readable. MCTS collection, revision and evaluation scripts are released under Apache-2.0. It also supports non-GUI text environments; this entry covers its WebShop interaction component. No separate weight release was verified.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ByteDance-Seed/Agent-R/HEAD/README.md", "target_id": "arxiv:2501.11425", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Trajectory collection and revision utilities that use tree search and model-guided critique to train agents to recover from mistakes, including WebShop tasks.", "topics": ["self-reflection", "data-generation", "web-navigation"], "updated_at": null, "url": "https://github.com/ByteDance-Seed/Agent-R"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/caap-agent/caap-agent/HEAD/LICENSE"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate. The inspected text does not establish -only versus -or-later for the project."}}, "repository_tree_sha": "9d65c0f30e3c119db533dd832863b4f3f7b74b18", "repository_url": "https://github.com/caap-agent/caap-agent"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "481db1288ffc769f48b19c83f0db4fb40a64dc1b05859f776b77f0774bfcdb90", "supports": ["name", "kind", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/caap-agent/caap-agent/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T15:37:02Z", "role": "primary", "sha256": "8d518cad746591e82b54b63b79d0a1d47531a75f02c8baf2b560877fb03918ba", "supports": ["identity"], "url": "https://api.github.com/repos/caap-agent/caap-agent"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:25:24Z", "role": "primary", "sha256": "c511d94ce58e73a87c9647169b1199c0ed0029e6b1df271ad1fcd3ca86f3458b", "supports": ["identity", "release_status", "details.repository_tree_sha"], "url": "https://api.github.com/repos/caap-agent/caap-agent/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["summary", "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"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:20Z", "role": "primary", "sha256": "5e6fa1bfd473724e67366c42761db27fe091ee00fdcfbe599b9dd5d5a9311d36", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/caap-agent/caap-agent/HEAD/LICENSE"}], "id": "github:caap-agent/caap-agent", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-18T04:27:14Z", "licenses": {"code": {"basis": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate. The inspected text does not establish -only versus -or-later for the project.", "declared": "AGPL-3.0", "retrieved_at": "2026-10-18T04:35:20Z", "url": "https://raw.githubusercontent.com/caap-agent/caap-agent/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/caap-agent/caap-agent"}], "name": "CAAP Agent", "notes": "Code is released for the two named browser benchmarks. Visual-observer checkpoints are linked separately through Google Drive and were not downloaded or access-tested. Root LICENSE begins with GNU AGPL v3 and includes third-party notices; this code declaration is not transferred to bundled/generated data or external weights.", "published_at": null, "relations": [{"evidence_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", "target_id": "arxiv:2406.06947", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Screenshot-driven agent implementation for MiniWoB++ and WebShop, using a visual observer and context-aware prompts to issue mouse and keyboard actions.", "topics": ["planning", "screen-understanding", "web-navigation"], "updated_at": null, "url": "https://github.com/caap-agent/caap-agent"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "fc3099072ca60acf2294a63dd2c3e343e90c901a7c7581f723cfdccc5a1edc2b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/EmergenceAI/Agent-E/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:04Z", "role": "primary", "sha256": "555ac6a8bf7ef4a3c82f4ee258be9d751042f8d2a6c901e5eac7c43596563c75", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/EmergenceAI/Agent-E/HEAD/LICENSE.txt"}], "id": "github:emergenceai/agent-e", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:04Z", "url": "https://raw.githubusercontent.com/EmergenceAI/Agent-E/HEAD/LICENSE.txt"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/EmergenceAI/Agent-E"}], "name": "Agent E", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/EmergenceAI/Agent-E/HEAD/README.md", "target_id": "arxiv:2407.13032", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Browser automation agent on the AG2 framework that pairs a planner agent with a browser navigation agent to fill forms, search shops and work through sites, also callable over a local HTTP API.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/EmergenceAI/Agent-E"}
{"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": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "9fe74ee087156efde4531a058ccd3abad504032e0bd994979992ed4164c8704b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/hyperbrowserai/HyperAgent/HEAD/README.md"}], "id": "github:hyperbrowserai/hyperagent", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/hyperbrowserai/HyperAgent"}], "name": "HyperAgent", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/hyperbrowserai/HyperAgent/HEAD/README.md", "target_id": "organization:hyperbrowser", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Library adding natural-language commands to Playwright: accessibility-tree single actions, multi-step tasks that can use screenshots, an MCP client and action caching for replay without LLM calls.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/hyperbrowserai/HyperAgent"}
{"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": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "7e4d38b95657b604f5f390e25762d94c8d59164fbb2ae6c140c126bb128dbd4f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/lavague-ai/LaVague/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/lavague-ai/LaVague/HEAD/LICENSE"}], "id": "github:lavague-ai/lavague", "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/lavague-ai/LaVague/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/lavague-ai/LaVague"}], "name": "LaVague", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Open-source web agent framework where a World Model turns an objective and the current page into instructions and an Action Engine compiles them into Selenium or Playwright code to run.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/lavague-ai/LaVague"}
{"aliases": [], "authors": [], "details": {"availability": "Public memory, prompting, browser-experiment and fine-tuning source.", "license_conflict": "README MIT badge versus NTUITIVE agreement; the agreement contains inconsistent distribution provisions.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/ltzheng/synapse/HEAD/README.md", "https://raw.githubusercontent.com/ltzheng/Synapse/HEAD/LICENSE"], "reason": "Unresolved conflict between README MIT badge and NTUITIVE non-commercial LICENSE, including inconsistent redistribution clauses. Keep permissions unknown rather than choose an interpretation."}}, "repository_revision": "08c3a252ff158822e50e9c109c1cd49dc2f6147d"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "ed5416d91665017fca4b354016b7c2c748acf32d7723483fecb5387a22a1fdff", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.code"], "url": "https://raw.githubusercontent.com/ltzheng/synapse/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "79a238a6f486d2a8970baba199f0968d28f14d42922c86ce71753eb687e0e9c6", "supports": ["identity"], "url": "https://api.github.com/repos/ltzheng/synapse"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:05Z", "role": "primary", "sha256": "785a7163dac9f41b1d0ca97f8dd9311b6df6a09f3a5163438dd64db226a4ceaf", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/ltzheng/Synapse/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["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": "2026-10-17T12:34:09Z", "role": "primary", "sha256": "cc4bce2274289d2b924fcc3c185c68b8e7c539365c5f8c72be865883244088ec", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ltzheng/Synapse/HEAD/LICENSE"}], "id": "github:ltzheng/synapse", "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/ltzheng/Synapse"}], "name": "Synapse", "notes": "Training, memory-building and evaluation scripts for MiniWoB++ and Mind2Web are present. External benchmark data, model/API credentials and legacy dependencies are still required. License remains unknown: the README shows MIT, while LICENSE is an NTUITIVE non-commercial agreement that both permits some distribution and separately says distribution is prohibited. No interpretation resolves those contradictions, and no commercial or non-commercial permission is asserted. This admission confirms the released GUI implementation, not a licensing grant.", "published_at": null, "relations": [{"evidence_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", "target_id": "arxiv:2306.07863", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A computer-control research implementation that retrieves example trajectories from memory, abstracts task-relevant state and prompts an agent to navigate browser tasks.", "topics": ["memory", "planning", "web-navigation", "imitation-learning"], "updated_at": null, "url": "https://github.com/ltzheng/Synapse"}
{"aliases": [], "authors": [], "details": {"availability": "Public conversational planning, candidate-generation and training/evaluation source.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/magicgh/Self-MAP/HEAD/LICENSE"], "reason": "Project source LICENSE; separate dataset license is not folded into the code slot."}}, "repository_revision": "6cf779682b70dae597626f96dfa5130b202689bc"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:55Z", "role": "primary", "sha256": "693197cb807e3ab4d5afda9de35dd51342b4feb18e68b8ccfcdaa370399dbeb3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/magicgh/self-map/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:54Z", "role": "primary", "sha256": "82c425841c94bf129c4fbd1e70029c581aa1488d5f85cdd23aa0f292092146f0", "supports": ["identity"], "url": "https://api.github.com/repos/magicgh/self-map"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:05Z", "role": "primary", "sha256": "8d844f227c457321640231ab3a4953457e3c6cfeab8d8f4803f8aa01dc15a8f8", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/magicgh/Self-MAP/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "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=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "f271c479b19a620600d18fc2265224205bf71483b969672d3f6acd9bc552b3f5", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/magicgh/Self-MAP/HEAD/LICENSE"}], "id": "github:magicgh/self-map", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T12:30:55Z", "licenses": {"code": {"basis": "Project source LICENSE; separate dataset license is not folded into the code slot.", "declared": "MIT", "retrieved_at": "2026-10-17T12:34:10Z", "url": "https://raw.githubusercontent.com/magicgh/Self-MAP/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/magicgh/Self-MAP"}], "name": "Self-MAP", "notes": "The source tree provides candidate-generation and action-prediction training/evaluation scripts. Its documented workflow requires separately downloaded MT-Mind2Web data and configured model/API paths. The README cites the ACL publication by DOI; a retained original arXiv abstract explicitly names Self-MAP and MT-Mind2Web and supplies the catalog paper relation without a new bibliography entity. MIT is the code license; the separate data artifact declares CC-BY-4.0.", "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=800&max_results=100&sortBy=submittedDate&sortOrder=descending", "target_id": "arxiv:2402.15057", "type": "documented_by"}, {"evidence_url": "https://raw.githubusercontent.com/magicgh/self-map/HEAD/README.md", "target_id": "url:bdfb50330deb97893ce4", "type": "uses"}], "review_status": "curated", "scope": "core", "summary": "A conversational web-agent implementation with candidate ranking and memory-assisted action planning, including optional memory simplification, refinement and self-reflection modules.", "topics": ["web-navigation", "memory", "self-reflection", "planning"], "updated_at": null, "url": "https://github.com/magicgh/Self-MAP"}
{"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": ["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": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:00Z", "role": "primary", "sha256": "ce11bdfe8354eb9c3b8a985dd5ddd21f88f94fb5fb9d00c9b01a76226f0b0a36", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/microsoft/fara/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:00Z", "role": "primary", "sha256": "98a96c38494df4963951e32b81ec4effebe4c8a812ab8d9e7d185d59e86fe2fc", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/microsoft/fara/HEAD/LICENSE"}], "id": "github:microsoft/fara", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:36:00Z", "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/microsoft/fara/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/microsoft/fara"}], "name": "Fara", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/microsoft/fara/HEAD/README.md", "target_id": "arxiv:2511.19663", "type": "documented_by"}, {"evidence_url": "https://raw.githubusercontent.com/microsoft/fara/HEAD/README.md", "target_id": "arxiv:2606.20785", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Microsoft browser-use agent models that act from screenshots by predicting click coordinates without accessibility trees; Fara1.5 comes in 4B, 9B and 27B, succeeding Fara-7B.", "topics": ["planning", "web-navigation"], "updated_at": null, "url": "https://github.com/microsoft/fara"}
{"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": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:34Z", "role": "primary", "sha256": "3095f7262b7d4ff916c6f58c55e6fc9902c08f77d09bf3951199945501148bb3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/MurtyShikhar/NNetnav/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:42Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/MurtyShikhar/NNetnav/HEAD/LICENSE"}], "id": "github:murtyshikhar/nnetnav", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T10:42:34Z", "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/MurtyShikhar/NNetnav/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/MurtyShikhar/NNetnav"}], "name": "NNetNav", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/MurtyShikhar/NNetnav/HEAD/README.md", "target_id": "arxiv:2410.02907", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Collects and labels browser-exploration trajectories for agent training, with code for data processing, fine-tuning, and web evaluation.", "topics": ["data-generation", "web-navigation"], "updated_at": null, "url": "https://github.com/MurtyShikhar/NNetnav"}
{"aliases": [], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "release_tag": "v0.3.0"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "4d8a0b0d6286c4bab116c155f087948062166e7e169d6bbaf504fedb4f32dc67", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/nanobrowser/nanobrowser/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "6b48386f32a971a724906de8b6d5ea53036f3813411ddc12931be3010c034400", "supports": ["identity"], "url": "https://api.github.com/repos/nanobrowser/nanobrowser"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "afd0b71b01d3d89033d8f32abaf9871f38e8fb0c50442ed3a8cbb3079d4a4fe0", "supports": ["release_status"], "url": "https://api.github.com/repos/nanobrowser/nanobrowser/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:00Z", "role": "primary", "sha256": "c7b28db1b1c8e78d100da066b08db713bc144baff99941b6c94489ef61baee8f", "supports": ["release_status"], "url": "https://api.github.com/repos/nanobrowser/nanobrowser/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:00Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/nanobrowser/nanobrowser/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T04:01:00Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/nanobrowser/nanobrowser/HEAD/LICENSE"}], "id": "github:nanobrowser/nanobrowser", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-17T03:59:54Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T04:01:00Z", "url": "https://raw.githubusercontent.com/nanobrowser/nanobrowser/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/nanobrowser/nanobrowser"}], "name": "Nanobrowser", "notes": "The official repository publishes extension source and a v0.3.0 ZIP release. It names Chrome and Edge as supported browsers. Apache-2.0 is declared for the code. Ask/Auto behavior and provider choices are documented, but no security guarantee, model performance or service-price claim is inferred.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A browser-extension agent that reads pages, interacts with elements and moves between tabs from a side-panel task interface.", "topics": ["web-navigation", "automation", "human-in-the-loop"], "updated_at": null, "url": "https://github.com/nanobrowser/nanobrowser"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:20Z", "role": "primary", "sha256": "2fb7c0d70ccfa93605d5250a8f6d5182324f0c53d70e1c35ba18cb00f4ada548", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/night-chen/WorkForceAgent-R1/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:19Z", "role": "primary", "sha256": "0ad3ef1d900200e3ecab86cd44f79038000e8748fb285cdd5f32882455d2e36e", "supports": ["identity"], "url": "https://api.github.com/repos/night-chen/WorkForceAgent-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:39Z", "role": "primary", "sha256": "521d065188a6354416c1aafe88a0eb34486f2c2ec011c1cdf8ed008760dd9c55", "supports": ["release_status"], "url": "https://api.github.com/repos/night-chen/WorkForceAgent-R1/git/trees/main?recursive=1"}], "id": "github:night-chen/workforceagent-r1", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:03:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/night-chen/WorkForceAgent-R1"}], "name": "WorkForceAgent-R1", "notes": "The release contains WorkArena preprocessing and training source. The inspected root has no explicit project license; dependencies retain their own terms. Although the README discusses trained 7B/14B models, it does not provide a verified checkpoint release, so this is cataloged only as a framework.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/night-chen/WorkForceAgent-R1/HEAD/README.md", "target_id": "arxiv:2505.22942", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Web-task reward functions, WorkArena data processing and GRPO training configurations for reasoning-oriented language agents.", "topics": ["reinforcement-learning", "web-navigation", "planning"], "updated_at": null, "url": "https://github.com/night-chen/WorkForceAgent-R1"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "5ecb73264cce10b75ec04dea59aaff77ffde552259fba6932507779a35681c5b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/nottelabs/notte/HEAD/README.md"}], "id": "github:nottelabs/notte", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/nottelabs/notte"}], "name": "Notte", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Web agent framework and cloud SDKs for managed Chrome sessions, adding a CLI, an MCP server for coding agents, serverless browser functions and an API that builds automations from English tasks.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/nottelabs/notte"}
{"aliases": [], "authors": [], "details": {"availability": "Public browser-agent source and npm release metadata.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/omxyz/lumen/HEAD/LICENSE"], "reason": "Publisher project-source license; external model and browser-service terms are separate."}}, "package_version": "0.2.0", "repository_revision": "b1ad26a0784645ac3a97d402db99cd5d17f86334"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:55Z", "role": "primary", "sha256": "7708f434142498959bb5ed5de35ac1b945f9ad6e070e436b4e73819b7db1df7f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/omxyz/lumen/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:55Z", "role": "primary", "sha256": "32b02ea43d19929d82e8b8c52bb8e448553aebdd96a8754d144927cb0f34c5b6", "supports": ["identity"], "url": "https://api.github.com/repos/omxyz/lumen"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:05Z", "role": "primary", "sha256": "9911953e501a245d657f3835b22c6822ad7d990b96f03e3cba93a15ebe4f70a6", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/omxyz/lumen/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "34a03b6dd8cd1960d5592a6569e2a1c1f6c1b2697da1386a58d0ca64157f059e", "supports": ["identity", "release_status", "details.package_version", "notes"], "url": "https://registry.npmjs.org/@omxyz%2Flumen/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "30dd5336a8a45b548b9051e751d45923a9864b3810a36c836df9b07447ca3c5c", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/omxyz/lumen/HEAD/LICENSE"}], "id": "github:omxyz/lumen", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-17T12:30:55Z", "licenses": {"code": {"basis": "Publisher project-source license; external model and browser-service terms are separate.", "declared": "MIT", "retrieved_at": "2026-10-17T12:34:10Z", "url": "https://raw.githubusercontent.com/omxyz/lumen/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/omxyz/lumen"}], "name": "Lumen browser agent", "notes": "Public TypeScript source and npm package metadata are available; the inspected package version is 0.2.0. The package declares Node ^20.19.0 or >=22.12.0 and requires a browser and model provider. Its README benchmark is explicitly preliminary and uses a small WebVoyager subset; this catalog makes no comparative performance or independent safety claim. Referenced research influences are not asserted as papers introducing Lumen.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A screenshot-driven browser agent over Chrome DevTools Protocol, with resumable sessions, context compression, action replay and configurable completion checks.", "topics": ["web-navigation", "memory", "automation", "observability"], "updated_at": null, "url": "https://github.com/omxyz/lumen"}
{"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"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "673b0e309540e57ec3b977689b33f89ef80e45ba1c0ffb05e7412e4862f6b99e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/SeeAct/HEAD/README.md"}], "id": "github:osu-nlp-group/seeact", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/SeeAct"}], "name": "SeeAct", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/SeeAct/HEAD/README.md", "target_id": "arxiv:2401.01614", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Generalist web agent framework for large multimodal models such as GPT-4V, with a Playwright tool that runs agents on live websites, text-choice and set-of-mark grounding, and a crawler mode.", "topics": ["web-navigation", "grounding"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/SeeAct"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "e06dbc9d58e713a0019d5040aa21220de85266e80aacc7f2d320a0518c781496", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/SkillWeaver/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "4dfa648e22431d1a4903eb1e11f432fbe8a0f39d24e2e3324c300285ffe458fd", "supports": ["identity"], "url": "https://api.github.com/repos/OSU-NLP-Group/SkillWeaver"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "7a1c94c67131a97516d92632c00a2a20d09226fec57bb09e77ff6d84a4e6e79a", "supports": ["release_status"], "url": "https://api.github.com/repos/OSU-NLP-Group/SkillWeaver/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "93b43d692b033b76129504448695bfe76ef22d18b11e51352bfab4b5622e5aaa", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/SkillWeaver/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:45:21Z", "role": "primary", "sha256": "822a60969166874dd8a26273b8d749ef8c69c46137a0915ed7178b3265bd037f", "supports": ["relations"], "url": "https://osu-nlp-group.github.io/SkillWeaver/"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "93b43d692b033b76129504448695bfe76ef22d18b11e51352bfab4b5622e5aaa", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/SkillWeaver/HEAD/LICENSE"}], "id": "github:osu-nlp-group/skillweaver", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:44:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T02:45:20Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/SkillWeaver/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/SkillWeaver"}], "name": "SkillWeaver", "notes": "Exploration, task execution, browser integration and evaluation code are published under MIT. Synthesized skill files are included, but no trained model weights are implied. The source depends on external web environments and model services; no runtime or browser tasks were executed.", "published_at": null, "relations": [{"evidence_url": "https://osu-nlp-group.github.io/SkillWeaver/", "target_id": "arxiv:2504.07079", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A web-agent framework that explores sites, synthesizes reusable interaction APIs and builds a skill library for later tasks.", "topics": ["web-navigation", "tool-integration", "data-generation"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/SkillWeaver"}
{"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": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T13:14:29Z", "role": "primary", "sha256": "75f99d9388aee3e0b124ebdb5f33e08fbe8ec999ce1e9eba5cd71c21e60522e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/PathOnAI/LiteWebAgent/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T13:29:25Z", "role": "primary", "sha256": "ec6ef142e3900e9f5555009b825387e41da67a77a24ee72ef6963c99b065e6eb", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/PathOnAIOrg/LiteWebAgent/HEAD/LICENSE"}], "id": "github:pathonaiorg/litewebagent", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T13:14:29Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T13:29:25Z", "url": "https://raw.githubusercontent.com/PathOnAIOrg/LiteWebAgent/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/PathOnAIOrg/LiteWebAgent"}], "name": "LiteWebAgent", "notes": "The repository is maintained by the PathOnAI.org research community; its README explicitly disclaims affiliation with a for-profit company.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/PathOnAI/LiteWebAgent/HEAD/README.md", "target_id": "arxiv:2503.02950", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Modular web-agent framework with browser control, planning, workflow memory, and tree-search components, plus web and extension interfaces.", "topics": ["web-navigation", "planning", "memory", "tool-integration"], "updated_at": null, "url": "https://github.com/PathOnAIOrg/LiteWebAgent"}
{"aliases": [], "authors": [], "details": {"availability": "Public agent and browser-environment source; legacy API compatibility is unverified.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/posgnu/rci-agent/HEAD/LICENSE"], "reason": "Repository code license; not a license for the remote language-model service."}}, "repository_revision": "31bf737922fa0337d03367a844de5cf1aff4ac52"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:12Z", "role": "primary", "sha256": "6a101c867546365d21b48527781bd3af324075929067fb465c6105872070f692", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/posgnu/rci-agent/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:11Z", "role": "primary", "sha256": "b826fc1747a1bbe38a8df8da567a68f79936509c5325ec76a683706118c526f5", "supports": ["identity"], "url": "https://api.github.com/repos/posgnu/rci-agent"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:11Z", "role": "primary", "sha256": "3dcb8cc06679490315f0f3636bb2795e2f717b77d4156ad0bf2429cc36e3273c", "supports": ["release_status"], "url": "https://api.github.com/repos/posgnu/rci-agent/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:12Z", "role": "primary", "sha256": "ffa8b2a603abd9ede8eb686e87c4a433110f1be47d7cc98df4977b7fb2ee72fc", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/posgnu/rci-agent/HEAD/LICENSE"}], "id": "github:posgnu/rci-agent", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-17T09:14:12Z", "licenses": {"code": {"basis": "Repository code license; not a license for the remote language-model service.", "declared": "MIT", "retrieved_at": "2026-10-17T09:14:12Z", "url": "https://raw.githubusercontent.com/posgnu/rci-agent/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/posgnu/rci-agent"}], "name": "RCI Agent", "notes": "The public release includes the agent, prompting utilities and a bundled MiniWoB++ environment wrapper. Its README documents older hosted model names and an OpenAI-key setup; contemporary endpoint compatibility was not tested. The repository notes that its MiniWoB++ dependency is not officially supported on Windows. MIT applies to repository code; no new trained weights are claimed.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/posgnu/rci-agent/HEAD/README.md", "target_id": "arxiv:2303.17491", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A MiniWoB++ agent that uses explicit and implicit recursive criticism-and-improvement prompts to refine action plans and ground actions to browser state.", "topics": ["web-navigation", "planning", "self-reflection", "training-free"], "updated_at": null, "url": "https://github.com/posgnu/rci-agent"}
{"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": {"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": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "feee31aa4d0801cb364578099a13e4f9cdcaa4eb5fbf4312b3fc3f8edd479a14", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/SqueezeAILab/plan-and-act/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "eada6ee9aa180f55daca059c5e7a4922b320cccb0514864108eba9040b52415c", "supports": ["identity"], "url": "https://api.github.com/repos/SqueezeAILab/plan-and-act"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "10ff0a1c1abe25454541949fdaac1869d68b5d4426fbb5ef1d487657afcd26f5", "supports": ["release_status"], "url": "https://api.github.com/repos/SqueezeAILab/plan-and-act/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:15Z", "role": "primary", "sha256": "f86e78202a51ab013e1cc47ba37386134601317f5a0c4e4ed287981b52159122", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/SqueezeAILab/plan-and-act/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:02:15Z", "role": "primary", "sha256": "f86e78202a51ab013e1cc47ba37386134601317f5a0c4e4ed287981b52159122", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/SqueezeAILab/plan-and-act/HEAD/LICENSE"}], "id": "github:squeezeailab/plan-and-act", "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": "2026-10-17T02:02:15Z", "url": "https://raw.githubusercontent.com/SqueezeAILab/plan-and-act/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/SqueezeAILab/plan-and-act"}], "name": "Plan-and-Act", "notes": "Runnable entry points, framework modules and synthetic-data tooling are published. The repository code declares MIT; separate model and dataset cards have their own declarations. Its setup targets WebArena-Lite and requires that external environment.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/SqueezeAILab/plan-and-act/HEAD/README.md", "target_id": "arxiv:2503.09572", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A web-agent framework that separates high-level planning from action execution, with replanning and synthetic planner-data generation components.", "topics": ["planning", "web-navigation", "data-generation"], "updated_at": null, "url": "https://github.com/SqueezeAILab/plan-and-act"}
{"aliases": [], "authors": [], "details": {"availability": "Public historical training and environment-wrapper source; legacy dependencies documented.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/LICENSE"], "reason": "Stanford WGE source license; not a blanket license for all third-party tasks, embeddings or demonstration resources."}}, "repository_revision": "bed526657a62a64a163f80deb4a8627cb63871a3"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:16Z", "role": "primary", "sha256": "50ee5234bf62320d096fbe4c11c7be21f50c9e09dae97754a16a5bbde2f1ab19", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:15Z", "role": "primary", "sha256": "66fcd6577d1867abf84e62fe9edf2e3f223292724cbe6b9d0ef02b582db5c8b2", "supports": ["identity"], "url": "https://api.github.com/repos/stanfordnlp/wge"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:19:50Z", "role": "primary", "sha256": "926aff11e14c529d05d9bc51fe12aadb007818119da55f99c2fa47d0387f466d", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/stanfordnlp/wge/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:17Z", "role": "primary", "sha256": "249b23ff62e21bd446b3e411b1d7ea7f173c7e19e42ed75e97800ed32201bb3a", "supports": ["details", "notes"], "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/third-party/miniwob-sandbox/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:05Z", "role": "primary", "sha256": "3df3dd1e3149cd6cc1a7b052a237a6653449a7c5f18edf19a3a79294c7d541c1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/LICENSE"}], "id": "github:stanfordnlp/wge", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T11:18:16Z", "licenses": {"code": {"basis": "Stanford WGE source license; not a blanket license for all third-party tasks, embeddings or demonstration resources.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T11:23:05Z", "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/stanfordnlp/wge"}], "name": "Workflow-Guided Exploration research code", "notes": "The released repository includes training policies, experiment configuration and a MiniWoB sandbox extended for the 2018 paper. This is the WGE implementation, not another copy of the already cataloged Farama MiniWoB++ benchmark and not the original OpenAI MiniWoB release. Its documented dependency is PyTorch 0.1.12, and current runtime compatibility was not tested. Stanford code declares Apache-2.0; bundled third-party browser tasks and externally downloaded demonstration/embedding data retain their own terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/README.md", "target_id": "arxiv:1802.08802", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Research code that learns browser-control policies by constraining exploration with workflows induced from demonstrations, with a Selenium-based MiniWoB environment and experiment harness.", "topics": ["reinforcement-learning", "web-navigation", "workflow", "history"], "updated_at": null, "url": "https://github.com/stanfordnlp/wge"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "668949f121cb5a8ac262a8d3222a31eaee6692bfc0de9b86e60ae04f885ccc9b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/steel-dev/steel-browser/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/steel-dev/steel-browser/HEAD/LICENSE"}], "id": "github:steel-dev/steel-browser", "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": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/steel-dev/steel-browser/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/steel-dev/steel-browser"}], "name": "Steel Browser", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/steel-dev/steel-browser/HEAD/README.md", "target_id": "organization:steel", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Open-source browser API managing Chrome sessions for agents, reachable through Puppeteer, Playwright or Selenium over CDP, with persisted cookies, proxy rotation, stealth and a session debugging UI.", "topics": ["sandbox", "web-navigation"], "updated_at": null, "url": "https://github.com/steel-dev/steel-browser"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "b92abb5fdf714e569cf9f0d2c85ac3fb91ed67365e371a4b322aafe6fab39a1d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/steel-dev/surf.new/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:04Z", "role": "primary", "sha256": "efae611b1201878994c77b110867dcb65fb24a098c398bf5dec6138a47662bbc", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/steel-dev/surf.new/HEAD/LICENSE"}], "id": "github:steel-dev/surf.new", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:04Z", "url": "https://raw.githubusercontent.com/steel-dev/surf.new/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/steel-dev/surf.new"}], "name": "Surf New", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Next.js and FastAPI playground from Steel for trying web agents in a controlled browser environment, with live views of browser interaction and support for providers such as Claude and GPT-4.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/steel-dev/surf.new"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "9d18b958cfefd47100aa32ab14374c22ababb997a637a80da98b1977dec83fe0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/THUDM/AutoWebGLM/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:04Z", "role": "primary", "sha256": "0d97778d2d5e7439d1fbb0ce628565e7bd49886b2e86e6e6ca3d481c7fdc3256", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/THUDM/AutoWebGLM/HEAD/LICENSE"}], "id": "github:thudm/autowebglm", "identifiers": {}, "kind": "agent", "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": "2026-10-16T06:42:04Z", "url": "https://raw.githubusercontent.com/THUDM/AutoWebGLM/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/AutoWebGLM"}], "name": "AutoWebGLM", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Web navigation agent built on ChatGLM3-6B that simplifies HTML for the model, is trained on hybrid human and AI data with RL and rejection sampling, and comes with the bilingual AutoWebBench.", "topics": ["web-navigation"], "updated_at": null, "url": "https://github.com/THUDM/AutoWebGLM"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "eb217b96c54c3c2ccfd3a27255c6fd740df40c2d6743ae811858339bf7f740b0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "42496cf0cff06a7f9e794f4313a10a670bc48df9909ec6d57763cb902bb229ea", "supports": ["identity"], "url": "https://api.github.com/repos/THUDM/WebRL"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "bb1683c6fb5ff658c89f0ebf0f8baf8026bb703840e174c9e46b223c85a16856", "supports": ["release_status"], "url": "https://api.github.com/repos/THUDM/WebRL/git/trees/main?recursive=1"}], "id": "github:thudm/webrl", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:02:06Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/WebRL"}], "name": "WebRL", "notes": "Training code and checkpoint links are released. No root code-license declaration was found in the inspected repository metadata; bundled dependency licenses do not establish a license for the whole project. Weights have separate Hub terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md", "target_id": "arxiv:2411.02337", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Training pipeline that evolves web tasks from prior attempts and uses an outcome reward model to refine browser policies.", "topics": ["reinforcement-learning", "web-navigation", "data-generation"], "updated_at": null, "url": "https://github.com/THUDM/WebRL"}
{"aliases": [], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "release_tag": "v0.39.1"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "4a9783ee8ebee10caaedfe2bdd2182539368e60e8eefadaee4cb7bab7fc75c12", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/vercel-labs/agent-browser/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "6b1120e122d5d110b0da568fbbf0560fd81a4cf8d50602c397a8382617343329", "supports": ["identity"], "url": "https://api.github.com/repos/vercel-labs/agent-browser"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "6a90f0b79089f895fab5d35443fd113faa056f404c807966b23210c1c427e62e", "supports": ["release_status"], "url": "https://api.github.com/repos/vercel-labs/agent-browser/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:55Z", "role": "primary", "sha256": "1a4213f48240031c7837e9651bda8cebb84d69d0d97a65f012082a53a7a1f702", "supports": ["release_status"], "url": "https://api.github.com/repos/vercel-labs/agent-browser/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:56Z", "role": "primary", "sha256": "014bb31e83d5c2e76aea1cc6e82217346ab41362f32cb355ad0f5c10aa0aeaff", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/vercel-labs/agent-browser/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T04:00:56Z", "role": "primary", "sha256": "014bb31e83d5c2e76aea1cc6e82217346ab41362f32cb355ad0f5c10aa0aeaff", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/vercel-labs/agent-browser/HEAD/LICENSE"}], "id": "github:vercel-labs/agent-browser", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T03:59:53Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T04:00:56Z", "url": "https://raw.githubusercontent.com/vercel-labs/agent-browser/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/vercel-labs/agent-browser"}], "name": "agent-browser", "notes": "The repository publishes a Rust-based CLI and platform-specific release binaries. The inspected release is v0.39.1. Apache-2.0 is declared for the code; Chrome and other external browser components retain separate terms. No package or browser was installed.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A browser-automation command-line interface that exposes browser state and interaction commands for AI agents.", "topics": ["web-navigation", "automation", "tool-integration"], "updated_at": null, "url": "https://github.com/vercel-labs/agent-browser"}
{"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": {"availability": "Public WebArena/Mind2Web pipeline, workflow induction and inference source.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/zorazrw/agent-workflow-memory/HEAD/LICENSE"], "reason": "Project source LICENSE, without transferring rights to external benchmark data or model services."}}, "repository_revision": "8c0ff8cd11d648c8fceb99e4e42f37e3b75381b1"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:56Z", "role": "primary", "sha256": "dd2cc4823b2a0b913f44fbaecf9f2cea98257871bdf974de6979459e49cb7b63", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/zorazrw/agent-workflow-memory/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:56Z", "role": "primary", "sha256": "f5c44d559932a22add3e94727b9b413d7c286f835db1396ae0fa1f05e20e773e", "supports": ["identity"], "url": "https://api.github.com/repos/zorazrw/agent-workflow-memory"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:06Z", "role": "primary", "sha256": "602e2238829ad29ef15c75ac7dc61f8bd3dee6a1a74adf0d8ce023067fde6da3", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/zorazrw/agent-workflow-memory/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "1e504b3bd326f7858b39e7f2bedf5ff55d69ffb126e11af364b4184fb980882e", "supports": ["notes", "release_status"], "url": "https://raw.githubusercontent.com/zorazrw/agent-workflow-memory/HEAD/webarena/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:11Z", "role": "primary", "sha256": "e047de93e1c73a39ab6bbf310f02e6023d15c2b06a2b1e80ad4a0875181a9533", "supports": ["notes", "release_status"], "url": "https://raw.githubusercontent.com/zorazrw/agent-workflow-memory/HEAD/mind2web/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/zorazrw/agent-workflow-memory/HEAD/LICENSE"}], "id": "github:zorazrw/agent-workflow-memory", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T12:30:56Z", "licenses": {"code": {"basis": "Project source LICENSE, without transferring rights to external benchmark data or model services.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T12:34:10Z", "url": "https://raw.githubusercontent.com/zorazrw/agent-workflow-memory/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/zorazrw/agent-workflow-memory"}], "name": "Agent Workflow Memory implementation", "notes": "The repository releases separate WebArena and Mind2Web pipelines for online or offline workflow induction and use. WebArena execution requires separately configured browser environments, while Mind2Web workflows require the original dataset and rankings. Publisher performance claims were not reproduced. Apache-2.0 is recorded for this implementation; external benchmark, model and service terms remain separate.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/zorazrw/agent-workflow-memory/HEAD/README.md", "target_id": "arxiv:2409.07429", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Research code that extracts reusable browser-task routines from demonstrations or prior executions, stores them as workflow memory and reuses them during subsequent navigation.", "topics": ["workflow", "memory", "web-navigation"], "updated_at": null, "url": "https://github.com/zorazrw/agent-workflow-memory"}
{"aliases": [], "authors": [], "details": {"access": "publisher-managed", "availability": "not independently tested", "pricing_url": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:54Z", "role": "primary", "sha256": "bd3e99c1e12588bc3136dfa27de2c0d5790b6f6f25f2696988eb182c6b1ef670", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://claude.com/chrome"}], "id": "url:07b6bbce7ce26a847f79", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-16T06:40:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://claude.com/chrome"}], "name": "Claude for Chrome", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://claude.com/chrome", "target_id": "organization:anthropic", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Claude browser extension for interacting with websites.", "topics": ["web-navigation"], "updated_at": null, "url": "https://claude.com/chrome"}
{"aliases": [], "authors": [], "details": {"access": "publisher-managed", "availability": "not independently tested", "pricing_url": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:54Z", "role": "primary", "sha256": "fa71ad6addd026ecd017059219c2457baf6b99c5c94f67350c0f9637c459c852", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://openai.com/index/introducing-chatgpt-atlas/"}], "id": "url:240a05c02f85d4da60dc", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-16T06:40:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://openai.com/index/introducing-chatgpt-atlas/"}], "name": "ChatGPT Atlas", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://openai.com/index/introducing-chatgpt-atlas/", "target_id": "organization:openai", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "OpenAI browser product integrating ChatGPT and agent capabilities.", "topics": ["web-navigation"], "updated_at": null, "url": "https://openai.com/index/introducing-chatgpt-atlas/"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-wa/tree/main", "gated": false, "hub_id": "stanfordnlp/nnetnav-wa", "languages": [], "modalities": ["text"], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:11:53Z", "role": "primary", "sha256": "3fe40a45aa2bf9d6f0cefffc65ea6382849c3543b44017200a4a8d43ab61e9eb", "supports": ["name", "kind"], "url": "https://huggingface.co/api/datasets/stanfordnlp/nnetnav-wa"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:34Z", "role": "primary", "sha256": "3095f7262b7d4ff916c6f58c55e6fc9902c08f77d09bf3951199945501148bb3", "supports": ["name", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/MurtyShikhar/NNetnav/HEAD/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T11:11:53Z", "role": "primary", "sha256": "3fe40a45aa2bf9d6f0cefffc65ea6382849c3543b44017200a4a8d43ab61e9eb", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/stanfordnlp/nnetnav-wa"}], "id": "url:25a4d0ab29b60250863a", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-16T11:11:53Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "apache-2.0", "retrieved_at": "2026-10-16T11:11:53Z", "url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-wa"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-wa"}], "name": "NNetNav WebArena data", "notes": "The Hub card supplies a license but no detailed schema. Dataset purpose is supported by the project's README; exact sample counts and annotation quality remain unverified.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/MurtyShikhar/NNetnav/HEAD/README.md", "target_id": "github:murtyshikhar/nnetnav", "type": "component_of"}], "review_status": "curated", "scope": "core", "summary": "NNetNav's WebArena-oriented browser-interaction data release, with separate training and test files.", "topics": ["data-generation", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-wa"}
{"aliases": [], "authors": ["Shaon Barman", "Sarah Chasins", "Ras Bodik", "Sumit Gulwani"], "details": {"publication": {"authors": ["Shaon Barman", "Sarah Chasins", "Ras Bodik", "Sumit Gulwani"], "date_basis": "Supplied month precision: 2016-11. year=2016, month=November", "date_parts": [2016, 11], "doi": "10.1145/2983990.2984020", "event_dates": null, "evidence_urls": ["https://www.microsoft.com/en-us/research/publication/ringer-web-automation-demonstration/bibtex/", "https://2016.splashcon.org/track/splash-2016-oopsla", "https://www.microsoft.com/en-us/research/publication/ringer-web-automation-demonstration/", "https://www.microsoft.com/en-us/research/wp-content/uploads/2016/12/oopsla16-web.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": "The original manuscript and MSR BibTeX print Sarah Chasins and Ras Bodik. The official conference profiles expand these to Sarah E. Chasins and Rastislav Bodík; the printed publication names are preserved. Page range, volume and exact publication day are not established by the retained citation sources.", "pages": null, "pdf_url": "https://www.microsoft.com/en-us/research/wp-content/uploads/2016/12/oopsla16-web.pdf", "publisher": null, "series": null, "title": "Ringer: Web Automation by Demonstration", "type": "conference-paper", "url": "https://doi.org/10.1145/2983990.2984020", "venue": "OOPSLA 2016", "volume": null}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:57:36Z", "role": "primary", "sha256": "633aa15e2e14346241de5385cc19d99f1fcdc9ef79fd06d27176dbfa8f38149a", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.type", "details.publication.venue"], "url": "https://www.microsoft.com/en-us/research/publication/ringer-web-automation-demonstration/bibtex/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:57:36Z", "role": "primary", "sha256": "94c6b6bbd127126c9dd259bd346a7757d75366b4d805bdee2959c4c73226df74", "supports": ["details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.url"], "url": "https://2016.splashcon.org/track/splash-2016-oopsla"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:24Z", "role": "primary", "sha256": "45f65483c3801dac142f4657c4cbe45e17b6a3e081497731e984c1e670c6e8f6", "supports": ["details.publication.identity_basis", "details.publication.notes", "details.publication.title"], "url": "https://www.microsoft.com/en-us/research/publication/ringer-web-automation-demonstration/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:55:00Z", "role": "primary", "sha256": "bfbb69f57aaed39e3a210fae832e1fd98ed62ba1c01217b792238fe57e1a0ad1", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.pdf_url"], "url": "https://www.microsoft.com/en-us/research/wp-content/uploads/2016/12/oopsla16-web.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:24Z", "role": "primary", "sha256": "45f65483c3801dac142f4657c4cbe45e17b6a3e081497731e984c1e670c6e8f6", "supports": ["details.publication.identity_basis", "details.publication.notes", "details.publication.title", "environments", "name", "scope", "topics"], "url": "https://www.microsoft.com/en-us/research/publication/ringer-web-automation-demonstration/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:57:36Z", "role": "primary", "sha256": "633aa15e2e14346241de5385cc19d99f1fcdc9ef79fd06d27176dbfa8f38149a", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.type", "details.publication.venue", "kind"], "url": "https://www.microsoft.com/en-us/research/publication/ringer-web-automation-demonstration/bibtex/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:57:36Z", "role": "primary", "sha256": "94c6b6bbd127126c9dd259bd346a7757d75366b4d805bdee2959c4c73226df74", "supports": ["details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.url"], "url": "https://2016.splashcon.org/track/splash-2016-oopsla"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:55:00Z", "role": "primary", "sha256": "bfbb69f57aaed39e3a210fae832e1fd98ed62ba1c01217b792238fe57e1a0ad1", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.pdf_url", "summary"], "url": "https://www.microsoft.com/en-us/research/wp-content/uploads/2016/12/oopsla16-web.pdf"}], "id": "url:347925d0f440d6396080", "identifiers": {"doi": "10.1145/2983990.2984020"}, "kind": "paper", "last_verified_at": "2026-10-17T09:53:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://doi.org/10.1145/2983990.2984020"}, {"rel": "publication_pdf", "url": "https://www.microsoft.com/en-us/research/wp-content/uploads/2016/12/oopsla16-web.pdf"}], "name": "Ringer: Web Automation by Demonstration", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. The original manuscript and MSR BibTeX print Sarah Chasins and Ras Bodik. The official conference profiles expand these to Sarah E. Chasins and Rastislav Bodík; the printed publication names are preserved.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Ringer, which converts demonstrated webpage interactions into scripts combining actions, target selectors, and waiting conditions. It evaluates replay robustness on tasks recorded from real websites.", "topics": ["web-navigation", "recording", "robustness"], "updated_at": null, "url": "https://doi.org/10.1145/2983990.2984020"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/tree/main", "gated": false, "hub_id": "xTRam1/plan-and-act-planner-70b", "license_review": {"weights": {"evidence_urls": ["https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md"], "reason": "The model card declares MIT in its metadata header (license: mit), beside a template comment asking the publisher to update it if needed. It is recorded as the publisher's declaration and does not override the Llama-3.3-70B-Instruct base model terms."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "4ef2d84d2b5d9d2972907e6135ea49f1ef088beaf6ef6042dc7d8fb4297a9a3a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.weights"], "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "d2f757790ded81244e7c9b9e7350ec9fdc304609e387a95c7918db705bda535c", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/xTRam1/plan-and-act-planner-70b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "feee31aa4d0801cb364578099a13e4f9cdcaa4eb5fbf4312b3fc3f8edd479a14", "supports": ["environments"], "url": "https://raw.githubusercontent.com/SqueezeAILab/plan-and-act/HEAD/README.md"}], "id": "url:3932d69c6b6f9b2a7ba8", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:02:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:02:28Z", "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b"}], "name": "Plan-and-Act Planner 70B", "notes": "The Hub manifest contains 17 safetensors shards and tokenizer/configuration files. Card metadata declares MIT and Llama-3.3-70B-Instruct as the base, but retains template comments asking the publisher to update those fields if needed. These are recorded as publisher declarations, without claiming they override upstream model terms.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md", "target_id": "arxiv:2503.09572", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "The planner component of Plan-and-Act, generating structured high-level plans that an executor turns into web actions.", "topics": ["planning", "web-navigation", "long-horizon"], "updated_at": null, "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b"}
{"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": {"access": "publisher-managed", "availability": "not independently tested", "pricing_url": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "7d67747fdfe89bcc69d2b60b41f4f2dccf758913ea6d491949cbeac20de372ef", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://hcompany.ai/newsroom/meet-holotab"}], "id": "url:4a47a6df845735984bf1", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-16T06:38:27Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://hcompany.ai/newsroom/meet-holotab"}], "name": "HoloTab", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://hcompany.ai/newsroom/meet-holotab", "target_id": "organization:h", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "H browser extension for executing tasks through website interfaces.", "topics": ["web-navigation"], "updated_at": null, "url": "https://hcompany.ai/newsroom/meet-holotab"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-live/tree/main", "gated": false, "hub_id": "stanfordnlp/nnetnav-live", "languages": [], "modalities": ["text"], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:11:53Z", "role": "primary", "sha256": "d8f8571dfc583d3a46d5b3a6a285e77f17057b36a8111d416d711513aa4ecf16", "supports": ["name", "kind"], "url": "https://huggingface.co/api/datasets/stanfordnlp/nnetnav-live"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:34Z", "role": "primary", "sha256": "3095f7262b7d4ff916c6f58c55e6fc9902c08f77d09bf3951199945501148bb3", "supports": ["name", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/MurtyShikhar/NNetnav/HEAD/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T11:11:53Z", "role": "primary", "sha256": "d8f8571dfc583d3a46d5b3a6a285e77f17057b36a8111d416d711513aa4ecf16", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/stanfordnlp/nnetnav-live"}], "id": "url:4f47abde8cd28a1ec1e1", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-16T11:11:53Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "apache-2.0", "retrieved_at": "2026-10-16T11:11:53Z", "url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-live"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-live"}], "name": "NNetNav Live data", "notes": "The Hub card supplies a license but no detailed schema. Dataset purpose is supported by the project's README; exact sample counts and annotation quality remain unverified.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/MurtyShikhar/NNetnav/HEAD/README.md", "target_id": "github:murtyshikhar/nnetnav", "type": "component_of"}], "review_status": "curated", "scope": "core", "summary": "Browser-interaction data released with NNetNav for its live-website setting, with separate training and test files.", "topics": ["data-generation", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/datasets/stanfordnlp/nnetnav-live"}
{"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": {"license_review": {"document": {"evidence_urls": ["https://raw.githubusercontent.com/huggingface/smolagents/12c1bc820eca50ace6f80a21d90426d41d74f845/docs/source/en/examples/web_browser.md", "https://raw.githubusercontent.com/huggingface/smolagents/12c1bc820eca50ace6f80a21d90426d41d74f845/LICENSE"], "reason": "The retained repository-root file contains Apache-2.0 terms, but the reviewed Markdown has no license notice and the retained sources do not explicitly establish that this license applies to the tutorial document. Document reuse terms remain unknown; this is not a claim that no license exists. Code, data, weights and hosted-service terms are not inferred."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:42:59.429802+00:00", "role": "primary", "sha256": "0bd3804794c37f263db5cba3429523721854e95fe1477127f9fb83d4d0510f50", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://raw.githubusercontent.com/huggingface/smolagents/12c1bc820eca50ace6f80a21d90426d41d74f845/docs/source/en/examples/web_browser.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:41:14.002361+00:00", "role": "primary", "sha256": "736281d4d96fc310050a09e830801f739729cc01217fef0be1fd45874e4c68f6", "supports": ["identity", "name", "summary", "environments", "topics", "scope", "notes"], "url": "https://huggingface.co/docs/smolagents/v1.26.0/examples/web_browser"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:41:17.450466+00:00", "role": "primary", "sha256": "cf08eff3ce4ca2fd85f85ba1eebca477277bca0cb2a54613e172741392061365", "supports": ["notes"], "url": "https://api.github.com/repos/huggingface/smolagents/git/ref/tags/v1.26.0"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T12:51:48.091044+00:00", "role": "primary", "sha256": "9def26af20ef4a6c4091e5873b53737940b360e6ff7db2d606add92d6586c8fb", "supports": ["notes"], "url": "https://huggingface.co/docs/smolagents/v1.8.0/examples/web_browser"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:43:02.729855+00:00", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.document"], "url": "https://raw.githubusercontent.com/huggingface/smolagents/12c1bc820eca50ace6f80a21d90426d41d74f845/LICENSE"}], "id": "url:5857c840372512cfe3fc", "identifiers": {}, "kind": "learning", "last_verified_at": "2026-10-20T13:42:59.429802+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/docs/smolagents/examples/web_browser"}], "name": "Web Browser Automation with Agents (smolagents tutorial)", "notes": "The reviewed v1.26.0 page matches all eight code blocks in its commit-pinned Markdown source. It uses InferenceClientModel and a vision model, unlike the retained v1.8.0 example. The shown setup starts a local visible Chrome session, and its installation command does not pin dependency versions. Runtime API compatibility, model-provider access and example outcomes were not tested. The inspected repository root contains Apache-2.0 terms, but explicit applicability to this Markdown tutorial was not established from retained sources; document reuse terms remain unknown. Model weights, visited websites and hosted-service terms are separate. No installation or notebook execution occurred.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Versioned tutorial building a vision-guided CodeAgent with Selenium/Helium browser tools, screenshot callbacks and example web-navigation tasks.", "topics": ["web-navigation", "screen-understanding"], "updated_at": null, "url": "https://huggingface.co/docs/smolagents/examples/web_browser"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/yifeizhou/pae-llava-7b/tree/main", "gated": false, "hub_id": "yifeizhou/pae-llava-7b"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "c0994582f81415c9b0d37e4e5018752e48e511178c4b3898c207a042992db845", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "1b76d2aebbe15954eff2b8859b884ec6504f8b67ed0b1045d8b3a83c0d132d3a", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/yifeizhou/pae-llava-7b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "3e5f24b3175afeba8635dacc7b66eba1ce89fea49996a20281ff2df5a41f1e24", "supports": ["environments", "summary"], "url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "c0994582f81415c9b0d37e4e5018752e48e511178c4b3898c207a042992db845", "supports": ["licenses.weights"], "url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md"}], "id": "url:5d61c1b5fc8dd650518d", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:02:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "cc-by-nc-4.0", "retrieved_at": "2026-10-17T02:02:28Z", "url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/yifeizhou/pae-llava-7b"}], "name": "PAE LLaVA 7B", "notes": "The model is publicly ungated and has seven safetensors shards plus tokenizer and configuration files. Its card declares CC-BY-NC-4.0 for this model. This weight declaration does not settle the separate PAE dataset or code licensing questions.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md", "target_id": "arxiv:2412.13194", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A released LLaVA checkpoint for visual browser interaction from the PAE autonomous task-proposal and policy-learning system.", "topics": ["web-navigation", "reinforcement-learning"], "updated_at": null, "url": "https://huggingface.co/yifeizhou/pae-llava-7b"}
{"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": ["James Allen", "Nathanael Chambers", "George Ferguson", "Lucian Galescu", "Hyuckchul Jung", "Mary Swift", "William Taysom"], "details": {"publication": {"authors": ["James Allen", "Nathanael Chambers", "George Ferguson", "Lucian Galescu", "Hyuckchul Jung", "Mary Swift", "William Taysom"], "date_basis": "Supplied year precision: 2007. Papers section: AAAI-2007", "date_parts": [2007], "doi": null, "event_dates": null, "evidence_urls": ["https://aaai.org/Papers/AAAI/2007/AAAI07-240.pdf", "https://www.cs.rochester.edu/research/cisd/projects/plow/"], "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": "The ordered names follow the original published paper, without expanding initials or replacing names from secondary indexes. DOI and volume number are not supplied in the retained primary sources. No exact publication month/day is established.", "pages": "1514-1519", "pdf_url": "https://cdn.aaai.org/AAAI/2007/AAAI07-240.pdf", "publisher": null, "series": null, "title": "PLOW: A Collaborative Task Learning Agent", "type": "conference-paper", "url": "https://aaai.org/Papers/AAAI/2007/AAAI07-240.pdf", "venue": "Twenty-Second AAAI National Conference on Artificial Intelligence", "volume": null}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:26Z", "role": "primary", "sha256": "795d275c9f85180561fd101652221d30fb98daca2c2a00faf2b57c397bdfd68d", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.url"], "url": "https://aaai.org/Papers/AAAI/2007/AAAI07-240.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:24Z", "role": "primary", "sha256": "c0b3571efd5f2e137e42b38107de29aca6668819bec84109912d514867d85b8c", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.notes", "details.publication.type", "details.publication.venue"], "url": "https://www.cs.rochester.edu/research/cisd/projects/plow/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:26Z", "role": "primary", "sha256": "795d275c9f85180561fd101652221d30fb98daca2c2a00faf2b57c397bdfd68d", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.url", "environments", "name", "scope", "summary", "topics"], "url": "https://aaai.org/Papers/AAAI/2007/AAAI07-240.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:24Z", "role": "primary", "sha256": "c0b3571efd5f2e137e42b38107de29aca6668819bec84109912d514867d85b8c", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.notes", "details.publication.type", "details.publication.venue", "environments", "kind"], "url": "https://www.cs.rochester.edu/research/cisd/projects/plow/"}], "id": "url:628b5251757bb3888e89", "identifiers": {}, "kind": "paper", "last_verified_at": "2026-10-17T09:53:26Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://aaai.org/Papers/AAAI/2007/AAAI07-240.pdf"}, {"rel": "publication_pdf", "url": "https://cdn.aaai.org/AAAI/2007/AAAI07-240.pdf"}], "name": "PLOW: A Collaborative Task Learning Agent", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. The ordered names follow the original published paper, without expanding initials or replacing names from secondary indexes.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes PLOW, which learns executable web-task models from a session combining demonstrations, natural-language explanations, and dialogue. Browser instrumentation connects user actions with DOM structure to support task generalization.", "topics": ["web-navigation", "human-in-the-loop", "planning"], "updated_at": null, "url": "https://aaai.org/Papers/AAAI/2007/AAAI07-240.pdf"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/LangAGI-Lab/WebShepherd_3B/tree/main", "gated": false, "hub_id": "LangAGI-Lab/WebShepherd_3B", "languages": [], "modalities": [], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "6f28e86a99b286a575c5e7f8d24d6901655971e9a091493950b59948f4b9ad9b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/api/models/LangAGI-Lab/WebShepherd_3B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:45:05Z", "role": "primary", "sha256": "647d9e7bdad5fe2c03a8cfcd34d93f3b2bf7c1553b517a4a843bc0113fa47456", "supports": ["identity", "summary", "relations"], "url": "https://huggingface.co/api/collections/LangAGI-Lab/web-shepherd-advancing-prms-for-reinforcing-web-agents-682b4f4ad607fc27c4dc49e8"}, {"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"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:49:45Z", "role": "primary", "sha256": "252ed290ec84ffd8fa686fb72df68afe7bbd8c19b80538c0b318db60cd107d09", "supports": ["details"], "url": "https://huggingface.co/LangAGI-Lab/WebShepherd_3B/raw/main/config.json"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T02:45:20Z", "role": "primary", "sha256": "6f28e86a99b286a575c5e7f8d24d6901655971e9a091493950b59948f4b9ad9b", "supports": ["details", "licenses.weights"], "url": "https://huggingface.co/api/models/LangAGI-Lab/WebShepherd_3B"}], "id": "url:72f7fe0211c6551b945f", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:45:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/LangAGI-Lab/WebShepherd_3B"}], "name": "WebShepherd 3B", "notes": "The official project-linked collection identifies this public 3B checkpoint. The README uses a separate unavailable WebShepherd/web-shepherd-base URL; redirect equivalence is not asserted. Two safetensors shards and a Qwen2 causal-language-model configuration are present. No model card or weight license is published in the inspected manifest. The separately announced multimodal 3B model is not assumed released.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/api/collections/LangAGI-Lab/web-shepherd-advancing-prms-for-reinforcing-web-agents-682b4f4ad607fc27c4dc49e8", "target_id": "arxiv:2505.15277", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A public text-based process-reward checkpoint from the Web-Shepherd collection for scoring web-agent decisions against task checklists.", "topics": ["reward-models", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/LangAGI-Lab/WebShepherd_3B"}
{"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": [], "authors": [], "details": {"architecture": "LlamaForCausalLM", "availability": "Ungated complete weight/config/tokenizer manifest.", "license_review": {"weights": {"evidence_urls": ["https://huggingface.co/McGill-NLP/Llama-3-8B-Web/raw/main/META_LLAMA_3_COMMUNITY_LICENSE_AGREEMENT.txt", "https://huggingface.co/McGill-NLP/Llama-3-8B-Web/raw/main/README.md"], "reason": "Weight license text and card declaration are retained separately from framework code licensing."}}, "parameter_label": "8B", "release_revision": "09856c28ec324cf3fe4c513bdc4437a75b78185f", "training_data_note": "Publisher describes a 24K curated subset of WebLINX interactions and dialogue; not a newly counted independent dataset."}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "1a445c7894c65c9a7c4247dba446c206c13ac2ecdc129ce93a0de943f2984a18", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "licenses.weights"], "url": "https://huggingface.co/McGill-NLP/Llama-3-8B-Web/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "69fe17825db0d5f32ccaa84b67fabfa10018e1410da0c70ec0ea9509703feba2", "supports": ["identity", "release_status", "details.release_revision", "details.architecture"], "url": "https://huggingface.co/api/models/McGill-NLP/Llama-3-8B-Web"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:36:55Z", "role": "primary", "sha256": "6bcca341999ab474e2973116128177b77f97cd64028eeace224035358ce6474f", "supports": ["licenses.weights"], "url": "https://huggingface.co/McGill-NLP/Llama-3-8B-Web/raw/main/META_LLAMA_3_COMMUNITY_LICENSE_AGREEMENT.txt"}], "id": "url:7c3e8c4ba52428584079", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T12:34:10Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "Weight license text and card declaration are retained separately from framework code licensing.", "declared": "Meta Llama 3 Community License Agreement", "retrieved_at": "2026-10-17T12:36:55Z", "url": "https://huggingface.co/McGill-NLP/Llama-3-8B-Web/raw/main/META_LLAMA_3_COMMUNITY_LICENSE_AGREEMENT.txt"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/McGill-NLP/Llama-3-8B-Web"}], "name": "Llama-3-8B-Web", "notes": "The ungated Hub manifest includes model.safetensors, pytorch_model.bin, config and tokenizer files; the two weight formats represent the same checkpoint and are not counted separately. The card describes an approximately 24K-example curated WebLINX training subset and evaluates dialogue-guided web navigation. Predictions still require parsing and execution in an external browser environment. The retained weight license is the Meta Llama 3 Community License Agreement, not WebLlama code MIT. No weights were downloaded, model used, or agreement accepted.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/McGill-NLP/Llama-3-8B-Web/raw/main/README.md", "target_id": "github:mcgill-nlp/weblinx", "type": "trained_on"}, {"evidence_url": "https://huggingface.co/McGill-NLP/Llama-3-8B-Web/raw/main/README.md", "target_id": "arxiv:2402.05930", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A Llama 3 8B instruction-model checkpoint fine-tuned on WebLINX web interactions and dialogue, predicting browser actions from a textual state representation.", "topics": ["web-navigation", "imitation-learning", "action-prediction"], "updated_at": null, "url": "https://huggingface.co/McGill-NLP/Llama-3-8B-Web"}
{"aliases": [], "authors": [], "details": {"access": "publisher-managed", "availability": "not independently tested", "pricing_url": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:28Z", "role": "primary", "sha256": "41ec857dacdb106ad99abbe1872f40a295056522ffaa1b1714b51dffb370f450", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://www.agentql.com/"}], "id": "url:8896d44370711136fd54", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-16T06:38:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://www.agentql.com/"}], "name": "AgentQL", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Natural-language querying and interaction tools for web interfaces.", "topics": ["grounding", "web-navigation"], "updated_at": null, "url": "https://www.agentql.com/"}
{"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": ["Sarah E. Chasins", "Maria Mueller", "Rastislav Bodik"], "details": {"publication": {"authors": ["Sarah E. Chasins", "Maria Mueller", "Rastislav Bodik"], "date_basis": "Supplied year precision: 2018. Rousillon entry: UIST 2018", "date_parts": [2018], "doi": "10.1145/3242587.3242661", "event_dates": null, "evidence_urls": ["https://schasins.com/papers/", "https://schasins.com/assets/papers/rousillon.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": "This is browser interaction learned from demonstrations, rather than a generic server-side crawling library. Final proceedings page range and volume are not established. Manuscript pages 1-13 are not substituted for publisher pagination. Exact publication month/day is not inferred from conference dates.", "pages": null, "pdf_url": "https://schasins.com/assets/papers/rousillon.pdf", "publisher": null, "series": null, "title": "Rousillon: Scraping Distributed Hierarchical Web Data", "type": "conference-paper", "url": "https://doi.org/10.1145/3242587.3242661", "venue": "Proceedings of the 31st Annual ACM Symposium on User Interface Software and Technology", "volume": null}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:57:38Z", "role": "primary", "sha256": "6b7e584bb3862e0de51ca164d0f311c2a853e97bd63d56658ce62f0b49d86557", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.notes", "details.publication.type", "details.publication.venue"], "url": "https://schasins.com/papers/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:55:12Z", "role": "primary", "sha256": "764dd1f5fe1db398188f403b4b1f6e3658e3238e4d8aac9ceed360dcad87e818", "supports": ["details.publication.authors", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pdf_url", "details.publication.title", "details.publication.url"], "url": "https://schasins.com/assets/papers/rousillon.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:55:12Z", "role": "primary", "sha256": "764dd1f5fe1db398188f403b4b1f6e3658e3238e4d8aac9ceed360dcad87e818", "supports": ["details.publication.authors", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pdf_url", "details.publication.title", "details.publication.url", "environments", "name", "scope", "summary", "topics"], "url": "https://schasins.com/assets/papers/rousillon.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:57:38Z", "role": "primary", "sha256": "6b7e584bb3862e0de51ca164d0f311c2a853e97bd63d56658ce62f0b49d86557", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.notes", "details.publication.type", "details.publication.venue", "kind"], "url": "https://schasins.com/papers/"}], "id": "url:97c0b1896525632434e2", "identifiers": {"doi": "10.1145/3242587.3242661"}, "kind": "paper", "last_verified_at": "2026-10-17T09:55:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://doi.org/10.1145/3242587.3242661"}, {"rel": "publication_pdf", "url": "https://schasins.com/assets/papers/rousillon.pdf"}], "name": "Rousillon: Scraping Distributed Hierarchical Web Data", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. This is browser interaction learned from demonstrations, rather than a generic server-side crawling library.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Rousillon, which generalizes a demonstration of collecting one hierarchical data row into a browser-automation program that traverses related pages. The paper develops relation-selection and generalization methods and evaluates users building such scripts.", "topics": ["web-navigation", "recording", "automation"], "updated_at": null, "url": "https://doi.org/10.1145/3242587.3242661"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/apurvaga/go-browse-wa/raw/main/README.md", "population": "train split", "qualifier": "exact", "unit": "step records", "value": 164533}], "evidence_urls": ["https://huggingface.co/datasets/apurvaga/go-browse-wa/raw/main/README.md", "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/README.md"], "formats": ["Parquet"], "modalities": ["text", "accessibility-tree observations", "actions"], "notes": "The card count is flattened step records, not trajectories or independent tasks. Both successful and unsuccessful trajectories are included according to the project README; no license is declared.", "splits": ["train"], "supervision": ["action completion", "trajectory outcome"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/apurvaga/go-browse-wa/tree/main", "gated": false, "hub_id": "apurvaga/go-browse-wa", "languages": [], "modalities": ["tabular", "text"], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "494f4ac9ab9e94831dcf768be001b7284e30e00704674358baf3bb2060344361", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/api/datasets/apurvaga/go-browse-wa"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "18ab64833acf31113a83b2463359f1c7e37c4aaabfe11c33db2ba3e7d32818c6", "supports": ["details"], "url": "https://huggingface.co/datasets/apurvaga/go-browse-wa/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "0d4b28c993ccb977c7910ff3fecb5de29db364a7a7ad55523c8e052c9bae2566", "supports": ["summary", "environments", "relations"], "url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "494f4ac9ab9e94831dcf768be001b7284e30e00704674358baf3bb2060344361", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/apurvaga/go-browse-wa"}], "id": "url:a13a187bc058fbb4f6af", "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/apurvaga/go-browse-wa"}], "name": "Go-Browse-WA processed data", "notes": "The public release contains nine Parquet shards. The short card supplies schema and split metadata but no license declaration; the official README describes successful and unsuccessful trajectories processed for fine-tuning. The raw screenshot/HTML corpus is a separate release and is not claimed to be included here.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/README.md", "target_id": "arxiv:2506.03533", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Flattened web-interaction training records with accessibility-tree observations, assistant completions, trajectory rewards and step-link metadata.", "topics": ["web-navigation", "imitation-learning"], "updated_at": null, "url": "https://huggingface.co/datasets/apurvaga/go-browse-wa"}
{"aliases": ["url:4338ec0f1a67bdb6a139", "microsoft/Fara-7B", "https://huggingface.co/microsoft/Fara-7B", "url:d416331cdf9730afd6b9", "microsoft/fara-7b", "https://huggingface.co/microsoft/fara-7b"], "authors": [], "details": {"artifact_type": "model-checkpoint", "disabled": false, "download_url": "https://huggingface.co/microsoft/Fara-7b/tree/main", "gated": false, "hub_id": "microsoft/Fara-7B", "languages": ["en"], "modalities": [], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "92bfa3c1f34e7e76a35f72389a3b3dd9fdab5fcff0569c9fc8aabe331c703f07", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/api/models/microsoft/Fara-7b"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "92bfa3c1f34e7e76a35f72389a3b3dd9fdab5fcff0569c9fc8aabe331c703f07", "supports": ["details", "licenses.weights"], "url": "https://huggingface.co/api/models/microsoft/Fara-7b"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:14:59Z", "role": "primary", "sha256": "cd1f83cd2ebf30d45501d1a31ecbe3bcb2039928d580c750733d2edb49916398", "supports": ["aliases"], "url": "https://huggingface.co/api/models/microsoft/Fara-7B"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:00Z", "role": "primary", "sha256": "cd1f83cd2ebf30d45501d1a31ecbe3bcb2039928d580c750733d2edb49916398", "supports": ["aliases"], "url": "https://huggingface.co/api/models/microsoft/Fara-7b"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:01Z", "role": "primary", "sha256": "cd1f83cd2ebf30d45501d1a31ecbe3bcb2039928d580c750733d2edb49916398", "supports": ["aliases"], "url": "https://huggingface.co/api/models/microsoft/fara-7b"}], "id": "url:a4da77ee63f879cc8e63", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:50:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card", "declared": "mit", "retrieved_at": "2026-10-16T06:50:49Z", "url": "https://huggingface.co/microsoft/Fara-7b"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/microsoft/Fara-7b"}], "name": "Fara 7B", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/api/models/microsoft/Fara-7b", "target_id": "github:microsoft/fara", "type": "variant_of"}], "review_status": "curated", "scope": "core", "summary": "Microsoft's Fara checkpoint, a Qwen2.5-VL-architecture image-text-to-text model with about 8.3B BF16 parameters, English, MIT-licensed and linked to arXiv paper 2511.19663.", "topics": ["planning", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/microsoft/Fara-7b"}
{"aliases": ["url:2f67f5ad7954632dfc4e", "THUDM/webrl-llama-3.1-8b", "https://huggingface.co/THUDM/webrl-llama-3.1-8b"], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/tree/main", "gated": false, "hub_id": "zai-org/webrl-llama-3.1-8b"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "4712446382f106a115a317d0ec6273a975545386b9cedf56946a96a56479dc9f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "6523b7f198d8debec2f3074190812955c6251590b61f8a6ea1bbb681b8ea128f", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/zai-org/webrl-llama-3.1-8b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "6523b7f198d8debec2f3074190812955c6251590b61f8a6ea1bbb681b8ea128f", "supports": ["identity"], "url": "https://huggingface.co/api/models/THUDM/webrl-llama-3.1-8b"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "4712446382f106a115a317d0ec6273a975545386b9cedf56946a96a56479dc9f", "supports": ["licenses.weights"], "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:06Z", "role": "primary", "sha256": "96621f060dddec91453d8c2b451b3793a34b34810d82c7cd13ce4cb05ca90a76", "supports": ["aliases"], "url": "https://huggingface.co/api/models/THUDM/webrl-llama-3.1-8b"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:06Z", "role": "primary", "sha256": "96621f060dddec91453d8c2b451b3793a34b34810d82c7cd13ce4cb05ca90a76", "supports": ["aliases"], "url": "https://huggingface.co/api/models/zai-org/webrl-llama-3.1-8b"}], "id": "url:a6ee20fea67438d91e06", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:05:40Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "other", "retrieved_at": "2026-10-17T02:05:40Z", "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b"}], "name": "WebRL Llama 3.1 8B", "notes": "The official Hub API resolves the former THUDM namespace to zai-org. Seven safetensors shards and configuration/tokenizer files are public and ungated. The card marks the weight license as other without naming complete terms, so no permissive weight license is inferred. The card’s evaluation prose also names GLM-4 despite the Llama checkpoint identity; no score is imported.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md", "target_id": "arxiv:2411.02337", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A WebRL browser-policy checkpoint for performing actions in the five website environments of WebArena.", "topics": ["web-navigation", "reinforcement-learning"], "updated_at": null, "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T12:51:44.325799+00:00", "role": "primary", "sha256": "1f0315e8758fe0edd97adb2c5253955dcdbe6d7bd8be393fe14dcd093483705f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://www.deeplearning.ai/courses/building-ai-browser-agents"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:41:10.403317+00:00", "role": "primary", "sha256": "0915d862638a308da6200d9603d668ab4d11bcabfa713ff0ccb82c7b21a5a92f", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://learn.deeplearning.ai/courses/building-ai-browser-agents/lesson/wwewm/building-a-simple-web-agent"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:42:56.029732+00:00", "role": "primary", "sha256": "20d17ca47d1d16826f8e2736d6639d0cf11e826ef28c788187e16adcac9d7561", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://learn.deeplearning.ai/courses/building-ai-browser-agents/lesson/lot5j/building-an-autonomous-web-agent"}], "id": "url:aa693ca3e8948623f31e", "identifiers": {}, "kind": "learning", "last_verified_at": "2026-10-20T12:51:44.325799+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://www.deeplearning.ai/courses/building-ai-browser-agents"}], "name": "Building AI Browser Agents", "notes": "Review covers the public syllabus and complete published transcripts of the simple-agent and autonomous-agent lessons, not the whole course or its video frames. The first uses Playwright and structured HTML extraction; the second uses a MultiOn session API and shows missed steps, continuations and clarifications. Hosted notebooks and helper files were not retrieved, and current service access was not tested. No account, enrollment, terms acceptance, installation or execution occurred. Course and example-code reuse terms remain unknown.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Browser-agent course with worked lessons on structured web extraction and multi-step browser interaction, followed by AgentQ search and self-correction concepts.", "topics": ["web-navigation", "planning"], "updated_at": null, "url": "https://www.deeplearning.ai/courses/building-ai-browser-agents"}
{"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": {"dataset_profile": {"counts": [], "evidence_urls": ["https://huggingface.co/datasets/xTRam1/plan-and-act-data/raw/main/README.md", "https://huggingface.co/api/datasets/xTRam1/plan-and-act-data"], "formats": ["JSON", "JSONL"], "modalities": ["text", "plans", "actions"], "notes": "The release contains planner and actor files for two web environments. No sample count is inferred from the package-file manifest. Card loading examples have repository/path inconsistencies.", "splits": ["train (QwQ-32B files)", "validation (QwQ-32B files)"], "supervision": ["high-level plans", "action execution"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/xTRam1/plan-and-act-data/tree/main", "gated": false, "hub_id": "xTRam1/plan-and-act-data"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:29Z", "role": "primary", "sha256": "2ec8f4a03b8331bdd2ba8d6f6800d6dea9d955cfd9e7c1d2168ecb64deb25448", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/xTRam1/plan-and-act-data/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "a98f98b7d4ceb81b00f217983ff271257f2f1e7d2c1a09836b872b2350d79ae9", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/xTRam1/plan-and-act-data"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:02:29Z", "role": "primary", "sha256": "2ec8f4a03b8331bdd2ba8d6f6800d6dea9d955cfd9e7c1d2168ecb64deb25448", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/xTRam1/plan-and-act-data/raw/main/README.md"}], "id": "url:ba6453cfdae0eb049ced", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:02:29Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:02:29Z", "url": "https://huggingface.co/datasets/xTRam1/plan-and-act-data/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/xTRam1/plan-and-act-data"}], "name": "Plan-and-Act training data", "notes": "The public manifest exposes JSON/JSONL data for planner and actor roles, including train and validation variants. The card declares MIT. Its loading examples use a different Hub repository identifier and omit the data/ prefix for one file; use the inspected xTRam1 manifest rather than copying those examples verbatim. Third-party site content retains its own terms.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/xTRam1/plan-and-act-data/raw/main/README.md", "target_id": "arxiv:2503.09572", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Planner and action-executor training examples for WebArena-Lite and WebVoyager, with model-specific processed files and raw trajectory annotations.", "topics": ["planning", "web-navigation", "data-generation"], "updated_at": null, "url": "https://huggingface.co/datasets/xTRam1/plan-and-act-data"}
{"aliases": [], "authors": [], "details": {"availability": "Ungated JSON conversation shards and ranking-file manifest.", "dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "train", "qualifier": "exact", "unit": "conversations", "value": 600}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "test-task", "qualifier": "exact", "unit": "conversations", "value": 34}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "test-website", "qualifier": "exact", "unit": "conversations", "value": 42}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "test-subdomain", "qualifier": "exact", "unit": "conversations", "value": 44}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "train", "qualifier": "exact", "unit": "dialogue turns", "value": 2896}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "test-task", "qualifier": "exact", "unit": "dialogue turns", "value": 191}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "test-website", "qualifier": "exact", "unit": "dialogue turns", "value": 218}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md", "population": "test-subdomain", "qualifier": "exact", "unit": "dialogue turns", "value": 216}], "evidence_urls": ["https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md"], "formats": ["JSON", "auxiliary PKL"], "modalities": ["dialogue text", "HTML", "element candidates", "actions"], "notes": "Conversation, dialogue-turn and action-step units differ. The released file manifest is inspected; underlying examples were not downloaded.", "splits": ["train", "test_task", "test_website", "test_subdomain"], "supervision": ["multi-turn web instruction following", "element/action prediction"]}, "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md"], "reason": "Publisher dataset-card declaration for the multi-turn annotation release."}}, "release_revision": "5ae9a3bedf431f54c2218fbb8ee474e5efbda3a8"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "8685e0a26a06b3339b5c078632a7b0ade1ead40e6b586ef1ff3676414699a4bc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "details.dataset_profile", "licenses.data"], "url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "4455e6daf1151694c03b2a1ebd4a7af40b5b079b0568706a1783a0222383bb1d", "supports": ["identity", "release_status", "details.release_revision"], "url": "https://huggingface.co/api/datasets/magicgh/MT-Mind2Web"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "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=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "url:bdfb50330deb97893ce4", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T12:34:10Z", "licenses": {"code": null, "data": {"basis": "Publisher dataset-card declaration for the multi-turn annotation release.", "declared": "cc-by-4.0", "retrieved_at": "2026-10-17T12:34:10Z", "url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web"}], "name": "MT-Mind2Web", "notes": "The ungated Hub manifest contains train, test-task, test-website and test-subdomain JSON shards plus a candidate-ranking pickle file. Counts distinguish conversation sessions from dialogue turns; neither is a screenshot or action-step count. No pickle or sample data was loaded. The card declares CC-BY-4.0 and identifies the original Mind2Web interactions as its source.", "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=800&max_results=100&sortBy=submittedDate&sortOrder=descending", "target_id": "arxiv:2402.15057", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Multi-turn web-navigation conversations constructed from Mind2Web interactions, with dialogue turns, HTML snapshots, element candidates and click/type/select action sequences.", "topics": ["web-navigation", "imitation-learning", "memory"], "updated_at": null, "url": "https://huggingface.co/datasets/magicgh/MT-Mind2Web"}
{"aliases": [], "authors": ["Greg Little", "Tessa A. Lau", "Allen Cypher", "James Lin", "Eben M. Haber", "Eser Kandogan"], "details": {"publication": {"authors": ["Greg Little", "Tessa A. Lau", "Allen Cypher", "James Lin", "Eben M. Haber", "Eser Kandogan"], "date_basis": "Supplied year precision: 2007. CHI 2007 proceedings heading and copyright year The IBM page has an unlabeled Date of 22 Oct 2007, while the paper gives April/May event dates. Their roles do not establish a reliable day-level publication date.", "date_parts": [2007], "doi": "10.1145/1240624.1240767", "event_dates": null, "evidence_urls": ["https://research.ibm.com/publications/koala-capture-share-automate-personalize-business-processes-on-the-web", "https://plunk.org/eben/PublishedPapers/KoalaCHI.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": "This proposal is the specific Koala paper. It does not automatically merge every later CoScripter publication or project record. Exact publication date is unresolved; only the clearly established year is proposed.", "pages": "943-946", "pdf_url": "https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf", "publisher": null, "series": null, "title": "Koala: Capture, Share, Automate, Personalize Business Processes on the Web", "type": "conference-paper", "url": "https://doi.org/10.1145/1240624.1240767", "venue": "CHI 2007 Proceedings", "volume": null}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:23Z", "role": "primary", "sha256": "ae61a94cca68e058858244b6ff54ed9409973a5cecaf25c0e325b27adaf0e3dd", "supports": ["details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.url"], "url": "https://research.ibm.com/publications/koala-capture-share-automate-personalize-business-processes-on-the-web"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:55:21Z", "role": "primary", "sha256": "944b6f74a6cdb88fc253009f364dcc71cc42186b4907e04720da128c5292e357", "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.venue"], "url": "https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:55:21Z", "role": "primary", "sha256": "944b6f74a6cdb88fc253009f364dcc71cc42186b4907e04720da128c5292e357", "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.venue", "environments", "kind", "name", "scope", "summary", "topics"], "url": "https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:23Z", "role": "primary", "sha256": "ae61a94cca68e058858244b6ff54ed9409973a5cecaf25c0e325b27adaf0e3dd", "supports": ["details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.url"], "url": "https://research.ibm.com/publications/koala-capture-share-automate-personalize-business-processes-on-the-web"}], "id": "url:cccd967aa993404d74d0", "identifiers": {"doi": "10.1145/1240624.1240767"}, "kind": "paper", "last_verified_at": "2026-10-17T09:55:21Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://doi.org/10.1145/1240624.1240767"}, {"rel": "publication_pdf", "url": "https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf"}], "name": "Koala: Capture, Share, Automate, Personalize Business Processes on the Web", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. This proposal is the specific Koala paper. It does not automatically merge every later CoScripter publication or project record.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Koala, a collaborative web automation system that records browser interactions as editable pseudo-natural-language scripts. Scripts are shared through a wiki and instantiated with personal or corporate data for reuse by other users.", "topics": ["web-navigation", "workflow", "recording"], "updated_at": null, "url": "https://doi.org/10.1145/1240624.1240767"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/zai-org/webrl-orm-llama-3.1-8b/tree/main", "gated": false, "hub_id": "zai-org/webrl-orm-llama-3.1-8b", "languages": [], "modalities": [], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "8bafcdf63fe484b6005ca7082da1310459f8c87d4c011e84293a3cc8fca6ae9e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/api/models/zai-org/webrl-orm-llama-3.1-8b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:29Z", "role": "primary", "sha256": "8bafcdf63fe484b6005ca7082da1310459f8c87d4c011e84293a3cc8fca6ae9e", "supports": ["identity"], "url": "https://huggingface.co/api/models/THUDM/webrl-orm-llama-3.1-8b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "eb217b96c54c3c2ccfd3a27255c6fd740df40c2d6743ae811858339bf7f740b0", "supports": ["summary", "environments", "relations"], "url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "8bafcdf63fe484b6005ca7082da1310459f8c87d4c011e84293a3cc8fca6ae9e", "supports": ["details", "licenses.weights"], "url": "https://huggingface.co/api/models/zai-org/webrl-orm-llama-3.1-8b"}], "id": "url:d1e85aeb6d96859c109d", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:05:40Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/zai-org/webrl-orm-llama-3.1-8b"}], "name": "WebRL outcome reward model 8B", "notes": "The official API resolves the former THUDM namespace to zai-org and lists four safetensors shards. No model card or weight-license declaration is present in the inspected manifest. The official WebRL README identifies this as its outcome-supervised reward model; it is a training component rather than an action policy.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md", "target_id": "arxiv:2411.02337", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "The outcome-supervision checkpoint used by WebRL to score browser trajectories during policy training.", "topics": ["reward-models", "reinforcement-learning", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/zai-org/webrl-orm-llama-3.1-8b"}
{"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": {"access": "publisher-managed", "availability": "not independently tested", "pricing_url": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:28Z", "role": "primary", "sha256": "461de0060fae1ded024dfa724624b001665b463b14f454afafebd0c9a2dce71d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://www.tinyfish.ai/"}], "id": "url:d92596cfe618e5322ec2", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-16T06:38:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://www.tinyfish.ai/"}], "name": "TinyFish", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://www.tinyfish.ai/", "target_id": "organization:tinyfish", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Browser-agent API for executing web tasks and extracting structured results.", "topics": ["web-navigation"], "updated_at": null, "url": "https://www.tinyfish.ai/"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/apurvaga/go-browse-wa-qwen-7B/tree/main", "gated": false, "hub_id": "apurvaga/go-browse-wa-qwen-7B"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "470ef6f975990518f9622b259a6b8e5761b26dcd22985c248f1df2beb4aa2760", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/apurvaga/go-browse-wa-qwen-7B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "d083bf53b20c6c6124f1df29a83ce783799d46a1ab5600d19d658729fc6d67bb", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/apurvaga/go-browse-wa-qwen-7B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "0d4b28c993ccb977c7910ff3fecb5de29db364a7a7ad55523c8e052c9bae2566", "supports": ["summary", "environments"], "url": "https://raw.githubusercontent.com/ApGa/Go-Browse/HEAD/README.md"}], "id": "url:db1b7b7efe3755ae5693", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:44:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/apurvaga/go-browse-wa-qwen-7B"}], "name": "Go-Browse WebArena Qwen 7B", "notes": "Seven public safetensors shards and model/tokenizer files are present. The card links the official code and paper but declares no weight license. The code repository’s MIT license is not inherited by this checkpoint, and the separate NNetNav-trained comparison model is not treated as an alias.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/apurvaga/go-browse-wa-qwen-7B/raw/main/README.md", "target_id": "arxiv:2506.03533", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A Qwen-2.5-7B-Instruct checkpoint fine-tuned on Go-Browse WebArena trajectories for website task execution.", "topics": ["web-navigation", "imitation-learning"], "updated_at": null, "url": "https://huggingface.co/apurvaga/go-browse-wa-qwen-7B"}
{"aliases": ["Nova Act"], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T17:18:21Z", "role": "primary", "sha256": "4e4e60092351a85d0b99186a745d6295b170958be59092a786eda369923bd691", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://aws.amazon.com/nova/act/"}], "id": "url:dec156cbf0cb1181f7b6", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-17T17:18:21Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://aws.amazon.com/nova/act/"}], "name": "Amazon Nova Act", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "AWS service for building and deploying agents that automate browser-based UI workflows from natural language and Python, powered by a custom Nova 2 Lite model.", "topics": ["automation", "web-navigation"], "updated_at": null, "url": "https://aws.amazon.com/nova/act/"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [], "evidence_urls": ["https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md", "https://huggingface.co/api/datasets/CognitiveKernel/WebEvolver"], "formats": ["JSONL"], "modalities": [], "notes": "Two purpose-specific files are released: webevolver_policy_sft.jsonl and webevolver_world_model_sft.jsonl. They are not assumed disjoint train/test splits, and no modality or example count is inferred from filenames alone.", "splits": [], "supervision": ["agent-policy supervised fine-tuning", "web-world-model supervised fine-tuning"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/tree/13f4f45b359e138a45b1c89bd4e913298c586ca9", "gated": false, "hub_id": "CognitiveKernel/WebEvolver", "hub_revision": "13f4f45b359e138a45b1c89bd4e913298c586ca9", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md"], "reason": "Publisher dataset-card declaration for this release; it does not establish rights to all upstream screenshots, applications or source content."}}, "repository_url": "https://github.com/Tencent/SelfEvolvingAgent/tree/main/WebEvolver"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:17Z", "role": "primary", "sha256": "36ccdce12c1b1512ad4b21326c5d0c0379c73e85f0b8150779d43240b54686de", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "licenses.data", "details.dataset_profile"], "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:29Z", "role": "primary", "sha256": "cbef3f8037405d311ed9fe8aa620ec3734f91c77ca8f4209fbe98167fb33bee8", "supports": ["identity", "release_status", "details.hub_id", "details.hub_revision", "details.gated", "details.disabled", "details.download_url", "details.dataset_profile"], "url": "https://huggingface.co/api/datasets/CognitiveKernel/WebEvolver"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "447dfa1e52789f7f3f111e9e0d453fb1f9fd8bf2d7b858f28fbe21a2239b4b2e", "supports": ["details.repository_url"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/WebEvolver/README.md"}], "id": "url:e126bbd9729ebdaf05ab", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-18T04:35:17Z", "licenses": {"code": null, "data": {"basis": "Publisher dataset-card declaration for this release; it does not establish rights to all upstream screenshots, applications or source content.", "declared": "MIT", "retrieved_at": "2026-10-18T04:35:17Z", "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver"}], "name": "WebEvolver SFT data", "notes": "The card and public manifest identify separate policy-model and world-model JSONL files. No row total or complete record schema is documented in the inspected card; the files were not downloaded as training samples. MIT is the publisher’s data declaration.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md", "target_id": "arxiv:2504.21024", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Two JSONL collections for supervised fine-tuning the WebEvolver agent policy and web world model.", "topics": ["web-navigation", "world-models", "imitation-learning"], "updated_at": null, "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver"}
{"aliases": [], "authors": [], "details": {"access": "publisher-managed", "availability": "not independently tested", "pricing_url": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "69e029d01dda3806f88abfaee01da08c671adcb3555f1314e975ab8942a686ae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://browser-use.com/"}], "id": "url:ef244017b1460269fc68", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-16T06:38:27Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://browser-use.com/"}], "name": "Browser Use Cloud", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://browser-use.com/", "target_id": "organization:browser-use", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Hosted browser-agent platform from Browser Use.", "topics": ["web-navigation"], "updated_at": null, "url": "https://browser-use.com/"}
{"aliases": [], "authors": [], "details": {"access": "publisher-managed", "availability": "not independently tested", "pricing_url": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "474a27cf644ac5f8400a023086fb8d356e6148de71d1851337215f88624852d9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://openai.com/index/introducing-operator/"}], "id": "url:f0584842138a52158f51", "identifiers": {}, "kind": "product", "last_verified_at": "2026-10-16T06:38:27Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://openai.com/index/introducing-operator/"}], "name": "Operator", "notes": "Historical product entry; the official page describes its integration into ChatGPT agent. Do not infer current standalone availability.", "published_at": null, "relations": [{"evidence_url": "https://openai.com/index/introducing-operator/", "target_id": "organization:openai", "type": "published_by"}], "review_status": "curated", "scope": "core", "summary": "Historical OpenAI browser-agent research preview.", "topics": ["web-navigation", "history"], "updated_at": null, "url": "https://openai.com/index/introducing-operator/"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/Hcompany/Holo1-7B/tree/main", "gated": false, "hub_id": "Hcompany/Holo1-7B"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "a7e185380bb02150d761651420cccc9c4a57f67c70a5dec3b7f85af63fb8773b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/Hcompany/Holo1-7B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "66076674ebbdcf9faeb408ebd137ac1283fb56b1011acd7037ebb73e09ba749d", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/Hcompany/Holo1-7B"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:44:50Z", "role": "primary", "sha256": "a7e185380bb02150d761651420cccc9c4a57f67c70a5dec3b7f85af63fb8773b", "supports": ["licenses.weights"], "url": "https://huggingface.co/Hcompany/Holo1-7B/raw/main/README.md"}], "id": "url:f35cfcbaeb79663eabeb", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:44:50Z", "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": "2026-10-17T02:44:50Z", "url": "https://huggingface.co/Hcompany/Holo1-7B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/Hcompany/Holo1-7B"}], "name": "Holo1 7B", "notes": "Four public safetensors shards and example navigation/localization files are present. The model card declares Apache-2.0. Reported Surfer-H results depend on the assembled policy/localizer/validator configuration and are not imported as standalone checkpoint scores.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/Hcompany/Holo1-7B/raw/main/README.md", "target_id": "arxiv:2506.02865", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A Qwen2.5-VL-based action model released for the Surfer-H web-agent system, supporting policy, interface-localization and validation roles.", "topics": ["web-navigation", "grounding", "planning"], "updated_at": null, "url": "https://huggingface.co/Hcompany/Holo1-7B"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_checklist_generation/raw/main/README.md", "population": "Hub split named test", "qualifier": "exact", "unit": "rows", "value": 3630}], "evidence_urls": ["https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_checklist_generation/raw/main/README.md", "https://huggingface.co/api/collections/LangAGI-Lab/web-shepherd-advancing-prms-for-reinforcing-web-agents-682b4f4ad607fc27c4dc49e8"], "formats": ["Parquet"], "modalities": ["screenshots", "text", "action history"], "notes": "This is the checklist-generation release rather than the separate preference-pair collection. No unique-task count or data permission is inferred.", "splits": ["test"], "supervision": ["checklist generation", "subgoal analysis"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_checklist_generation/tree/main", "gated": false, "hub_id": "LangAGI-Lab/WebPRMCollection_checklist_generation", "languages": [], "modalities": ["image", "text"], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:33Z", "role": "primary", "sha256": "2e4142675f2cb994c21a7620a5188b10147ab03326a2047946886b5916789ec2", "supports": ["name", "kind", "identity", "release_status", "details", "summary"], "url": "https://huggingface.co/api/datasets/LangAGI-Lab/WebPRMCollection_checklist_generation"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:33Z", "role": "primary", "sha256": "3693cb307065a6ced63a4789a7f592b5eba0ad6a5ffe1a3caf0649decd5cb6e8", "supports": ["details"], "url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_checklist_generation/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:45:05Z", "role": "primary", "sha256": "647d9e7bdad5fe2c03a8cfcd34d93f3b2bf7c1553b517a4a843bc0113fa47456", "supports": ["identity", "summary"], "url": "https://huggingface.co/api/collections/LangAGI-Lab/web-shepherd-advancing-prms-for-reinforcing-web-agents-682b4f4ad607fc27c4dc49e8"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:04Z", "role": "primary", "sha256": "07431058d8b9a71b80e75b7a74a40a8346c1503eec240628459cc4d4e2f0802f", "supports": ["summary", "environments", "topics", "scope"], "url": "https://raw.githubusercontent.com/kyle8581/Web-Shepherd/HEAD/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T04:01:33Z", "role": "primary", "sha256": "2e4142675f2cb994c21a7620a5188b10147ab03326a2047946886b5916789ec2", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/LangAGI-Lab/WebPRMCollection_checklist_generation"}], "id": "url:fcc98b149272bbf31b05", "identifiers": {}, "kind": "dataset", "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/LangAGI-Lab/WebPRMCollection_checklist_generation"}], "name": "WebPRM checklist-generation data", "notes": "Five public Parquet shards are present. This collection provides checklist-generation supervision and is distinct from the previously cataloged chosen/rejected preference data. The Hub calls its split test; that label is preserved. No data license is declared.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Web-task context, starting observations and action history paired with structured checklists and subgoal analyses for checklist generation.", "topics": ["reward-models", "data-generation", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/datasets/LangAGI-Lab/WebPRMCollection_checklist_generation"}