{"aliases": [], "authors": ["Hongxin Li", "Jingran Su", "Yuntao Chen", "Qing Li", "Zhaoxiang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "320ff40872677e9d32efae360dc289ca345076e037e7044188868f765e816008", "submitted_at": "2023-05-30T17:59:30Z", "updated_at": "2023-10-30T06:36:24Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2305.19308v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:56Z", "role": "primary", "sha256": "beb8c811bf2d070845aa6fcc7e94ffd43ca4f9f60ce42c72d958598790fbdffc", "supports": ["environments", "notes"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md"}], "id": "arxiv:2305.19308", "identifiers": {"arxiv": "2305.19308"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2305.19308"}], "name": "SheetCopilot: Bringing Software Productivity to the Next Level through Large Language Models", "notes": "The contribution was reviewed from the retained paper abstract. The original Windows/Excel implementation is identified in the separately retained project README; the later Ubuntu evaluator is not treated as a paper result. No results were independently reproduced.", "published_at": "2023-05-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SheetCopilot, an agent that steers spreadsheet software via atomic virtual APIs, along with a 221-task spreadsheet manipulation benchmark.", "topics": ["workflow", "tool-integration", "evaluation"], "updated_at": "2023-10-30", "url": "https://arxiv.org/abs/2305.19308"}
{"aliases": [], "authors": ["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": ["Yiduo Guo", "Zekai Zhang", "Yaobo Liang", "Dongyan Zhao", "Nan Duan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "58dfa75fd82a2cc1d41954f10feb1d987de6469782d11d04c3b8bfffc8a3b745", "submitted_at": "2023-11-03T08:06:35Z", "updated_at": "2023-11-07T10:13:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2311.01767v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2311.01767", "identifiers": {"arxiv": "2311.01767"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2311.01767"}], "name": "PPTC Benchmark: Evaluating Large Language Models for PowerPoint Task Completion", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-11-03", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Evaluates multi-turn PowerPoint creation and editing, scoring the resulting presentation files rather than requiring a fixed API sequence.", "topics": ["evaluation", "tool-integration"], "updated_at": "2023-11-07", "url": "https://arxiv.org/abs/2311.01767"}
{"aliases": [], "authors": ["Grégoire Mialon", "Clémentine Fourrier", "Craig Swift", "Thomas Wolf", "Yann LeCun", "Thomas Scialom"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "29ddf2b7b4d495ee18bbd263cd77f86a71b3014b58a6f140c3bfce7cbae9c922", "submitted_at": "2023-11-21T20:34:47Z", "updated_at": "2023-11-21T20:34:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2311.12983v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2311.12983", "identifiers": {"arxiv": "2311.12983"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2311.12983"}], "name": "GAIA: a benchmark for General AI Assistants", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-11-21", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces questions that test assistant reasoning, multimodal understanding, browsing, and tool use beyond screen-action execution alone.", "topics": ["evaluation", "tool-integration"], "updated_at": "2023-11-21", "url": "https://arxiv.org/abs/2311.12983"}
{"aliases": [], "authors": ["Zhiyong Wu", "Chengcheng Han", "Zichen Ding", "Zhenmin Weng", "Zhoumianze Liu", "Shunyu Yao", "Tao Yu", "Lingpeng Kong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "674b005f670a3593bd215a8ff75eac05c90ee068578fc26afdd998c40f5b7b71", "submitted_at": "2024-02-12T07:29:22Z", "updated_at": "2024-02-15T09:30:48Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2402.07456v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2402.07456", "identifiers": {"arxiv": "2402.07456"}, "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/2402.07456"}], "name": "OS-Copilot: Towards Generalist Computer Agents with Self-Improvement", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-02-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Provides an operating-system agent framework connecting applications, files, terminals, and web tools, with accumulated skills for subsequent tasks.", "topics": ["tool-integration", "planning"], "updated_at": "2024-02-15", "url": "https://arxiv.org/abs/2402.07456"}
{"aliases": [], "authors": ["Zhen Xiang", "Linzhi Zheng", "Yanjie Li", "Junyuan Hong", "Qinbin Li", "Han Xie", "Jiawei Zhang", "Zidi Xiong", "Chulin Xie", "Carl Yang", "Dawn Song", "Bo Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "50a8ea492d3f1c2ff2bd00ebfb18e19077bacc96b44c751d985bfe0f7bef0247", "submitted_at": "2024-06-13T14:49:26Z", "updated_at": "2025-05-29T03:09:05Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2406.09187v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:42:42Z", "role": "primary", "sha256": "d1c1bc3fee56d124427abd2aabbcf876418633a8adf7a03f4ef0317cba6a42ae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2008.05132%2C2010.04295%2C2101.04893%2C2112.05692%2C2306.01242%2C2311.03079%2C2401.05459%2C2402.03746%2C2402.15506%2C2403.20329%2C2404.06395%2C2404.10179%2C2405.10292%2C2406.09187%2C2406.14596%2C2408.03326%2C2409.12089%2C2501.10893%2C2504.09352&max_results=19"}], "id": "arxiv:2406.09187", "identifiers": {"arxiv": "2406.09187"}, "kind": "paper", "last_verified_at": "2026-10-16T06:42:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.09187"}], "name": "GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoning", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The web-agent evidence is Mind2Web-SC; the healthcare benchmark does not independently establish GUI control.", "published_at": "2024-06-13", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces GuardAgent, which translates safety requests into executable guardrail code using plans and retrieved demonstrations. Mind2Web-SC supplies a web-agent safety-policy evaluation alongside a separate healthcare access-control benchmark.", "topics": ["safety-security", "evaluation", "tool-integration"], "updated_at": "2025-05-29", "url": "https://arxiv.org/abs/2406.09187"}
{"aliases": [], "authors": ["Shihan Deng", "Weikai Xu", "Hongda Sun", "Wei Liu", "Tao Tan", "Jianfeng Liu", "Ang Li", "Jian Luan", "Bin Wang", "Rui Yan", "Shuo Shang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "90ce9900644c21004e0ba9e749e10d94238d2447d0e84bd7205ff55dcbd243f4", "submitted_at": "2024-07-01T06:10:01Z", "updated_at": "2024-07-01T06:10:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.00993v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.00993", "identifiers": {"arxiv": "2407.00993"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.00993"}], "name": "Mobile-Bench: An Evaluation Benchmark for LLM-based Mobile Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates mobile agents using combined UI and API actions, tasks with varying app scope, and checkpoints for intermediate progress.", "topics": ["evaluation", "cross-app", "tool-integration"], "updated_at": "2024-07-01", "url": "https://arxiv.org/abs/2407.00993"}
{"aliases": [], "authors": ["Jiayi Zhang", "Chuang Zhao", "Yihan Zhao", "Zhaoyang Yu", "Ming He", "Jianping Fan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aa74755c53e20b3a9ff4b72e03a557f455ecceaefc96a50587aefd78ac3c8176", "submitted_at": "2024-07-04T13:12:19Z", "updated_at": "2024-07-04T13:12:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.03913v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.03913", "identifiers": {"arxiv": "2407.03913"}, "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/2407.03913"}], "name": "MobileExperts: A Dynamic Tool-Enabled Agent Team in Mobile Devices", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-07-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MobileExperts, which assembles mobile-agent teams, lets agents explore and formulate tools, and coordinates them through two levels of planning. A benchmark organizes tasks by hierarchical intelligence levels.", "topics": ["mobile-navigation", "planning", "tool-integration"], "updated_at": "2024-07-04", "url": "https://arxiv.org/abs/2407.03913"}
{"aliases": [], "authors": ["Harsh Trivedi", "Tushar Khot", "Mareike Hartmann", "Ruskin Manku", "Vinty Dong", "Edward Li", "Shashank Gupta", "Ashish Sabharwal", "Niranjan Balasubramanian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e59087ecf28d07baf81b60e4166a866d4bea4d8cd934bea06c2149dd61f38f1a", "submitted_at": "2024-07-26T17:55:45Z", "updated_at": "2024-07-26T17:55:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.18901v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.18901", "identifiers": {"arxiv": "2407.18901"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.18901"}], "name": "AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-26", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces simulated application APIs and state-based task evaluation for interactive coding agents performing everyday digital workflows.", "topics": ["evaluation", "tool-integration", "cross-app"], "updated_at": "2024-07-26", "url": "https://arxiv.org/abs/2407.18901"}
{"aliases": [], "authors": ["Junting Lu", "Zhiyang Zhang", "Fangkai Yang", "Jue Zhang", "Lu Wang", "Chao Du", "Qingwei Lin", "Saravan Rajmohan", "Dongmei Zhang", "Qi Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "618f41c2253a532780766b38396f241a48b587c580d6b4d7065d1519e6149751", "submitted_at": "2024-09-25T17:58:08Z", "updated_at": "2025-05-19T16:12:40Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2409.17140v2"}, "publication": {"authors": ["Junting Lu", "Zhiyang Zhang", "Fangkai Yang", "Jue Zhang", "Lu Wang", "Chao Du", "Qingwei Lin", "Saravan Rajmohan", "Dongmei Zhang", "Qi Zhang"], "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.381", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.acl-long.381/"], "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": "7711-7743", "pdf_url": "https://aclanthology.org/2025.acl-long.381.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "AXIS: Efficient Human-Agent-Computer Interaction with API-First LLM-Based Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.acl-long.381/", "venue": "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)", "volume": null}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:52Z", "role": "primary", "sha256": "31ab26306d5c41b25731db1ad7f73495330e637c1694c9da644f4e238a73917b", "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.381/"}, {"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:2409.17140", "identifiers": {"arxiv": "2409.17140"}, "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/2409.17140"}, {"rel": "publication", "url": "https://aclanthology.org/2025.acl-long.381/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.acl-long.381.pdf"}], "name": "AXIS: Efficient Human-Agent-Computer Interaction with API-First LLM-Based Agents", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-09-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AXIS, an API-first application agent that can use UI actions and expand its available APIs through automated application exploration. The work evaluates the approach on Microsoft Word tasks.", "topics": ["tool-integration", "automation", "efficiency"], "updated_at": "2025-05-19", "url": "https://arxiv.org/abs/2409.17140"}
{"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": ["Chengyou Jia", "Minnan Luo", "Zhuohang Dang", "Qiushi Sun", "Fangzhi Xu", "Junlin Hu", "Tianbao Xie", "Zhiyong Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ece9f64cc037dad8633388edf1b54d88818d8092b77e19b15780e929b23aa593", "submitted_at": "2024-10-24T09:58:40Z", "updated_at": "2024-10-24T09:58:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2410.18603v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.18603", "identifiers": {"arxiv": "2410.18603"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.18603"}], "name": "AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a platform that integrates specialized agents under a coordinating MetaAgent for varied computer tasks.", "topics": ["automation", "tool-integration"], "updated_at": "2024-10-24", "url": "https://arxiv.org/abs/2410.18603"}
{"aliases": [], "authors": ["Xiaoqiang Wang", "Bang Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "66fa51b108ac79cc43a895ba757ca463d5008bfe6db5cde4bdf3a3bb38c444eb", "submitted_at": "2024-10-24T17:58:08Z", "updated_at": "2024-10-24T17:58:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2410.18963v1"}}, "environments": ["desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.18963", "identifiers": {"arxiv": "2410.18963"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.18963"}], "name": "OSCAR: Operating System Control via State-Aware Reasoning and Re-Planning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "OSCAR translates user instructions into Python control code while observing screen images and acting through mouse and keyboard operations. A state-machine controller handles errors and replans tasks during execution across desktop and mobile applications.", "topics": ["planning", "state-tracking", "tool-integration", "robustness"], "updated_at": "2024-10-24", "url": "https://arxiv.org/abs/2410.18963"}
{"aliases": [], "authors": ["Adam Fourney", "Gagan Bansal", "Hussein Mozannar", "Cheng Tan", "Eduardo Salinas", "Erkang", "Zhu", "Friederike Niedtner", "Grace Proebsting", "Griffin Bassman", "Jack Gerrits", "Jacob Alber", "Peter Chang", "Ricky Loynd", "Robert West", "Victor Dibia", "Ahmed Awadallah", "Ece Kamar", "Rafah Hosn", "Saleema Amershi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4a2c499237dbcbe1482824a2710f5185dee05c9ed66a0882d1e422f30afe056a", "submitted_at": "2024-11-07T06:36:19Z", "updated_at": "2024-11-07T06:36:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2411.04468v1"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2411.04468", "identifiers": {"arxiv": "2411.04468"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.04468"}], "name": "Magentic-One: A Generalist Multi-Agent System for Solving Complex 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": "2024-11-07", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces an orchestrated team of browsing, file, and coding agents with progress tracking and error recovery for general digital tasks.", "topics": ["planning", "tool-integration"], "updated_at": "2024-11-07", "url": "https://arxiv.org/abs/2411.04468"}
{"aliases": [], "authors": ["Weikai Xie", "Li Zhang", "Shihe Wang", "Rongjie Yi", "Mengwei Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9c22d6a9aafb89e5628198fb4f6ea50fe40b1e955a2fa367571459753cea61df", "submitted_at": "2024-11-30T08:55:39Z", "updated_at": "2024-11-30T08:55:39Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2412.00402v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2412.00402", "identifiers": {"arxiv": "2412.00402"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.00402"}], "name": "DroidCall: A Dataset for LLM-powered Android Intent Invocation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-11-30", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces data and a generation pipeline for translating natural-language requests into Android intent invocations using compact language models.", "topics": ["tool-integration", "data-generation"], "updated_at": "2024-11-30", "url": "https://arxiv.org/abs/2412.00402"}
{"aliases": [], "authors": ["Hao Wen", "Shizuo Tian", "Borislav Pavlov", "Wenjie Du", "Yixuan Li", "Ge Chang", "Shanhui Zhao", "Jiacheng Liu", "Yunxin Liu", "Ya-Qin Zhang", "Yuanchun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3190c452dbfa24b131cc34f66192bbdb9c538d30a1919082a9fed334f6169fff", "submitted_at": "2024-12-24T02:54:56Z", "updated_at": "2025-05-06T12:24:43Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2412.18116v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2412.18116", "identifiers": {"arxiv": "2412.18116"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.18116"}], "name": "AutoDroid-V2: Boosting SLM-based GUI Agents via Code Generation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-12-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Recasts mobile UI automation as on-device code generation, using app-specific documentation and synthetic tasks to guide smaller models.", "topics": ["mobile-navigation", "tool-integration"], "updated_at": "2025-05-06", "url": "https://arxiv.org/abs/2412.18116"}
{"aliases": [], "authors": ["Pranjal Aggarwal", "Sean Welleck"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9e4091fc740249b6406dd9e96f2692d7fac9323759ec6fad6ba5a05f3e445844", "submitted_at": "2025-02-24T18:41:33Z", "updated_at": "2025-10-03T02:09:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2502.18525v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2502.18525", "identifiers": {"arxiv": "2502.18525"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.18525"}], "name": "Programming with Pixels: Can Computer-Use Agents do Software Engineering?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a visual IDE environment and benchmark for software-engineering tasks, comparing screenshot-based control with access to editing and shell tools.", "topics": ["evaluation", "tool-integration", "automation"], "updated_at": "2025-10-03", "url": "https://arxiv.org/abs/2502.18525"}
{"aliases": [], "authors": ["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": ["Chaoyun Zhang", "Shilin He", "Liqun Li", "Si Qin", "Yu Kang", "Qingwei Lin", "Saravan Rajmohan", "Dongmei Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "43a533866496d3ebd3e34c04142a702e4a3aaa387c3830e3bb993fdbb0ce8432", "submitted_at": "2025-03-14T04:26:21Z", "updated_at": "2025-06-23T13:01:02Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2503.11069v2"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.11069", "identifiers": {"arxiv": "2503.11069"}, "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/2503.11069"}], "name": "API Agents vs. GUI Agents: Divergence and Convergence", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares API-driven and GUI-driven agents across architecture, development, and interaction considerations, discussing when hybrid approaches are useful.", "topics": ["tool-integration", "survey", "automation"], "updated_at": "2025-06-23", "url": "https://arxiv.org/abs/2503.11069"}
{"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": ["Chaoyun Zhang", "He Huang", "Chiming Ni", "Jian Mu", "Si Qin", "Shilin He", "Lu Wang", "Fangkai Yang", "Pu Zhao", "Chao Du", "Liqun Li", "Yu Kang", "Zhao Jiang", "Suzhen Zheng", "Rujia Wang", "Jiaxu Qian", "Minghua Ma", "Jian-Guang Lou", "Qingwei Lin", "Saravan Rajmohan", "Dongmei Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dc47a80b48a6a5f126e36fa14639d56d8c8025a453fbcee2cbdb73eb67945561", "submitted_at": "2025-04-20T13:04:43Z", "updated_at": "2025-04-25T05:14:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.14603v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:47Z", "role": "primary", "sha256": "27144160e8981bb9f36b60c391689774913ff3f026eeeffc2923154c404a6529", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2504.03206%2C2504.03601%2C2504.06821%2C2504.09723%2C2504.11301%2C2504.11543%2C2504.12609%2C2504.14233%2C2504.14348%2C2504.14603%2C2504.16073%2C2504.16419%2C2504.17892%2C2504.18580%2C2504.20965%2C2505.00019%2C2505.00693%2C2505.01578%2C2505.02094%2C2505.03570%2C2505.03574%2C2505.05026%2C2505.05849%2C2505.06507%2C2505.07003%2C2505.07473%2C2505.10468%2C2505.10887%2C2505.12053%2C2505.12490%2C2505.12728%2C2505.13291%2C2505.14030%2C2505.14454%2C2505.14646%2C2505.15801%2C2505.15816%2C2505.17020%2C2505.18746%2C2505.19095%2C2505.19713%2C2505.20023%2C2505.20732%2C2505.20795%2C2505.21055%2C2505.22304%2C2505.22429%2C2505.23450%2C2505.23559%2C2506.00641&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:07Z", "role": "primary", "sha256": "0ff3084c62a2ae0fbda31be5d409fb8ca3a6887da897a238beec93a8d7be7825", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2504.14603v2"}], "id": "arxiv:2504.14603", "identifiers": {"arxiv": "2504.14603"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:07Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.14603"}], "name": "UFO2: The Desktop AgentOS", "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": "2025-04-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces UFO2, a Windows desktop agent architecture combining a coordinating HostAgent, application-specialized agents, and a shared GUI/API action layer. It integrates UI Automation, visual parsing, speculative planning, and an isolated picture-in-picture desktop.", "topics": ["automation", "cross-app", "tool-integration"], "updated_at": "2025-04-25", "url": "https://arxiv.org/abs/2504.14603"}
{"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": ["Kai Mei", "Xi Zhu", "Hang Gao", "Shuhang Lin", "Yongfeng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2c382ce428e6dd41ec07af7575461c69906ce386fd2bcd8ad9f0d792d2b8f0b3", "submitted_at": "2025-05-24T18:56:00Z", "updated_at": "2025-10-31T21:48:48Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.18829v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.18829", "identifiers": {"arxiv": "2505.18829"}, "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/2505.18829"}], "name": "LiteCUA: Computer as MCP Server for Computer-Use Agent on AIOS", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "AIOS 1.0 exposes computer states and actions through an MCP server to provide structured environmental context for agents. LiteCUA is a lightweight agent built on this interface, evaluated on OSWorld to study the effect of computer-environment contextualization.", "topics": ["tool-integration", "automation"], "updated_at": "2025-10-31", "url": "https://arxiv.org/abs/2505.18829"}
{"aliases": [], "authors": ["Yunhe Yan", "Shihe Wang", "Jiajun Du", "Yexuan Yang", "Yuxuan Shan", "Qichen Qiu", "Xianqing Jia", "Xinge Wang", "Xin Yuan", "Xu Han", "Mao Qin", "Yinxiao Chen", "Chen Peng", "Shangguang Wang", "Mengwei Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2334e017fe49b39dc5d460963988f05d11051f2101bebe23be76005809fafd45", "submitted_at": "2025-06-09T11:50:33Z", "updated_at": "2025-06-09T11:50:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.07672v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:12:09Z", "role": "primary", "sha256": "f64464b1516369a87f0fbe28828e32ff61687dae4ce063aedddd8bd45a7d40eb", "supports": ["scope", "summary", "environments"], "url": "https://arxiv.org/html/2506.07672"}], "id": "arxiv:2506.07672", "identifiers": {"arxiv": "2506.07672"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.07672"}], "name": "MCPWorld: A Unified Benchmarking Testbed for API, GUI, and Hybrid Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced. The platform or GUI evaluation setting was additionally checked in the primary full text.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "MCPWorld evaluates API, GUI, and hybrid computer-use agents with 201 tasks in applications whose source can be instrumented and extended with MCP support. It verifies completion from application behavior and packages the testbed in containers with GPU support.", "topics": ["evaluation", "tool-integration", "sandbox"], "updated_at": "2025-06-09", "url": "https://arxiv.org/abs/2506.07672"}
{"aliases": [], "authors": ["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": ["Linxin Song", "Yutong Dai", "Viraj Prabhu", "Jieyu Zhang", "Taiwei Shi", "Li Li", "Junnan Li", "Silvio Savarese", "Zeyuan Chen", "Jieyu Zhao", "Ran Xu", "Caiming Xiong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b6e4272a69c1a93f22bd35032547ffb6021a79c97d179a1075e14aee69dea83a", "submitted_at": "2025-08-05T21:33:36Z", "updated_at": "2026-02-20T10:14:08Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2508.03923v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.03923", "identifiers": {"arxiv": "2508.03923"}, "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/2508.03923"}], "name": "CoAct-1: Computer-using Multi-Agent System with Coding 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": "2025-08-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "CoAct-1 delegates subtasks between a GUI operator and a programmer that executes Python or Bash. Its OSWorld evaluation studies a hybrid action space in which scripts handle suitable file and data operations while visual interaction handles other steps.", "topics": ["tool-integration", "automation", "efficiency", "planning"], "updated_at": "2026-02-20", "url": "https://arxiv.org/abs/2508.03923"}
{"aliases": [], "authors": ["Hanyu Lai", "Xiao Liu", "Yanxiao Zhao", "Han Xu", "Hanchen Zhang", "Bohao Jing", "Yanyu Ren", "Shuntian Yao", "Yuxiao Dong", "Jie Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "383d8e80619d58e72ebb9c49229cd9522a0b3586b4bfa8f00b29b1834adca5b9", "submitted_at": "2025-08-19T17:59:45Z", "updated_at": "2025-10-21T05:18:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.14040v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.14040", "identifiers": {"arxiv": "2508.14040"}, "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/2508.14040"}], "name": "ComputerRL: Scaling End-to-End Online Reinforcement Learning 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": "2025-08-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines API and GUI actions with distributed desktop reinforcement learning and alternating supervised training to address training instability.", "topics": ["reinforcement-learning", "tool-integration"], "updated_at": "2025-10-21", "url": "https://arxiv.org/abs/2508.14040"}
{"aliases": [], "authors": ["Abhinav Kumar", "Jaechul Roh", "Ali Naseh", "Amir Houmansadr", "Eugene Bagdasarian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0f6ff68c59cede9cf59c039416a87ea39b926e3a3179b9f81af9354767befb93", "submitted_at": "2025-09-01T16:56:16Z", "updated_at": "2025-09-01T16:56:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.01619v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.01619", "identifiers": {"arxiv": "2509.01619"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.01619"}], "name": "Throttling Web Agents Using Reasoning Gates", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced. The contribution is an access-cost mechanism spanning websites and MCP servers, not a GUI navigation policy.", "published_at": "2025-09-01", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces reasoning gates that require agents to solve generated, verifiable puzzles before receiving access to resources. The throttling approach is deployed on a custom website and MCP servers and evaluated with web agents.", "topics": ["safety-security", "evaluation", "tool-integration"], "updated_at": "2025-09-01", "url": "https://arxiv.org/abs/2509.01619"}
{"aliases": [], "authors": ["Haoming Wang", "Haoyang Zou", "Huatong Song", "Jiazhan Feng", "Junjie Fang", "Junting Lu", "Longxiang Liu", "Qinyu Luo", "Shihao Liang", "Shijue Huang", "Wanjun Zhong", "Yining Ye", "Yujia Qin", "Yuwen Xiong", "Yuxin Song", "Zhiyong Wu", "Aoyan Li", "Bo Li", "Chen Dun", "Chong Liu", "Daoguang Zan", "Fuxing Leng", "Hanbin Wang", "Hao Yu", "Haobin Chen", "Hongyi Guo", "Jing Su", "Jingjia Huang", "Kai Shen", "Kaiyu Shi", "Lin Yan", "Peiyao Zhao", "Pengfei Liu", "Qinghao Ye", "Renjie Zheng", "Shulin Xin", "Wayne Xin Zhao", "Wen Heng", "Wenhao Huang", "Wenqian Wang", "Xiaobo Qin", "Yi Lin", "Youbin Wu", "Zehui Chen", "Zihao Wang", "Baoquan Zhong", "Xinchun Zhang", "Xujing Li", "Yuanfan Li", "Zhongkai Zhao", "Chengquan Jiang", "Faming Wu", "Haotian Zhou", "Jinlin Pang", "Li Han", "Qi Liu", "Qianli Ma", "Siyao Liu", "Songhua Cai", "Wenqi Fu", "Xin Liu", "Yaohui Wang", "Zhi Zhang", "Bo Zhou", "Guoliang Li", "Jiajun Shi", "Jiale Yang", "Jie Tang", "Li Li", "Qihua Han", "Taoran Lu", "Woyu Lin", "Xiaokang Tong", "Xinyao Li", "Yichi Zhang", "Yu Miao", "Zhengxuan Jiang", "Zili Li", "Ziyuan Zhao", "Chenxin Li", "Dehua Ma", "Feng Lin", "Ge Zhang", "Haihua Yang", "Hangyu Guo", "Hongda Zhu", "Jiaheng Liu", "Junda Du", "Kai Cai", "Kuanye Li", "Lichen Yuan", "Meilan Han", "Minchao Wang", "Shuyue Guo", "Tianhao Cheng", "Xiaobo Ma", "Xiaojun Xiao", "Xiaolong Huang", "Xinjie Chen", "Yidi Du", "Yilin Chen", "Yiwen Wang", "Zhaojian Li", "Zhenzhu Yang", "Zhiyuan Zeng", "Chaolin Jin", "Chen Li", "Hao Chen", "Haoli Chen", "Jian Chen", "Qinghao Zhao", "Guang Shi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "800e417f4a2f720beb436ddd2e7bc60f0dd2483b44ac007a72a5eab2c6001ff8", "submitted_at": "2025-09-02T17:44:45Z", "updated_at": "2025-09-05T14:59:27Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.02544v2"}}, "environments": ["web", "desktop", "windows", "android", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.02544", "identifiers": {"arxiv": "2509.02544"}, "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/2509.02544"}], "name": "UI-TARS-2 Technical Report: Advancing GUI Agent with 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": "2025-09-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-TARS-2 combines scalable trajectory generation, stabilized multi-turn reinforcement learning, and a sandbox that supports GUI, filesystem, and terminal actions. The report evaluates the resulting agent across web, desktop, Android, and game tasks and analyzes training stability.", "topics": ["reinforcement-learning", "data-generation", "sandbox", "tool-integration"], "updated_at": "2025-09-05", "url": "https://arxiv.org/abs/2509.02544"}
{"aliases": [], "authors": ["Pengxiang Zhao", "Guangyi Liu", "YaoZhen Liang", "Weiqing He", "Zhengxi Lu", "WenHao Wang", "Yuehao Huang", "Yuxiang Chai", "Zhaolu Kang", "Yaxuan Guo", "Hao Wang", "Kexin Zhang", "Liang Liu", "Yong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b1e0e807f5f396bfd9108ab6cfacd48ed4780abfb792589da0b41d7d246c717", "submitted_at": "2025-09-08T09:43:48Z", "updated_at": "2026-04-15T11:26:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.06477v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.06477", "identifiers": {"arxiv": "2509.06477"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.06477"}], "name": "MAS-Bench: A Unified Benchmark for Shortcut-Augmented Hybrid Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAS-Bench evaluates mobile agents that combine GUI actions with APIs, deep links, and reusable scripts. Its 139 tasks across 11 apps test use of 88 predefined shortcuts as well as the ability to generate new reusable workflows.", "topics": ["evaluation", "mobile-navigation", "tool-integration", "workflow"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2509.06477"}
{"aliases": [], "authors": ["Haochen Gong", "Chenxiao Li", "Rui Chang", "Wenbo Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "922c1d9b60db7c17b9019a320a4cfc2dd4eac325fd946f024ca82d43785f5eba", "submitted_at": "2025-09-26T12:19:27Z", "updated_at": "2026-01-14T06:07:30Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2509.22256v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.22256", "identifiers": {"arxiv": "2509.22256"}, "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/2509.22256"}], "name": "Secure and Efficient Access Control for Computer-Use Agents via Context Space", "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": "2025-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CSAgent, which combines static access policies with user intent and context and enforces them through an operating-system service. A supporting toolchain constructs and refines policies for agents using API, CLI and GUI interfaces.", "topics": ["safety-security", "tool-integration"], "updated_at": "2026-01-14", "url": "https://arxiv.org/abs/2509.22256"}
{"aliases": [], "authors": ["Zhen Yang", "Zi-Yi Dou", "Di Feng", "Forrest Huang", "Anh Nguyen", "Keen You", "Omar Attia", "Yuhao Yang", "Michael Feng", "Haotian Zhang", "Ram Ramrakhya", "Chao Jia", "Jeffrey Nichols", "Alexander Toshev", "Yinfei Yang", "Zhe Gan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "daa35e6bcd86fbdb709652d3592776ed93aebe2a0ebb3595fed91478dc0e3e0a", "submitted_at": "2025-09-30T17:13:56Z", "updated_at": "2025-09-30T17:13:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.26539v1"}}, "environments": ["android", "web", "desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.26539", "identifiers": {"arxiv": "2509.26539"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.26539"}], "name": "Ferret-UI Lite: Lessons from Building Small On-Device GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Ferret-UI Lite develops a 3B GUI agent using mixed real and synthetic data, reasoning with visual tools at inference, and reinforcement learning. It evaluates grounding and navigation across mobile, web, and desktop settings, including AndroidWorld and OSWorld.", "topics": ["grounding", "reinforcement-learning", "tool-integration", "efficiency"], "updated_at": "2025-09-30", "url": "https://arxiv.org/abs/2509.26539"}
{"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": ["Bin Lei", "Nuo Xu", "Ali Payani", "Mingyi Hong", "Chunhua Liao", "Yu Cao", "Caiwen Ding"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "798fc5ca4fe483a1235f0979b0c641e9c1732660780c13340c5221319a1b28fc", "submitted_at": "2025-10-05T05:15:45Z", "updated_at": "2025-10-05T05:15:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.04039v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "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:2510.04039", "identifiers": {"arxiv": "2510.04039"}, "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/2510.04039"}], "name": "\\textsc{GUI-Spotlight}: Adaptive Iterative Focus Refinement for Enhanced GUI Visual Grounding", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-Spotlight trains a grounding model to reason over images and call tools that iteratively narrow its visual focus. The paper evaluates this refinement method on ScreenSpot-Pro after training on 18,500 samples.", "topics": ["grounding", "planning", "tool-integration"], "updated_at": "2025-10-05", "url": "https://arxiv.org/abs/2510.04039"}
{"aliases": [], "authors": ["Yuan Wang", "Mingyu Li", "Haibo Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1c22547bf82eb2fb87e8bf6a3846d25c17cac296763726ed9a25faf88bdca33c", "submitted_at": "2025-10-06T09:14:58Z", "updated_at": "2026-03-25T11:29:08Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.04607v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "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:2510.04607", "identifiers": {"arxiv": "2510.04607"}, "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/2510.04607"}], "name": "From Imperative to Declarative: Towards LLM-friendly OS Interfaces for Boosted Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces a Declarative Model Interface that maps existing GUIs to access, state and observation primitives while handling low-level interactions. Evaluation uses Word, PowerPoint and Excel on Windows with a GUI-agent baseline.", "topics": ["automation", "tool-integration", "planning"], "updated_at": "2026-03-25", "url": "https://arxiv.org/abs/2510.04607"}
{"aliases": [], "authors": ["Yuhao Yang", "Zhen Yang", "Zi-Yi Dou", "Anh Nguyen", "Keen You", "Omar Attia", "Andrew Szot", "Michael Feng", "Ram Ramrakhya", "Alexander Toshev", "Chao Huang", "Yinfei Yang", "Zhe Gan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "33569f6ccca9c7e2675796191124424e76636828179c300d35da01ad9ccef1ce", "submitted_at": "2025-10-20T17:48:26Z", "updated_at": "2026-05-26T11:29:37Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2510.17790v3"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%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:2510.17790", "identifiers": {"arxiv": "2510.17790"}, "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/2510.17790"}], "name": "UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Action", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "UltraCUA trains a shared policy for primitive GUI actions and high-level tool calls using synthesized tasks and hybrid trajectories. Supervised fine-tuning followed by online reinforcement learning is evaluated on OSWorld and WindowsAgentArena.", "topics": ["tool-integration", "reinforcement-learning", "data-generation"], "updated_at": "2026-05-26", "url": "https://arxiv.org/abs/2510.17790"}
{"aliases": [], "authors": ["Hongrui Jia", "Jitong Liao", "Xi Zhang", "Haiyang Xu", "Tianbao Xie", "Chaoya Jiang", "Ming Yan", "Si Liu", "Wei Ye", "Fei Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d115356f723746ce95c3e40fed933786a0747b3b560c8f73a6b7af05ca22d24f", "submitted_at": "2025-10-28T15:56:36Z", "updated_at": "2025-11-11T04:10:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.24563v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.24563", "identifiers": {"arxiv": "2510.24563"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.24563"}], "name": "OSWorld-MCP: Benchmarking MCP Tool Invocation In Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "OSWorld-MCP evaluates how agents choose between GUI operations and MCP tool calls in computer tasks. It combines generated and curated tools, with 158 manually validated tools across seven applications, to examine tool use alongside task completion.", "topics": ["evaluation", "tool-integration", "planning"], "updated_at": "2025-11-11", "url": "https://arxiv.org/abs/2510.24563"}
{"aliases": [], "authors": ["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": ["Haolong Yan", "Jia Wang", "Xin Huang", "Yeqing Shen", "Ziyang Meng", "Zhimin Fan", "Kaijun Tan", "Jin Gao", "Lieyu Shi", "Mi Yang", "Shiliang Yang", "Zhirui Wang", "Brian Li", "Kang An", "Chenyang Li", "Lei Lei", "Mengmeng Duan", "Danxun Liang", "Guodong Liu", "Hang Cheng", "Hao Wu", "Jie Dong", "Junhao Huang", "Mei Chen", "Renjie Yu", "Shunshan Li", "Xu Zhou", "Yiting Dai", "Yineng Deng", "Yingdan Liang", "Zelin Chen", "Wen Sun", "Chengxu Yan", "Chunqin Xu", "Dong Li", "Fengqiong Xiao", "Guanghao Fan", "Guopeng Li", "Guozhen Peng", "Hongbing Li", "Hang Li", "Hongming Chen", "Jingjing Xie", "Jianyong Li", "Jingyang Zhang", "Jiaju Ren", "Jiayu Yuan", "Jianpeng Yin", "Kai Cao", "Liang Zhao", "Liguo Tan", "Liying Shi", "Mengqiang Ren", "Min Xu", "Manjiao Liu", "Mao Luo", "Mingxin Wan", "Na Wang", "Nan Wu", "Ning Wang", "Peiyao Ma", "Qingzhou Zhang", "Qiao Wang", "Qinlin Zeng", "Qiong Gao", "Qiongyao Li", "Shangwu Zhong", "Shuli Gao", "Shaofan Liu", "Shisi Gao", "Shuang Luo", "Xingbin Liu", "Xiaojia Liu", "Xiaojie Hou", "Xin Liu", "Xuanti Feng", "Xuedan Cai", "Xuan Wen", "Xianwei Zhu", "Xin Liang", "Xin Liu", "Xin Zhou", "Yifan Sui", "Yingxiu Zhao", "Yukang Shi", "Yunfang Xu", "Yuqing Zeng", "Yixun Zhang", "Zejia Weng", "Zhonghao Yan", "Zhiguo Huang", "Zhuoyu Wang", "Zihan Yan", "Zheng Ge", "Jing Li", "Yibo Zhu", "Binxing Jiao", "Xiangyu Zhang", "Daxin Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2fc824bdeabca3451fd895447a0321c507da1ac7bc7cc24184c7768992549daf", "submitted_at": "2025-12-17T13:26:30Z", "updated_at": "2025-12-19T17:36:21Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2512.15431v2"}}, "environments": ["android", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.15431", "identifiers": {"arxiv": "2512.15431"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.15431"}], "name": "Step-GUI Technical Report", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Step-GUI uses calibrated trajectory rewards to turn generated GUI interaction data into training signals for compact agent models. The report also introduces GUI-MCP for hierarchical GUI operations and local task delegation, and AndroidDaily for static-action and end-to-end mobile evaluation.", "topics": ["data-generation", "reward-models", "tool-integration", "evaluation", "mobile-navigation"], "updated_at": "2025-12-19", "url": "https://arxiv.org/abs/2512.15431"}
{"aliases": [], "authors": ["Quyu Kong", "Xu Zhang", "Zhenyu Yang", "Nolan Gao", "Chen Liu", "Panrong Tong", "Chenglin Cai", "Hanzhang Zhou", "Jianan Zhang", "Liangyu Chen", "Zhidan Liu", "Steven Hoi", "Yue Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cf3258fd2286b0ee6babb69b9bfc459ecf40536eec498a3af3403d2971791f3d", "submitted_at": "2025-12-22T14:31:28Z", "updated_at": "2025-12-30T07:15:15Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2512.19432v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.19432", "identifiers": {"arxiv": "2512.19432"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.19432"}], "name": "MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive and MCP-Augmented Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces long mobile workflows that combine agent-user interaction, cross-application tasks, and MCP-based tool access.", "topics": ["evaluation", "tool-integration", "long-horizon"], "updated_at": "2025-12-30", "url": "https://arxiv.org/abs/2512.19432"}
{"aliases": [], "authors": ["Hanzhang Zhou", "Xu Zhang", "Panrong Tong", "Jianan Zhang", "Liangyu Chen", "Quyu Kong", "Chenglin Cai", "Chen Liu", "Yue Wang", "Jingren Zhou", "Steven Hoi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a89eb3fd173680c127245933d252d4e245b7179d77bca4b463d50800c78f7238", "submitted_at": "2025-12-26T14:51:52Z", "updated_at": "2025-12-26T14:51:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.22047v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.22047", "identifiers": {"arxiv": "2512.22047"}, "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/2512.22047"}], "name": "MAI-UI Technical Report: Real-World Centric Foundation GUI Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAI-UI combines GUI navigation, native user interaction, and MCP tool calls in a family of policies trained through generated data and online reinforcement learning. The report also studies device-cloud routing and scaling of parallel environments and interaction budgets, with grounding and Android navigation evaluation.", "topics": ["human-in-the-loop", "tool-integration", "reinforcement-learning", "data-generation", "efficiency"], "updated_at": "2025-12-26", "url": "https://arxiv.org/abs/2512.22047"}
{"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": ["Bowen Yang", "Kaiming Jin", "Zhenyu Wu", "Zhaoyang Liu", "Qiushi Sun", "Zehao Li", "JingJing Xie", "Zhoumianze Liu", "Fangzhi Xu", "Kanzhi Cheng", "Qingyun Li", "Yian Wang", "Yu Qiao", "Zun Wang", "Zichen Ding"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cfdaabaf3c6df80eb718e51dc666c54fb2ecf38ec520f6a71ccf5e5ce54248c8", "submitted_at": "2026-01-12T17:55:51Z", "updated_at": "2026-01-12T17:55:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.07779v1"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "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:2601.07779", "identifiers": {"arxiv": "2601.07779"}, "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/2601.07779"}], "name": "OS-Symphony: A Holistic Framework for Robust and Generalist Computer-Using Agent", "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": "OS-Symphony coordinates milestone-based reflection and memory with tools that build visually aligned tutorials through browser interaction. The framework evaluates trajectory correction and adaptation on online computer-use benchmarks including OSWorld.", "topics": ["memory", "self-reflection", "planning", "tool-integration"], "updated_at": "2026-01-12", "url": "https://arxiv.org/abs/2601.07779"}
{"aliases": [], "authors": ["Chen Chen", "Jiawei Shao", "Dakuan Lu", "Haoyi Hu", "Xiangcheng Liu", "Hantao Yao", "Wu Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0d5ed7e72534fc55d028d4f25cb98b2373c88b622cf76396933d7a49d8ecf2e8", "submitted_at": "2026-01-14T14:27:28Z", "updated_at": "2026-01-14T14:27:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.09770v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.09770", "identifiers": {"arxiv": "2601.09770"}, "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/2601.09770"}], "name": "GUI-Eyes: Tool-Augmented Perception for Visual Grounding in GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-Eyes trains GUI grounders to decide when and how to use visual tools such as crops and zooms. Its coarse-to-fine policy uses dense spatial rewards based on location proximity and region overlap, with evaluation on ScreenSpot-Pro.", "topics": ["grounding", "reinforcement-learning", "tool-integration", "screen-understanding"], "updated_at": "2026-01-14", "url": "https://arxiv.org/abs/2601.09770"}
{"aliases": [], "authors": ["Tianyi Chen", "Yinheng Li", "Michael Solodko", "Sen Wang", "Nan Jiang", "Tingyuan Cui", "Junheng Hao", "Jongwoo Ko", "Sara Abdali", "Leon Xu", "Suzhen Zheng", "Hao Fan", "Pashmina Cameron", "Justin Wagle", "Kazuhito Koishida"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5a3efa3cd0add437e327ea49896a3ee794daa178664dfb4096aba66fed27554b", "submitted_at": "2026-01-28T23:38:25Z", "updated_at": "2026-02-02T23:11:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.21123v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.21123", "identifiers": {"arxiv": "2601.21123"}, "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/2601.21123"}], "name": "CUA-Skill: Develop Skills for Computer Using Agent", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "CUA-Skill encodes Windows application knowledge in parameterized skills and execution/composition graphs. Its agent retrieves and instantiates these skills and uses memory for failure recovery, with end-to-end evaluation on WindowsAgentArena.", "topics": ["tool-integration", "planning", "memory", "automation"], "updated_at": "2026-02-02", "url": "https://arxiv.org/abs/2601.21123"}
{"aliases": [], "authors": ["Chaoqun Cui", "Jing Huang", "Shijing Wang", "Liming Zheng", "Qingchao Kong", "Zhixiong Zeng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "86225f2e52e1394e2a0a56b1e82327d756bff95739e01ce7cc5898af49944aba", "submitted_at": "2026-01-31T07:36:54Z", "updated_at": "2026-07-25T06:54:35Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.00575v2"}}, "environments": ["desktop", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.00575", "identifiers": {"arxiv": "2602.00575"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.00575"}], "name": "Agentic Reward Modeling: Verifying GUI Agent via Progressive Trajectory-Grounded Interaction", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "VAGEN verifies GUI trajectories by first extracting readily available evidence and escalating to tool-based environment probes when needed. Evaluation on OSWorld-Verified and AndroidWorld studies verification accuracy and its cost.", "topics": ["reward-models", "tool-integration", "evaluation", "efficiency"], "updated_at": "2026-07-25", "url": "https://arxiv.org/abs/2602.00575"}
{"aliases": [], "authors": ["Xiaoce Wang", "Guibin Zhang", "Junzhe Li", "Jinzhe Tu", "Chun Li", "Ming Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "49b0dbff31dcf8c966bebd367ae05984e46a7fd13b2cfb1ad5ef061332b43f00", "submitted_at": "2026-01-30T08:38:05Z", "updated_at": "2026-01-30T08:38:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.02548v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.02548", "identifiers": {"arxiv": "2602.02548"}, "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/2602.02548"}], "name": "ToolTok: Tool Tokenization for Efficient and Generalizable GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "ToolTok represents GUI operations with learned tool tokens and frames interaction as progressive tool selection. Semantic anchoring and an easy-to-hard curriculum teach tool meanings before visual pathfinding, with evaluation across grounding scenarios.", "topics": ["tool-integration", "grounding", "data-generation"], "updated_at": "2026-01-30", "url": "https://arxiv.org/abs/2602.02548"}
{"aliases": [], "authors": ["Renjun Xu", "Yang Yan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f0edc70265a8d46a7c22a9c6384de8aa19f79dbb6987f8c23f4a260647a6a3f0", "submitted_at": "2026-02-12T21:33:25Z", "updated_at": "2026-06-02T16:14:54Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2602.12430v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}], "id": "arxiv:2602.12430", "identifiers": {"arxiv": "2602.12430"}, "kind": "survey", "last_verified_at": "2026-10-17T12:31:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.12430"}], "name": "Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent survey of the skill layer: CUA deployment and GUI grounding are explicit parts of the review, but its overall scope extends well beyond interface agents.", "published_at": "2026-02-12", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "This survey organizes agent skills by architecture, acquisition, deployment and security, including their role in computer-use systems and GUI grounding. It also proposes a permission and lifecycle governance framework for skill deployment.", "topics": ["survey", "tool-integration", "safety-security", "planning"], "updated_at": "2026-06-02", "url": "https://arxiv.org/abs/2602.12430"}
{"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": ["Haiyang Xu", "Xi Zhang", "Haowei Liu", "Junyang Wang", "Zhaozai Zhu", "Shengjie Zhou", "Xuhao Hu", "Feiyu Gao", "Junjie Cao", "Zihua Wang", "Zhiyuan Chen", "Jitong Liao", "Qi Zheng", "Jiahui Zeng", "Ze Xu", "Shuai Bai", "Junyang Lin", "Jingren Zhou", "Ming Yan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "eeaddd8a32bdd4b53356ca43e849ea3697ac3d40a05b32e6c930dc445c243898", "submitted_at": "2026-02-15T01:52:19Z", "updated_at": "2026-02-15T01:52:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.16855v1"}}, "environments": ["cross-platform"], "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", "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=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.16855", "identifiers": {"arxiv": "2602.16855"}, "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/2602.16855"}], "name": "Mobile-Agent-v3.5: Multi-platform Fundamental GUI Agents", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2026-02-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents GUI-Owl-1.5 variants and a hybrid data pipeline for multimodal GUI interaction, tool use, memory, and multi-platform tasks.", "topics": ["tool-integration", "memory", "reinforcement-learning"], "updated_at": "2026-02-15", "url": "https://arxiv.org/abs/2602.16855"}
{"aliases": [], "authors": ["Alexandre De Masi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5944045ee4562641ee02612969d1c2d038e019fa02cc8d48c671c1d5397557ad", "submitted_at": "2026-03-11T11:24:31Z", "updated_at": "2026-03-11T11:24:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.10664v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}], "id": "arxiv:2603.10664", "identifiers": {"arxiv": "2603.10664"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:16Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.10664"}], "name": "Terminal Is All You Need: Design Properties for Human-AI Agent Collaboration", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent conceptual HCI work about agent-facing interfaces; the abstract does not report GUI task experiments or a new evaluated CUA.", "published_at": "2026-03-11", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "This position paper derives interface design principles from terminal-based human-agent collaboration: representational compatibility, transparent actions and low participation barriers. It argues that graphical and other agent-facing modalities should provide these properties explicitly.", "topics": ["human-in-the-loop", "ui-design", "observability", "tool-integration"], "updated_at": "2026-03-11", "url": "https://arxiv.org/abs/2603.10664"}
{"aliases": [], "authors": ["Tiantian He", "Yihang Chen", "Keyue Jiang", "Ka Yiu Lee", "Kaiwen Zhou", "Kun Shao", "Shuai Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dd2bc0ee1955dbc8e98a9306a3d6260251cbab89d36ac117e4dc49b6fa99766e", "submitted_at": "2026-04-10T18:46:05Z", "updated_at": "2026-04-10T18:46:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.09815v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.09815", "identifiers": {"arxiv": "2604.09815"}, "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/2604.09815"}], "name": "EE-MCP: Self-Evolving MCP-GUI Agents via Automated Environment Generation and Experience 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": "2026-04-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "EE-MCP studies application-dependent mixtures of GUI actions and MCP calls using generated environments, trajectory collection and filtered training. An experience bank extracts rules from trajectory comparisons for inference-time improvement, distinct from model distillation.", "topics": ["tool-integration", "data-generation", "imitation-learning", "memory"], "updated_at": "2026-04-10", "url": "https://arxiv.org/abs/2604.09815"}
{"aliases": [], "authors": ["Fei Tang", "Zhiqiong Lu", "Boxuan Zhang", "Weiming Lu", "Jun Xiao", "Yueting Zhuang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "df56b38203bc5ec8d7e871eb2ad9fca9891a049df62b7d72796fe400eb8bce5e", "submitted_at": "2026-04-13T17:52:04Z", "updated_at": "2026-04-13T17:52:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.11784v1"}}, "environments": ["android", "harmonyos", "ios"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.11784", "identifiers": {"arxiv": "2604.11784"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.11784"}], "name": "ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents ClawGUI, a shared harness for GUI-agent reinforcement learning, benchmark evaluation and deployment on mobile devices. It combines parallel virtual and physical environments, process rewards, and hybrid CLI-GUI execution with persistent memory.", "topics": ["reinforcement-learning", "evaluation", "tool-integration", "mobile-navigation"], "updated_at": "2026-04-13", "url": "https://arxiv.org/abs/2604.11784"}
{"aliases": [], "authors": ["Zhengxi Lu", "Fei Tang", "Guangyi Liu", "Kaitao Song", "Xu Tan", "Jin Ma", "Wenqi Zhang", "Weiming Lu", "Jun Xiao", "Yueting Zhuang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d6e81b448c9c22e2fd25dce519db3fdc0585b0265b94d448751d7a37ed910dfa", "submitted_at": "2026-04-15T12:55:46Z", "updated_at": "2026-04-15T12:55:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.13822v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2604.13822", "identifiers": {"arxiv": "2604.13822"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.13822"}], "name": "UI-Copilot: Advancing Long-Horizon GUI Automation via Tool-Integrated Policy Optimization", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-Copilot separates GUI execution from on-demand memory retrieval and numerical assistance. Its training procedure treats tool selection and multi-turn task execution as distinct optimization targets while keeping persistent observations outside transient context.", "topics": ["tool-integration", "memory", "planning", "long-horizon", "mobile-navigation"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2604.13822"}
{"aliases": [], "authors": ["Haomin Zhuang", "Hanwen Xing", "Xiangliang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7b1788e6fbb4c14a6d993ce329bf7bf05e3a1e6ffbfc7c14a9fa9d36fbf7ef16", "submitted_at": "2026-04-15T18:22:50Z", "updated_at": "2026-04-15T18:22:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.16520v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2604.16520", "identifiers": {"arxiv": "2604.16520"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.16520"}], "name": "AgentClick: A Skill-Based Human-in-the-Loop Review Layer for Terminal AI Agents", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent supervision interface: the browser UI is for human review of terminal agents; the abstract does not claim that those agents operate graphical interfaces.", "published_at": "2026-04-15", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "AgentClick gives terminal-based agents a browser review interface for plans, drafts, execution traces and editable memory. A local server and skill plugin let users inspect proposed actions and provide structured feedback.", "topics": ["human-in-the-loop", "observability", "workflow", "memory", "tool-integration"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2604.16520"}
{"aliases": [], "authors": ["Qijun Han", "Haoqin Tu", "Zijun Wang", "Haoyue Dai", "Yiyang Zhou", "Nancy Lau", "Alvaro A. Cardenas", "Yuhui Xu", "Ran Xu", "Caiming Xiong", "Zeyu Zheng", "Huaxiu Yao", "Yuyin Zhou", "Cihang Xie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7e675476121ab3ef51a8627ccfcc86d76eddde73987ea882398f72e3cb31c5d9", "submitted_at": "2026-04-23T07:42:37Z", "updated_at": "2026-09-24T23:58:02Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.21375v3"}}, "environments": ["desktop", "linux", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:02Z", "role": "primary", "sha256": "e436adf2735cb537ee4b24aa0c3f36930a50c2f683cc6868e7c40df602bdb6b8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2604.00892%2C2604.01221%2C2604.01676%2C2604.02345%2C2604.03465%2C2604.05157%2C2604.06134%2C2604.06240%2C2604.06995%2C2604.07277%2C2604.07831%2C2604.07929%2C2604.08005%2C2604.09581%2C2604.10261%2C2604.10577%2C2604.10988%2C2604.11201%2C2604.18847%2C2604.21375%2C2604.24441%2C2604.26020%2C2605.00551%2C2605.02168%2C2605.07505%2C2605.10347%2C2605.12294%2C2605.15040%2C2605.16883%2C2605.27134%2C2605.28258%2C2606.01790%2C2606.02031%2C2606.02208%2C2606.03203%2C2606.03236%2C2606.03951%2C2606.04391%2C2606.04627%2C2606.04701%2C2606.05233%2C2606.05597%2C2606.05697%2C2606.06560%2C2606.06708%2C2606.07027%2C2606.07074%2C2606.09399%2C2606.09447%2C2606.09764&max_results=50"}], "id": "arxiv:2604.21375", "identifiers": {"arxiv": "2604.21375"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:02Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.21375"}], "name": "VLAA-GUI: Knowing When to Stop, Recover, and Search, A Modular Framework for GUI Automation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "VLAA-GUI combines mandatory checks for visually supported completion, multi-tier loop recovery and on-demand workflow search. Optional coding and grounding agents support execution, with component ablations on Linux and Windows GUI tasks.", "topics": ["robustness", "self-reflection", "tool-integration", "evaluation"], "updated_at": "2026-09-24", "url": "https://arxiv.org/abs/2604.21375"}
{"aliases": [], "authors": ["Jackson Vonderhorst", "Kuangshi Ai", "Haichao Miao", "Shusen Liu", "Chaoli Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8e7b07a811df7e74a28f91a5b6c59c69b7ba066bd6c64452a299e678a3187733", "submitted_at": "2026-04-30T15:22:28Z", "updated_at": "2026-06-26T20:31:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.27996v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2604.27996", "identifiers": {"arxiv": "2604.27996"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.27996"}], "name": "Exploring LLM Agent Designs and Interaction Modalities for Scientific Visualization", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-04-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study compares specialized, computer-use and coding agents on scientific visualization workflows. It examines GUI, CLI, code and tool-call interaction modes, including how persistent memory changes performance across repeated trials.", "topics": ["science", "evaluation", "tool-integration", "memory"], "updated_at": "2026-06-26", "url": "https://arxiv.org/abs/2604.27996"}
{"aliases": [], "authors": ["Chiyeong Heo", "Jaechang Kim", "Junhyuk Kwon", "Hoyoung Kim", "Dongmin Park", "Jonghyun Lee", "Jungseul Ok"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0bd75e6bf9d5a8185a2204cdd99d42f9d96c102f7297abcadb9f8425e9246349", "submitted_at": "2026-05-08T10:57:19Z", "updated_at": "2026-05-08T10:57:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.10966v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2605.10966", "identifiers": {"arxiv": "2605.10966"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.10966"}], "name": "MMTB: Evaluating Terminal Agents on Multimedia-File Tasks", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified. Adjacent computer-automation benchmark using terminal workflows, explicitly separated from graphical-interface control.", "published_at": "2026-05-08", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "MultiMedia-TerminalBench evaluates agents that use terminal tools to work with audio and video files. The Terminus-MM harness adds multimedia perception to study how accessible evidence changes executable workflows and task outcomes.", "topics": ["evaluation", "automation", "video", "tool-integration"], "updated_at": "2026-05-08", "url": "https://arxiv.org/abs/2605.10966"}
{"aliases": [], "authors": ["Xuhao Hu", "Xi Zhang", "Haiyang Xu", "Kyle Qiao", "Jingyi Yang", "Xuanjing Huang", "Jing Shao", "Ming Yan", "Jieping Ye"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "09128e90b925ac0cd2cf6e967853180bbef1e97a0960c22065f1b0b50c30d58b", "submitted_at": "2026-05-12T17:57:04Z", "updated_at": "2026-09-19T14:53:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.12481v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.12481", "identifiers": {"arxiv": "2605.12481"}, "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/2605.12481"}], "name": "ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-05-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "ToolCUA synthesizes interleaved GUI/tool trajectories and trains agents to choose between graphical actions and higher-level tools. Staged supervised and reinforcement learning culminate in online hybrid-environment training with a reward for efficient execution paths.", "topics": ["tool-integration", "data-generation", "imitation-learning", "reinforcement-learning", "planning", "efficiency"], "updated_at": "2026-09-19", "url": "https://arxiv.org/abs/2605.12481"}
{"aliases": [], "authors": ["Julien Piet", "Annabella Chow", "Yiwei Hou", "Muxi Lyu", "Sylvie Venuto", "Jinhao Zhu", "Raluca Ada Popa", "David Wagner"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "572be49190554bbfb6accac4c87ef91608f88e90d06250f229db2922e19fcd56", "submitted_at": "2026-05-14T02:48:57Z", "updated_at": "2026-05-14T02:48:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.14290v1"}}, "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:2605.14290", "identifiers": {"arxiv": "2605.14290"}, "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/2605.14290"}], "name": "Web Agents Should Adopt the Plan-Then-Execute Paradigm", "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. Directly analyzes browser-agent prompt-injection boundaries and WebArena task structure, but advocates typed website operations rather than demonstrating a new GUI-control policy.", "published_at": "2026-05-14", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Argues for compiling web tasks into predefined execution graphs before exposing an agent to untrusted runtime content. Its WebArena task analysis identifies typed, semantically complete website tools as a prerequisite and distinguishes this interface proposal from reactive click-level control.", "topics": ["planning", "safety-security", "tool-integration"], "updated_at": "2026-05-14", "url": "https://arxiv.org/abs/2605.14290"}
{"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": ["Yuhao Yang", "Tianyu Fan", "Chao Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f7490150e82a5abfc510c60401cccf21c821d829e66cc5f2b40fd2fbbf4b0df4", "submitted_at": "2026-06-02T16:30:33Z", "updated_at": "2026-06-02T16:30:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.03854v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.03854", "identifiers": {"arxiv": "2606.03854"}, "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/2606.03854"}], "name": "CLI-Anything: Towards Agent-Native Computer Use", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced. Adjacent computer-interface architecture: the described application command harnesses concern computer operation, but no GUI execution or comparative benchmark result is established in the abstract.", "published_at": "2026-06-02", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "CLI-Anything argues for application interfaces built around structured commands, explicit state and execution feedback. It describes transforming existing software into command-line harnesses and organizing that approach through CLI-Hub.", "topics": ["automation", "tool-integration", "workflow"], "updated_at": "2026-06-02", "url": "https://arxiv.org/abs/2606.03854"}
{"aliases": [], "authors": ["Wanli Li", "Bowen Zhou", "Yunyao Yu", "Zhou Xu", "Yifan Yang", "Dongsheng Li", "Caihua Shan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73b751efd63d98fc81166a1a80cb0c162d68bf1ec4ecca0a1a1a7717ad62e680", "submitted_at": "2026-06-08T12:39:23Z", "updated_at": "2026-07-06T14:25:02Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2606.09426v3"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.09426", "identifiers": {"arxiv": "2606.09426"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.09426"}], "name": "WeaveBench: A Long-Horizon, Real-World Benchmark for Computer-Use Agents with Hybrid Interfaces", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "WeaveBench requires agents to combine GUI, command-line and code operations within long tasks on an Ubuntu desktop. A trajectory-aware judge checks deliverables and interaction evidence, including shortcut behavior that outcome-only grading can miss.", "topics": ["evaluation", "long-horizon", "workflow", "tool-integration"], "updated_at": "2026-07-06", "url": "https://arxiv.org/abs/2606.09426"}
{"aliases": [], "authors": ["Jiaxin Ai", "Tao Hu", "Xuemeng Yang", "Shu Zou", "Hairong Zhang", "Daocheng Fu", "Yu Yang", "Hongbin Zhou", "Nianchen Deng", "Pinlong Cai", "Zhongyuan Wang", "Botian Shi", "Kaipeng Zhang", "Licheng Wen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "edf7e63803efba89551aa297bff1da9b0babf8ec898a2971fa6c8a20cc58dea8", "submitted_at": "2026-06-11T11:53:32Z", "updated_at": "2026-06-30T03:31:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.13239v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.13239", "identifiers": {"arxiv": "2606.13239"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.13239"}], "name": "ComAct: Reframing Professional Software Manipulation via COM-as-Action Paradigm", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-11", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "ComAct proposes COM program synthesis as an alternative interface for manipulating professional software and introduces ComCADBench with GUI-based comparison agents. ComActor adds a staged training and correction approach, supported by a Windows-container training platform.", "topics": ["automation", "evaluation", "tool-integration", "science"], "updated_at": "2026-06-30", "url": "https://arxiv.org/abs/2606.13239"}
{"aliases": [], "authors": ["Chenxin Li", "Zhengyao Fang", "Zhengyang Tang", "Pengyuan Lyu", "Xingran Zhou", "Xin Lai", "Fei Tang", "Liang Wu", "Yiduo Guo", "Weinong Wang", "Junyi Li", "Yi Zhang", "Yang Ding", "Huawen Shen", "Sunqi Fan", "Shangpin Peng", "Zheng Ruan", "Anran Zhang", "Benyou Wang", "Chengquan Zhang", "Han Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "280ec12bcb246f1f07cf3627011945e8a9ff4afed93b1fb2937f2d6b1bc1323b", "submitted_at": "2026-06-12T15:01:32Z", "updated_at": "2026-06-12T15:01:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.14832v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.14832", "identifiers": {"arxiv": "2606.14832"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.14832"}], "name": "PhoneHarness: Harnessing Phone-Use Agents through Mixed GUI, CLI, and Tool Actions", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "PhoneHarness routes phone workflows across GUI actions, device-side commands and host tools while retaining auditable execution traces. Its accompanying benchmark checks observable task side effects, separating mixed-action execution from plausible final-answer generation.", "topics": ["mobile-navigation", "tool-integration", "evaluation", "workflow"], "updated_at": "2026-06-12", "url": "https://arxiv.org/abs/2606.14832"}
{"aliases": [], "authors": ["Ziyan Jiang", "Li An", "Yujian Liu", "Jiabao Ji", "Qiucheng Wu", "Jacob Andreas", "Yang Zhang", "Shiyu Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2605cedb4359aa0cf1e694455903871ed334dd37b7b8e15edf31bac9d5e51b13", "submitted_at": "2026-06-16T19:57:07Z", "updated_at": "2026-06-16T19:57:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.18448v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.18448", "identifiers": {"arxiv": "2606.18448"}, "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/2606.18448"}], "name": "VISUALSKILL: Multimodal Skills for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "VISUALSKILL organizes application-specific text and figures into hierarchical skills retrieved on demand through an MCP tool. Skills combine authored documentation with live UI exploration, and matched multimodal/text-only comparisons test their utility for interface identification and workflow-state verification.", "topics": ["memory", "tool-integration", "screen-understanding", "long-horizon"], "updated_at": "2026-06-16", "url": "https://arxiv.org/abs/2606.18448"}
{"aliases": [], "authors": ["Li Gu", "Zihuan Jiang", "Linqiang Guo", "Zhixiang Chi", "Ziqiang Wang", "Huan Liu", "Yuanhao Yu", "Tse-Hsun Chen", "Yang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9bf8b07e08b275d90fcbfc63d867f54f5579509132cc9cf694d01447b7f91b78", "submitted_at": "2026-06-16T02:36:22Z", "updated_at": "2026-06-16T02:36:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.19388v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.19388", "identifiers": {"arxiv": "2606.19388"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.19388"}], "name": "Beyond the GUI Paradigm: Do Mobile Agents Need the Phone Screen?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-16", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "This study compares general coding agents using mobile command-line access with reproducible GUI agents on AndroidWorld and MobileWorld. It also supplies oracle CLI analyses and a separate task suite for operations such as bulk changes, filtering and hidden device state.", "topics": ["evaluation", "automation", "cross-app", "tool-integration"], "updated_at": "2026-06-16", "url": "https://arxiv.org/abs/2606.19388"}
{"aliases": [], "authors": ["Yikun Fu", "Bowen Fu", "Zhenyu Wu", "Shuang Cheng", "Xiaowei Sun", "Bowen Yang", "Zehao Li", "Yibo Zhao", "Zichen Ding", "Zhoumianze Liu", "Shijie Wang", "Biqing Qi", "Bowen Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6dc52f58bb2237d12041fa8b994b20bd1e41290aaab0c70d2c5dd79fcaca37b6", "submitted_at": "2026-06-21T15:32:57Z", "updated_at": "2026-06-21T15:32:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.22557v1"}}, "environments": ["desktop", "macos"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}], "id": "arxiv:2606.22557", "identifiers": {"arxiv": "2606.22557"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.22557"}], "name": "MacAgentBench: Benchmarking AI Agents on Real-World macOS Desktop", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces macOS tasks spanning GUI and command-line interaction, with deterministic checks and intermediate-progress scoring for application workflows.", "topics": ["evaluation", "cross-app", "tool-integration"], "updated_at": "2026-06-21", "url": "https://arxiv.org/abs/2606.22557"}
{"aliases": [], "authors": ["Xiao Zhou", "Siyue Zhang", "Yilun Zhao", "Jinbiao Wei", "Tingyu Song", "Arman Cohan", "Chen Zhao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "77f777f05b1c169ee4ad7ae7d77e9d1fc80a0696e9a1f6eefb5028a5c3e6fe5d", "submitted_at": "2026-06-22T17:05:07Z", "updated_at": "2026-06-22T17:05:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.24551v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.24551", "identifiers": {"arxiv": "2606.24551"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.24551"}], "name": "GUI vs. CLI: Execution Bottlenecks in Screen-Only and Skill-Mediated Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "This benchmark matches desktop goals, initial states and final-state verifiers across screen-only GUI and skill-mediated CLI agents. Its experiments separate graphical execution failures from missing CLI skill coverage, including a verifier-guided skill-augmentation condition.", "topics": ["evaluation", "tool-integration", "robustness", "workflow"], "updated_at": "2026-06-22", "url": "https://arxiv.org/abs/2606.24551"}
{"aliases": [], "authors": ["Yogeswar Reddy Thota"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6f3f2ae1438986c11146fec508dd61fd3e848ddcbf830f86a7dfa759280d730a", "submitted_at": "2026-06-29T03:19:40Z", "updated_at": "2026-06-29T03:19:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.30697v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.30697", "identifiers": {"arxiv": "2606.30697"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.30697"}], "name": "LUMOS: A Semantic Operating-System Layer for Accessibility-Grounded AI 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": "2026-06-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "LUMOS converts native accessibility metadata and browser UI structure into semantic interface blueprints with stable element identifiers and available actions. It combines live pointer-to-element grounding with an observe-act loop constrained to visible-UI primitives.", "topics": ["grounding", "screen-understanding", "tool-integration", "automation"], "updated_at": "2026-06-29", "url": "https://arxiv.org/abs/2606.30697"}
{"aliases": [], "authors": ["Yong Liu", "Zhenyi Zhong", "Zhanpeng Shi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a779b2c1ff9ba3b531d399204b0986e3affcea81c67fdd7951e3b7da343b0e17", "submitted_at": "2026-07-16T00:35:25Z", "updated_at": "2026-07-16T00:35:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.14443v1"}}, "environments": ["desktop", "macos"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.14443", "identifiers": {"arxiv": "2607.14443"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.14443"}], "name": "Tactile: Giving Computer-Using Agents Hands and Feet", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-07-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Tactile turns accessibility, OCR and visual evidence into explicit desktop targets with executable actions and verification cues. Its execution loop chooses semantic actions or coordinate fallbacks while retaining provenance for replay and diagnosis.", "topics": ["grounding", "automation", "observability", "tool-integration", "robustness", "evaluation"], "updated_at": "2026-07-16", "url": "https://arxiv.org/abs/2607.14443"}
{"aliases": [], "authors": ["Xiao Yu", "Baolin Peng", "Ruize Xu", "Hao Zou", "Qianhui Wu", "Hao Cheng", "Wenlin Yao", "Nikhil Singh", "Zhou Yu", "Jianfeng Gao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "965fe5aabfb44b4c82290fc347a463eb58fb2be7a3761496efd38167d0f7d5b6", "submitted_at": "2026-07-23T17:38:30Z", "updated_at": "2026-08-07T17:40:50Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2607.21557v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}], "id": "arxiv:2607.21557", "identifiers": {"arxiv": "2607.21557"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:09Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.21557"}], "name": "OpenForgeRL: Train Harness-native Agents in Any Environment", "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": "2026-07-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "OpenForgeRL connects stateful agent harnesses to reinforcement-learning infrastructure through a model-call recording proxy and isolated rollout containers. Its validation includes browser and computer-use agents, alongside tool-oriented agents, and examines how harness choice affects learned reliability.", "topics": ["reinforcement-learning", "sandbox", "tool-integration", "robustness"], "updated_at": "2026-08-07", "url": "https://arxiv.org/abs/2607.21557"}
{"aliases": [], "authors": ["Yan Yang", "Xiangru Jian", "Ziyang Luo", "Zirui Zhao", "Yutong Dai", "Ziji Shi", "Hanshu Yan", "Jun Hao Liew", "Silvio Savarese", "Junnan Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a010fbef09d8359550cf156bf0dcba873f25a0d6fb8f118d53bf6e23715cf203", "submitted_at": "2026-07-24T14:17:03Z", "updated_at": "2026-07-24T14:17:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.22798v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.22798", "identifiers": {"arxiv": "2607.22798"}, "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/2607.22798"}], "name": "StateAct: Program State, before Pixels, for Long-Horizon Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces StateAct, a code-first computer-use harness that manipulates program state directly and delegates screenshot-and-click subgoals to a GUI agent. An independent completion gate checks saved outputs, while fresh subagents handle subgoals to limit accumulated context.", "topics": ["tool-integration", "state-tracking", "long-horizon"], "updated_at": "2026-07-24", "url": "https://arxiv.org/abs/2607.22798"}
{"aliases": [], "authors": ["Chenrui Shi", "Yuwei Wu", "Yang Liu", "Ruining Feng", "Zirui Shang", "Zhi Gao", "Lifeng Fan", "Che Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d1688a528d35a8213a8027385b1b1327536567f6ec307aa6562eea543ee71d7f", "submitted_at": "2026-07-28T16:01:38Z", "updated_at": "2026-07-29T05:13:57Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.25904v2"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}], "id": "arxiv:2607.25904", "identifiers": {"arxiv": "2607.25904"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:09Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.25904"}], "name": "Interactive Reward Agent: GUI Task Evaluation via Environment-State Verification", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces an Interactive Reward Agent that proposes completion conditions and verifies them using system, application, and GUI tools after execution. GUI-RewardBench provides annotated desktop trajectories for evaluating these evidence-based judgments and their use in reinforcement learning.", "topics": ["reward-models", "evaluation", "tool-integration"], "updated_at": "2026-07-29", "url": "https://arxiv.org/abs/2607.25904"}
{"aliases": [], "authors": ["Hanzhang Zhou", "Panrong Tong", "Xu Zhang", "Quyu Kong", "Chenglin Cai", "Tianyu Xia", "Gongjie Zhang", "Jianan Zhang", "Long Li", "Long Chen", "Lei Wang", "Gaole Dai", "Pengxiang Li", "Liangyu Chen", "Yue Wang", "Steven Hoi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "52fb1b49501132f1dbe3564daa02cca22a93fc8d7c28dec8ee6070a23889574a", "submitted_at": "2026-07-30T13:58:41Z", "updated_at": "2026-07-30T13:58:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.28227v1"}}, "environments": ["cross-platform", "web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.28227", "identifiers": {"arxiv": "2607.28227"}, "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/2607.28227"}], "name": "Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a foundation GUI agent with combined GUI and command-line actions, large-scale interaction infrastructure, and iterative environment and task generation.", "topics": ["reinforcement-learning", "long-horizon", "tool-integration"], "updated_at": "2026-07-30", "url": "https://arxiv.org/abs/2607.28227"}
{"aliases": [], "authors": ["Siqi Fan", "Minghao Li", "Xiaoqian Ma", "Wenhui Tan", "Xiusheng Huang", "Juntong Wu", "Liujie Zhang", "Shuo Shang", "Weihang Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2c8692712194d0ff822091d13324e6bb540339103a824f1030f7a971699d0a0b", "submitted_at": "2026-08-04T08:35:51Z", "updated_at": "2026-08-06T07:26:24Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.03327v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.03327", "identifiers": {"arxiv": "2608.03327"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.03327"}], "name": "Screenshots or Tools? Eliciting Tool Use and Managing Multimodal Context in Hybrid GUI-MCP Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study separates tool-adoption behavior from task competence in hybrid GUI–MCP agents. It tests tool-use incentives and screenshot-history compression, showing why changed action preferences alone need not improve held-out accuracy.", "topics": ["tool-integration", "reinforcement-learning", "efficiency", "interaction-history", "evaluation"], "updated_at": "2026-08-06", "url": "https://arxiv.org/abs/2608.03327"}
{"aliases": [], "authors": ["Lin-Fa Lee", "YI-YU Chang", "Kuo-Hui Yeh"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1bddfe5b5b2312e9b71d01dbb6bb23a8b0ad1c8cd787e4e98449b71a2ac2d678", "submitted_at": "2026-08-25T03:16:45Z", "updated_at": "2026-08-25T03:16:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.24017v1"}}, "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:2608.24017", "identifiers": {"arxiv": "2608.24017"}, "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/2608.24017"}], "name": "WebMCP-Phalanx: Enforcing and Characterizing Trust Boundaries for Browser-Integrated 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": "2026-08-25", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Proposes WebMCP-Phalanx, a browser-integrated runtime that binds page-exposed tools to their registering principals and separates content inspection from privileged invocation. It studies lifecycle and prompt-injection attacks, including a timing bypass that motivates validating all visible metadata before tool use.", "topics": ["safety-security", "tool-integration"], "updated_at": "2026-08-25", "url": "https://arxiv.org/abs/2608.24017"}
{"aliases": [], "authors": ["Rui Xie", "Lu Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e1626d5190f145c668c9b7f363b7f84904c12daae485e63b858e440ded5ea6df", "submitted_at": "2026-08-27T11:38:19Z", "updated_at": "2026-08-27T11:38:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.26991v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:35Z", "role": "primary", "sha256": "6499b4f4917eec0765d16d6e26a08d7edc5146a30753b19cf8218517656bb070", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.16211%2C2608.16391%2C2608.17067%2C2608.17975%2C2608.19741%2C2608.21500%2C2608.22301%2C2608.22510%2C2608.23035%2C2608.23320%2C2608.26238%2C2608.26546%2C2608.26991%2C2608.27260%2C2608.29814%2C2608.31076%2C2609.02292%2C2609.04173%2C2609.04298%2C2609.05776%2C2609.06059%2C2609.07529%2C2609.08149%2C2609.08418%2C2609.09187%2C2609.11561%2C2609.12533%2C2609.15096%2C2609.15544%2C2609.20732%2C2609.21293%2C2609.23986%2C2609.24971%2C2609.24997%2C2609.27455%2C2609.29647%2C2609.29964%2C2609.30798%2C2609.32835%2C2609.34428%2C2609.35002%2C2609.35232%2C2609.35394%2C2609.35581%2C2609.36071%2C2609.36352%2C2609.36777%2C2609.37017%2C2609.38923%2C2609.39222&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:35Z", "role": "primary", "sha256": "6499b4f4917eec0765d16d6e26a08d7edc5146a30753b19cf8218517656bb070", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2608.16211%2C2608.16391%2C2608.17067%2C2608.17975%2C2608.19741%2C2608.21500%2C2608.22301%2C2608.22510%2C2608.23035%2C2608.23320%2C2608.26238%2C2608.26546%2C2608.26991%2C2608.27260%2C2608.29814%2C2608.31076%2C2609.02292%2C2609.04173%2C2609.04298%2C2609.05776%2C2609.06059%2C2609.07529%2C2609.08149%2C2609.08418%2C2609.09187%2C2609.11561%2C2609.12533%2C2609.15096%2C2609.15544%2C2609.20732%2C2609.21293%2C2609.23986%2C2609.24971%2C2609.24997%2C2609.27455%2C2609.29647%2C2609.29964%2C2609.30798%2C2609.32835%2C2609.34428%2C2609.35002%2C2609.35232%2C2609.35394%2C2609.35581%2C2609.36071%2C2609.36352%2C2609.36777%2C2609.37017%2C2609.38923%2C2609.39222&max_results=50"}], "id": "arxiv:2608.26991", "identifiers": {"arxiv": "2608.26991"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:35Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.26991"}], "name": "ASIL: Replacing Screenshot-and-Click with Structured State and Semantic Actions", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "ASIL exposes application state as structured JSON and provides code-executable semantic actions. Its application benchmark compares this interface with screenshot control and application-native APIs, and studies supervised and reinforcement learning on the structured interface.", "topics": ["automation", "tool-integration", "evaluation", "reinforcement-learning", "imitation-learning"], "updated_at": "2026-08-27", "url": "https://arxiv.org/abs/2608.26991"}
{"aliases": [], "authors": ["Haoting Shi", "Wenhao Wang", "Weicheng Fang", "Yaozhong Liang", "Tian Jin", "Pengxiang Zhao", "Guangyi Liu", "Siheng Chen", "Yanfeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "69eab19b6e45e1b2ddf9f29f6f0937d51e06a5bb7dea34dadb761361c82f08d4", "submitted_at": "2026-09-04T17:26:18Z", "updated_at": "2026-09-04T17:26:18Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.05374v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2609.05374", "identifiers": {"arxiv": "2609.05374"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.05374"}], "name": "CUA-Universe: A Scalable and Dynamic Environment for Hybrid GUI+CLI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Turns desktop applications into reproducible GUI-and-CLI environments and generates verified hybrid trajectories for agent post-training.", "topics": ["tool-integration", "data-generation", "sandbox"], "updated_at": "2026-09-04", "url": "https://arxiv.org/abs/2609.05374"}
{"aliases": [], "authors": ["Juyong Lee", "Woogyeol Jin", "Kimin Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "14790f9565861cdba0b24661f6bfe87d7572cbad9d76d9a0b2d7216e9bac7ea7", "submitted_at": "2026-09-06T19:22:53Z", "updated_at": "2026-09-06T19:22:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.06792v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.06792", "identifiers": {"arxiv": "2609.06792"}, "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/2609.06792"}], "name": "Improving Proficiency and Efficiency of Android GUI Agents via Self-Generating Tool Actions", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "DroidTool generates application-state tools through proposal, implementation, testing and repair. Relational tests connect the tools’ preconditions and effects, and the resulting tools are evaluated as additions to Android GUI agents.", "topics": ["tool-integration", "automation", "mobile-navigation", "efficiency"], "updated_at": "2026-09-06", "url": "https://arxiv.org/abs/2609.06792"}
{"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": ["Shuai Bai", "Jiayong Deng", "Sicheng Fan", "Yikun Fu", "Chang Gao", "Xuhao Hu", "Mianqiu Huang", "Yizhen Jiang", "Yuheng Jing", "Dehui Kong", "Keliang Li", "Ning Li", "Wanli Li", "Dayiheng Liu", "Dunjie Lu", "Changwei Luo", "Que Shen", "Zheyuan Wang", "Zijian Wang", "Jie Wu", "Gao Wu", "Zhihui Xie", "Rui Xie", "Haiyang Xu", "An Yang", "Jiakang Yuan", "Yanming Zhang", "Jiajun Zhang", "Xi Zhang", "Zhenru Zhang", "Zhuo Zhen", "Mingkang Zhu", "Bowen Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3b312c5a4d638ec9f9a7408fd9ce2d8cbf74fa462d48c8aa37f59e4ad3bd8edc", "submitted_at": "2026-09-18T17:00:56Z", "updated_at": "2026-09-21T07:09:10Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.22000v2"}}, "environments": ["desktop", "linux", "macos", "windows", "android", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.22000", "identifiers": {"arxiv": "2609.22000"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.22000"}], "name": "RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "RecreationWorld gives hybrid agents running reference applications to explore and recreate using native GUI controls and coding tools. Its multi-platform benchmark checks interaction-dependent behavior with programmatic and visual assertions, while trajectory-training transfer is evaluated separately.", "topics": ["evaluation", "data-generation", "tool-integration", "workflow"], "updated_at": "2026-09-21", "url": "https://arxiv.org/abs/2609.22000"}
{"aliases": [], "authors": ["Mingfei Gao", "Rui Tian", "Haiming Gang", "Bohan Zhai", "Le Zhang", "Yuanzheng Gong", "Di Feng", "Ege Özsoy", "Kaixin Ma", "Vishwesh Kirthivasan", "Oğuzhan Fatih Kar", "Roman Bachmann", "Anders Boesen Lindbo Larsen", "Afshin Dehghan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4dfb0703133032426ff8f40ed6a2c8131025b07658c0b643603f8d50ba163d16", "submitted_at": "2026-09-18T17:59:34Z", "updated_at": "2026-09-18T17:59:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.22083v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}], "id": "arxiv:2609.22083", "identifiers": {"arxiv": "2609.22083"}, "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/2609.22083"}], "name": "MintAct: A Unified Visual Agent for Digital Environments", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MintAct trains a visual-agent family for grounding, navigation and tool use across mobile, desktop and web environments. Its shared infrastructure supports trajectory collection and asynchronous online RL while controlling cross-domain sampling and off-policy drift.", "topics": ["grounding", "reinforcement-learning", "data-generation", "tool-integration"], "updated_at": "2026-09-18", "url": "https://arxiv.org/abs/2609.22083"}
{"aliases": [], "authors": ["Yuchen Sun", "Chenglin Cai", "Gongjie Zhang", "Tianyu Xia", "Quyu Kong", "Panrong Tong", "Zhengwen Zeng", "Long Chen", "Steven Hoi", "Chongyang Zhang", "Yue Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "838635f067b706c7b36a948c8100b30bf4cacd218470aaed6970e5186fa44fb9", "submitted_at": "2026-09-25T06:50:04Z", "updated_at": "2026-09-25T06:50:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.30887v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:30.162100+00:00", "role": "primary", "sha256": "096098b62f82867ee39dfd7cf44095edc5c6a9307351265592e485d82eadf111", "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%5B202609200000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T10:53:44.035673+00:00", "role": "primary", "sha256": "2b6b8147cf7b0e1fcded85f397c272bcf535ed179789e76c3ff4f15aa3fe53a7", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2609.30887v1"}], "id": "arxiv:2609.30887", "identifiers": {"arxiv": "2609.30887"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:30.162100+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.30887"}], "name": "From Tapping to Hopping: Augmenting Mobile GUI Agents with App-Native Deeplinks", "notes": "Reviewed the complete retained Atom abstract and selected exact-v1 full text. The Atom has incomplete clauses; its text/hash remain unchanged and this annotation uses the explicit methods. Evaluation is Android-only, with a selected real-app subset. Landing-Activity verification is not task completion. Model/catalog release is prospective in the paper. No independent reproduction.", "published_at": "2026-09-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-Hopper combines verified Android deeplinks with GUI actions, using landing-screen descriptions, hybrid trajectory training and reinforcement learning to select navigation shortcuts and recover through GUI control.", "topics": ["mobile-navigation", "tool-integration", "data-generation", "imitation-learning", "reinforcement-learning", "efficiency"], "updated_at": "2026-09-25", "url": "https://arxiv.org/abs/2609.30887"}
{"aliases": [], "authors": ["Prince Zizhuang Wang", "Chenhao Liang", "Zelong Xu", "Aojie Yuan", "Xiaolin Zhou", "Haiyue Zhang", "Yue Zhao", "Xiyang Hu", "Shuli Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d53f2512b99828849f047f63412264437a59b730c117c801de252e6dc80a33fc", "submitted_at": "2026-09-26T13:18:00Z", "updated_at": "2026-09-26T13:18:00Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32600v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.32600", "identifiers": {"arxiv": "2609.32600"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32600"}], "name": "CUA-SWE: When Computer-Use Agents Meet Visual Software Engineering", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CUA-SWE, a benchmark and execution environment that combines code changes with graphical interaction and visual inspection of running software. Task-specific tests verify repairs and changes whose required information may be available only through the application interface.", "topics": ["evaluation", "workflow", "tool-integration"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32600"}
{"aliases": [], "authors": ["Jianxing Chen", "Xiao Yu", "Shipra Agrawal", "Zhou Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e92c63d4cf58375aa7068570f8a98c145e24f8d393c5807c0aff3ed67f18b9f0", "submitted_at": "2026-09-28T20:01:35Z", "updated_at": "2026-09-28T20:01:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.36201v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.36201", "identifiers": {"arxiv": "2609.36201"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.36201"}], "name": "SCOUT: Synergizing Reasoning and Tool-Use for Computer-Use Safety", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SCOUT, a safety verifier that first derives task-specific completion and safety rubrics from a computer-use trajectory, then actively probes the resulting environment for evidence. It evaluates whether these two stages detect harmful consequences that screenshots alone can miss.", "topics": ["safety-security", "evaluation", "tool-integration"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.36201"}
{"aliases": [], "authors": ["Tongbo Chen", "Junbo Niu", "Zhengxi Lu", "Niu Lian", "Fei Tang", "Yuchen Yan", "Yike Hong", "Yong Du", "Yizhou Liu", "Bofan Chen", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b394677ebe37782950cfc56c0be65ca7a93f34e5550ec9425d91921bf1799847", "submitted_at": "2026-09-29T16:57:25Z", "updated_at": "2026-09-29T16:57:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.38008v1"}}, "environments": ["desktop", "windows"], "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:2609.38008", "identifiers": {"arxiv": "2609.38008"}, "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/2609.38008"}], "name": "HybridCUA: Learning to Orchestrate GUI and CLI for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains computer agents on verified GUI, command-line, and interleaved trajectories to learn when and how to combine interaction modes.", "topics": ["tool-integration", "reinforcement-learning", "data-generation"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.38008"}
{"aliases": [], "authors": ["Yadong Lu", "Theodore Lee", "Yifei Li", "Lawrence Keunho Jang", "Tianci Xue", "Yu Su", "Huan Sun", "Ahmed Hassan Awadallah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "97af38af8647b769c129d917b65f8f744035308f6ace88706a4888a96fe345b9", "submitted_at": "2026-10-02T22:43:06Z", "updated_at": "2026-10-02T22:43:06Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.04116v1"}}, "environments": ["desktop", "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:2610.04116", "identifiers": {"arxiv": "2610.04116"}, "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/2610.04116"}], "name": "CUAWright: A Minimal Unified Interface for Digital Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-02", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Studies a minimal terminal-based execution framework that lets digital agents construct tools and manage context across browser, desktop, and CAD tasks.", "topics": ["tool-integration", "efficiency"], "updated_at": "2026-10-02", "url": "https://arxiv.org/abs/2610.04116"}
{"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": ["Chenyang Yan", "Yingying Zhang", "Kanzhi Cheng", "Qiushi Sun", "Zhengyuan Pan", "Zheng Ma", "Hang Yan", "Yian Wang", "Nuo Chen", "Jialin Cao", "Xingdong Gong", "Wenpo Song", "Han Chen", "Zichen Ding", "Fangzhi Xu", "Shujian Huang", "Xinyu Dai", "Yichen Liu", "Tiankuo Yao", "Bo Wang", "Ben Kao", "Jianbing Zhang", "Lewei Lu", "Dahua Lin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6f7d940d3e341fd8c9f23a4829a5ab1150fc07ee231ccda9b108a3f0d785efb9", "submitted_at": "2026-10-09T17:26:38Z", "updated_at": "2026-10-09T17:26:38Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.14157v1"}}, "environments": ["android", "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"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:46:06Z", "role": "primary", "sha256": "9b385e9f6aa2f7169f616a76887df3b76df605cb2bea071d99ee7ca3f9b659cc", "supports": ["environments"], "url": "https://os-copilot.github.io/OpenMobile2-Home/"}], "id": "arxiv:2610.14157", "identifiers": {"arxiv": "2610.14157"}, "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/2610.14157"}], "name": "OpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native Tools", "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 OpenMobile-2, combining simulated mobile apps, emulator tasks, released training data, and policies that can alternate GUI interaction with app-native tools.", "topics": ["mobile-navigation", "tool-integration", "data-generation", "reinforcement-learning"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2610.14157"}
{"aliases": [], "authors": ["Hyobin Ahn", "Dongjae Lee"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ff5470e792d14ced2cdff648d97e2606eea4b3c822046b7c10272fd2a5f1695f", "submitted_at": "2026-10-11T10:09:44Z", "updated_at": "2026-10-11T10:09:44Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.15502v1"}}, "environments": ["android", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T12:56:03.829575Z", "role": "primary", "sha256": "a8636ac679d2110181d4417d4aef040ddb24464597a7fa6fe6d2c35a2401a3b1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22graphical+user+interface%22+OR+ti%3A%22screen+grounding%22+OR+ti%3A%22UI+grounding%22+OR+ti%3A%22desktop+agents%22+OR+ti%3A%22Android+agent%22+OR+ti%3A%22phone+agent%22+OR+ti%3A%22mobile+GUI%22+OR+ti%3A%22web+navigation%22+OR+ti%3A%22GUI+automation%22+OR+ti%3A%22browser+automation%22%29+AND+submittedDate%3A%5B202608210000+TO+202610201255%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:58.069698Z", "role": "primary", "sha256": "a38b3ddc9d5d6744eefc937f86e174917f9b5de149323632d5d1b20283d79c8e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.15502v1"}], "id": "arxiv:2610.15502", "identifiers": {"arxiv": "2610.15502"}, "kind": "paper", "last_verified_at": "2026-10-20T13:12:58.069698Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.15502"}], "name": "From Observations to Authorized Actions: Measuring Source Information in Android Agents", "notes": "The two prototypes use different workloads and trust assumptions on patched Android/WebView components. Approval is simulated; authenticated browser delivery and atomic capture remain unimplemented. Coarse source history can withhold benign work and persist across tasks. Direct-call checks and limited live-agent trials are separate, and writes are in-memory. No production vulnerability, general defense rate or released artifact is established.", "published_at": "2026-10-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Measures source-attribution limits in Android accessibility and separately tests application-history and selected-WebView-object checks for authorizing AppFunctions actions.", "topics": ["safety-security", "tool-integration", "observability"], "updated_at": "2026-10-11", "url": "https://arxiv.org/abs/2610.15502"}
{"aliases": [], "authors": ["Hanzhang Zhou", "Long Li", "Quyu Kong", "Lei Wang", "Panrong Tong", "Long Chen", "Tianyu Xia", "Gongjie Zhang", "Xu Zhang", "Chen Liu", "Jianan Zhang", "Chenglin Cai", "Zhengwen Zeng", "Steven Hoi", "Yue Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "beab15a2bf8a6aca19d219dbe0a047e8c6d6bd84a580df33633b5d366c175a0e", "submitted_at": "2026-10-15T16:51:35Z", "updated_at": "2026-10-15T16:51:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22200v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.22200", "identifiers": {"arxiv": "2610.22200"}, "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/2610.22200"}], "name": "OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2026-10-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces an open generalist computer-use stack combining visual and programmatic interaction, with model-controlled observation and task-state decisions.", "topics": ["tool-integration", "planning"], "updated_at": "2026-10-15", "url": "https://arxiv.org/abs/2610.22200"}
{"aliases": [], "authors": ["Bowen Yang", "Kaiming Jin", "Zichen Ding", "Zhenyu Wu", "Zehao Li", "Zhoumianze Liu", "Kanzhi Cheng", "Fangzhi Xu", "Yanan Sun", "Qiushi Sun", "Yu Qiao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2b7f3542ca2faa63df81e25721a01928bfb35e9e3de458e8be27fd3f9010adc8", "submitted_at": "2026-10-15T17:14:02Z", "updated_at": "2026-10-15T17:14:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22234v1"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "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:2610.22234", "identifiers": {"arxiv": "2610.22234"}, "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/2610.22234"}], "name": "OS-Helix: Unifying GUI Interaction and Code Execution in Native Computer-Using Agents", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2026-10-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains native computer-use models to coordinate GUI actions and code execution using hybrid demonstrations and online reinforcement learning.", "topics": ["tool-integration", "reinforcement-learning"], "updated_at": "2026-10-15", "url": "https://arxiv.org/abs/2610.22234"}
{"aliases": [], "authors": ["Xinyu Fang", "Yuchi Wang", "Weihao Yu", "Rongyao Fang", "Zhaohai Li", "Zhixun Li", "Tiancheng Gu", "Xilin Wei", "Yuhang Zang", "Hongsheng Li", "Dahua Lin", "Guofeng Zhang", "Shuai Bai", "Zhibo Yang", "Dayiheng Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6dfa0c59ec1262cebcf2a4e61ab1e0b58c647513d8e6eedb2ce7e6a1a675558c", "submitted_at": "2026-10-14T12:46:54Z", "updated_at": "2026-10-14T12:46:54Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.23986v1"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T13:48:11Z", "role": "primary", "sha256": "89e110b6fa882639fb73936cb27e5373d11b632938c31968f563152534bf5670", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.14966%2C2610.23986&max_results=2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:09:58.148629Z", "role": "primary", "sha256": "bef4399f087e484b62ff5b8dc2c283f92e0357b94c93b464b86fd97fe7b069cb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.23986v1"}], "id": "arxiv:2610.23986", "identifiers": {"arxiv": "2610.23986"}, "kind": "paper", "last_verified_at": "2026-10-20T15:09:58.148629Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.23986"}], "name": "BlenderWorld: Benchmarking Visual Agents for 3D Creation from Real-World References", "notes": "The evaluated agents use Linux shell/Python/MCP tools and can inspect viewport screenshots; graphical CUA control and Windows/macOS are future work. Each agent is run once per task, with different Codex/Claude Code harnesses across systems. Scores combine weighted rubric satisfaction and model-rated visual quality. Tool counts include unsuccessful calls and do not establish matched compute or successful actions. No independent reproduction or artifact-release verification.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Benchmarks agents creating editable Blender projects through shell, Python and MCP tools, judging delivered scenes with multi-view renders, task-specific rubrics and visual-quality ratings.", "topics": ["evaluation", "tool-integration", "sandbox"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.23986"}
{"aliases": [], "authors": ["Zhanhui Zhou", "Weichen Zhang", "Haoran Liu", "Lingjie Chen", "Tianneng Shi", "Kevin Lin", "Zhengyuan Yang", "Lijuan Wang", "Dawn Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0a297401f7ad4dcd2f485ade6bab295e365280bb803dcecc111aebf0693dd203", "submitted_at": "2026-10-19T09:30:54Z", "updated_at": "2026-10-19T09:30:54Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26071v1"}}, "environments": ["desktop", "linux", "windows", "web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:28:55.967592+00:00", "role": "primary", "sha256": "d4231fa9677e916ca8cc27493dfbbe9fb34c7430207150e6b9be10c2bad43bfb", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26071v1"}], "id": "arxiv:2610.26071", "identifiers": {"arxiv": "2610.26071"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26071"}], "name": "CUA-Lite: An Open Platform for Computer Use Agents", "notes": "Reviewed the complete official abstract and selected v1 methods, training and limitations sections. WebVoyager uses offline mirrors; VM-free sandboxes are Linux-only, and reported desktop/mobile gains do not establish unseen-app transfer. No results or released artifacts were independently reproduced.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "CUA-Lite unifies environment interfaces, supervised data formats and per-model harnesses for evaluation, reinforcement learning and supervised fine-tuning across desktop, browser and mobile tasks.", "topics": ["tool-integration", "data-generation", "evaluation", "imitation-learning", "reinforcement-learning", "sandbox"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26071"}
{"aliases": [], "authors": ["Zhaoyang Wang", "Qianhui Wu", "Haorui Wang", "Simon Yu", "Baolin Peng", "Canwen Xu", "Hao Cheng", "Wenlin Yao", "Boyi Liu", "Yite Wang", "Jianfeng Gao", "Huaxiu Yao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1f1ff33859352b61f017709fff91fd8e32f08636e9423fb044e7ca8e28f06905", "submitted_at": "2026-10-19T17:55:35Z", "updated_at": "2026-10-19T17:55:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26828v1"}}, "environments": ["web", "desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:29:05.337543+00:00", "role": "primary", "sha256": "b680e07a5a883e7bb0210a1734a628be6c47d91b7df0487a2044d56d1b7e227e", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26828v1"}], "id": "arxiv:2610.26828", "identifiers": {"arxiv": "2610.26828"}, "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/2610.26828"}], "name": "Agent World Model in the Wild: Computer-Use Agent Environment Synthesis", "notes": "Reviewed the complete abstract and selected v1 synthesis, training and evaluation sections. The system uses hybrid GUI/MCP/CLI interfaces. Evaluation includes benchmark subsets and model-based judgments, so reported scores are not treated as directly comparable GUI-only results. No independent reproduction.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "WildAWM synthesizes resettable GUI/MCP/CLI environments and cross-interface tasks, using execution-based checks of solvability, difficulty and verifier consistency before supervised and reinforcement learning.", "topics": ["data-generation", "sandbox", "tool-integration", "long-horizon", "imitation-learning", "reinforcement-learning"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26828"}
{"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": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md", "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/LICENSE"], "reason": "The retained README restricts SheetCopilot and its dataset to noncommercial use, while the retained repository LICENSE contains the GNU GPL version 3 text. These declarations do not establish one unambiguous code license."}}}, "environments": ["desktop", "windows", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:59:56Z", "role": "primary", "sha256": "beb8c811bf2d070845aa6fcc7e94ffd43ca4f9f60ce42c72d958598790fbdffc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "licenses.code", "notes"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:47Z", "role": "primary", "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-18T11:03:47Z", "role": "primary", "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/LICENSE"}], "id": "github:bravegroup/sheetcopilot", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-18T10:59:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/BraveGroup/SheetCopilot"}], "name": "SheetCopilot", "notes": "The original agent and the later Ubuntu evaluator have different roles; the repository does not establish that the original agent runs on Ubuntu. Code licensing is unresolved because the retained README and LICENSE contain conflicting declarations.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/BraveGroup/SheetCopilot/HEAD/README.md", "target_id": "arxiv:2305.19308", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Spreadsheet automation agent for Windows/Excel, with benchmark assets and a separately released headless Ubuntu evaluation backend.", "topics": ["workflow", "tool-integration", "evaluation"], "updated_at": null, "url": "https://github.com/BraveGroup/SheetCopilot"}
{"aliases": [], "authors": [], "details": {"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": {"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": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:53Z", "role": "primary", "sha256": "059a0349daaef6bda98eae6d9384ed349ac94b8033b18be817c317a01b472024", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ChromeDevTools/chrome-devtools-mcp/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:07Z", "role": "primary", "sha256": "58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ChromeDevTools/chrome-devtools-mcp/HEAD/LICENSE"}], "id": "github:chromedevtools/chrome-devtools-mcp", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-16T06:40:53Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:07Z", "url": "https://raw.githubusercontent.com/ChromeDevTools/chrome-devtools-mcp/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ChromeDevTools/chrome-devtools-mcp"}], "name": "Chrome DevTools MCP", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "MCP tools for browser control and inspection through Chrome DevTools.", "topics": ["automation", "tool-integration"], "updated_at": null, "url": "https://github.com/ChromeDevTools/chrome-devtools-mcp"}
{"aliases": [], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "release_tag": "v0.8.8"}, "environments": ["windows", "desktop", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "bbec20828c8a0f5483be68f6a00c47b202e9431268d96c1195fbf0745aca1364", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/CursorTouch/Windows-MCP/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "ef43250ad8e57c3267f2a8e0aaeb655eb3445ade8d2dee8c243dc6e241a9d2d8", "supports": ["identity"], "url": "https://api.github.com/repos/CursorTouch/Windows-MCP"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "652b580c99ac2de9d56d124225ba302f0dd83430ccc9b5162e16f683c56b952f", "supports": ["release_status"], "url": "https://api.github.com/repos/CursorTouch/Windows-MCP/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:58Z", "role": "primary", "sha256": "db3de978027106a991da3f7f2d78b216cfffee9b08604872fa3216927acc7a2d", "supports": ["release_status"], "url": "https://api.github.com/repos/CursorTouch/Windows-MCP/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:59Z", "role": "primary", "sha256": "20ced59173768e20e91a43e3a80443aa1200b129c134f07f0cc45779747a1743", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/CursorTouch/Windows-MCP/HEAD/LICENSE.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T04:00:59Z", "role": "primary", "sha256": "20ced59173768e20e91a43e3a80443aa1200b129c134f07f0cc45779747a1743", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/CursorTouch/Windows-MCP/HEAD/LICENSE.md"}], "id": "github:cursortouch/windows-mcp", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T03:59:53Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T04:00:59Z", "url": "https://raw.githubusercontent.com/CursorTouch/Windows-MCP/HEAD/LICENSE.md"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/CursorTouch/Windows-MCP"}], "name": "Windows-MCP", "notes": "Public source and official release metadata are present. The README explicitly supports Windows 10 and 11 and includes a browser DOM inspection mode. Code is MIT-licensed. This is an execution tool, not a trained model; runtime behavior and latency claims were not independently tested.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A Windows MCP bridge exposing application, window, keyboard and mouse operations plus native UI-state inspection for agents.", "topics": ["tool-integration", "automation", "screen-understanding"], "updated_at": null, "url": "https://github.com/CursorTouch/Windows-MCP"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android", "ios"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "0d05433c730e74f13be753fa6156737f797a69d61d8793e47f5a56b3a63ee790", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/droidrun/mobile-harness/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "de215a2914d50cb87fcab9e19efc856b9e600a68eefec5c63e284dae34977133", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/droidrun/mobile-harness/HEAD/LICENSE"}], "id": "github:droidrun/mobile-harness", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T10:48:41Z", "url": "https://raw.githubusercontent.com/droidrun/mobile-harness/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/droidrun/mobile-harness"}], "name": "Mobile Harness", "notes": "The publisher describes this as a Markdown harness, not an agent runtime. Published benchmark results have not been independently reproduced.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Portable agent instructions and integration tooling for local or cloud phone control through mobilerun-core.", "topics": ["automation", "tool-integration"], "updated_at": null, "url": "https://github.com/droidrun/mobile-harness"}
{"aliases": ["Droidrun"], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "release_tag": "v0.7.0"}, "environments": ["android", "ios"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "80052560e06db353277a55f3f8f6a4b5231aa6dc1376d5182559b72adcb5c3a1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/droidrun/mobilerun/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:56Z", "role": "primary", "sha256": "e48b9154a58b4b7021a4707ca185af32b9dd81c828d53ba36a86f195c1720344", "supports": ["identity"], "url": "https://api.github.com/repos/droidrun/mobilerun"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "7f40d42acd62b4405b5077a0f594a35aae73fb057589346afa5f6fb22d710a0c", "supports": ["release_status"], "url": "https://api.github.com/repos/droidrun/mobilerun/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:56Z", "role": "primary", "sha256": "f85c8b766edc107c455a86e363c56af3d1f7fce669b55377f4010982aa3a7e44", "supports": ["release_status"], "url": "https://api.github.com/repos/droidrun/mobilerun/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:57Z", "role": "primary", "sha256": "b3ebb19fda8286efa415d457529eafff41e1b1a27fe4e00d99f34e1559b6cdd9", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/droidrun/mobilerun/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "e48b9154a58b4b7021a4707ca185af32b9dd81c828d53ba36a86f195c1720344", "supports": ["identity", "aliases", "notes"], "url": "https://api.github.com/repos/droidrun/droidrun"}, {"method": "license_header", "retrieved_at": "2026-10-17T04:00:57Z", "role": "primary", "sha256": "b3ebb19fda8286efa415d457529eafff41e1b1a27fe4e00d99f34e1559b6cdd9", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/droidrun/mobilerun/HEAD/LICENSE"}], "id": "github:droidrun/mobilerun", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T03:59:53Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T04:00:57Z", "url": "https://raw.githubusercontent.com/droidrun/mobilerun/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/droidrun/mobilerun"}], "name": "Mobilerun (formerly Droidrun)", "notes": "The former droidrun/droidrun repository resolves to droidrun/mobilerun through the official API. Source and release v0.7.0 are public. Android and iOS setup paths differ, and the companion Portal app is a separate device bridge. MIT applies to framework code; model-provider and cloud-service terms are separate.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A framework that combines screenshots and accessibility state with agent tools for mobile-device tasks through command-line and Python interfaces.", "topics": ["mobile-navigation", "tool-integration", "planning"], "updated_at": null, "url": "https://github.com/droidrun/mobilerun"}
{"aliases": ["Droidrun Portal"], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/droidrun/mobilerun-portal/HEAD/LICENSE"], "reason": "GitHub classifies this license as NOASSERTION, but the actual program-specific notice is explicit. Do not reduce the permission to a generic unknown or silently transfer the framework MIT declaration."}}, "release_tag": "v0.7.25"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "fd2257dd735dbb755f2de36b44e425e5862004a04130e53e92b330f1419479a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/droidrun/mobilerun-portal/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:56Z", "role": "primary", "sha256": "be519be28e277bd979717996e9587879c6a5d9235b95944305252b18595da413", "supports": ["identity"], "url": "https://api.github.com/repos/droidrun/mobilerun-portal"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "65e81cfc9f227350e33da9e09030a7098e82a6021aa9b04cf0485ed2acee95a0", "supports": ["release_status"], "url": "https://api.github.com/repos/droidrun/mobilerun-portal/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:01Z", "role": "primary", "sha256": "31b8d75541bbfb7dba2f80adc83aa099ef83ca057b467eb2cb98b586f6d9d045", "supports": ["release_status"], "url": "https://api.github.com/repos/droidrun/mobilerun-portal/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:01:01Z", "role": "primary", "sha256": "945331c1cfa589694eb6ccc51574a72ee215a01b10bd5556f1b8be8d11bd71ab", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/droidrun/mobilerun-portal/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:54Z", "role": "primary", "sha256": "be519be28e277bd979717996e9587879c6a5d9235b95944305252b18595da413", "supports": ["identity", "aliases", "notes"], "url": "https://api.github.com/repos/droidrun/droidrun-portal"}], "id": "github:droidrun/mobilerun-portal", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T03:59:54Z", "licenses": {"code": {"basis": "The program-specific copyright notice explicitly permits GNU AGPL version 3 or any later version.", "declared": "AGPL-3.0-or-later", "retrieved_at": "2026-10-17T04:01:01Z", "url": "https://raw.githubusercontent.com/droidrun/mobilerun-portal/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/droidrun/mobilerun-portal"}], "name": "Mobilerun Portal (formerly Droidrun Portal)", "notes": "The official API resolves the former Droidrun Portal path to mobilerun-portal. Release metadata lists a v0.7.25 Android APK. Although GitHub labels its license NOASSERTION, the inspected LICENSE explicitly grants AGPL version 3 or later for the program. Accessibility-dependent UI functions require the service; no permissions were granted or device operated.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "An Android accessibility service that exposes UI-element inspection, device-control interfaces and screen feedback to mobile automation agents.", "topics": ["mobile-navigation", "screen-understanding", "tool-integration"], "updated_at": null, "url": "https://github.com/droidrun/mobilerun-portal"}
{"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": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "63349e7557bd99240d4628de3994a53b69e54e3bd809895be11e9879bfba40cb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "f872ff2b1a064f46a45eee86f2cb2562fc0513285b80036c6cf7df9a5ffc6c70", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/LICENSE"}], "id": "github:gydpku/pptc", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T10:48:41Z", "url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/gydpku/PPTC"}], "name": "PPTC", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/gydpku/PPTC/HEAD/README.md", "target_id": "arxiv:2311.01767", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "PowerPoint creation and editing benchmark with API-based execution and checks of the resulting presentation files.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://github.com/gydpku/PPTC"}
{"aliases": [], "authors": [], "details": {"availability": "Public 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": "2026-10-16T06:40:53Z", "role": "primary", "sha256": "b33aa71a13d655e96e6b736d0c9de1e5ddea52f205916ca20b876387ff830e87", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/microsoft/playwright-mcp/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:07Z", "role": "primary", "sha256": "9a7110fc2d2f964038e5dc49128f908f29f47a574c961cba16085914e879cbda", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/microsoft/playwright-mcp/HEAD/LICENSE"}], "id": "github:microsoft/playwright-mcp", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-16T06:40:53Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:07Z", "url": "https://raw.githubusercontent.com/microsoft/playwright-mcp/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/microsoft/playwright-mcp"}], "name": "Playwright MCP", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "MCP server that lets LLM clients drive Playwright browsers through structured accessibility snapshots instead of screenshots, with opt-in vision, PDF and devtools capabilities.", "topics": ["automation", "tool-integration"], "updated_at": null, "url": "https://github.com/microsoft/playwright-mcp"}
{"aliases": ["Webwright", "CUAWright"], "authors": [], "details": {}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "7c0f87f684e5ed4d77b175b63a5dcaf0a268644c3662c723713de4cf0ac1de40", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/microsoft/Webwright/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c2cfccb812fe482101a8f04597dfc5a9991a6b2748266c47ac91b6a5aae15383", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/microsoft/Webwright/HEAD/LICENSE"}], "id": "github:microsoft/webwright", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/microsoft/Webwright/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/microsoft/Webwright"}], "name": "CUAWright (formerly Webwright)", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/microsoft/Webwright/HEAD/README.md", "target_id": "arxiv:2610.04116", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Agent framework for browser and desktop tasks using automation code.", "topics": ["tool-integration"], "updated_at": null, "url": "https://github.com/microsoft/Webwright"}
{"aliases": [], "authors": [], "details": {"availability": "Published source; official release metadata inspected", "release_tag": "1.0.11"}, "environments": ["android", "ios"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "bc24fa665d1a49fa806d58ec569d93e134058e0cb1a7bc9324339ba0e35ff107", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/mobile-next/mobile-mcp/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "80e8ad254f71f5c80b91a6b80edd09f96116ef4967ae4e6832e5960357f7b99a", "supports": ["identity"], "url": "https://api.github.com/repos/mobile-next/mobile-mcp"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T03:59:53Z", "role": "primary", "sha256": "49684448428c28afd4bf191f695c3d45fbe46910fa93951d2a02f6b8afcf9f16", "supports": ["release_status"], "url": "https://api.github.com/repos/mobile-next/mobile-mcp/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:57Z", "role": "primary", "sha256": "43a5c02407afe86b6f1fedb4299f28420778ce21825cd87516deafe05507a7bc", "supports": ["release_status"], "url": "https://api.github.com/repos/mobile-next/mobile-mcp/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:00:58Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/mobile-next/mobile-mcp/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T04:00:58Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/mobile-next/mobile-mcp/HEAD/LICENSE"}], "id": "github:mobile-next/mobile-mcp", "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": "2026-10-17T04:00:58Z", "url": "https://raw.githubusercontent.com/mobile-next/mobile-mcp/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/mobile-next/mobile-mcp"}], "name": "Mobile Next MCP", "notes": "Source and release 1.0.11 are public. The README specifies iOS and Android device setup and supports local devices as well as a separate cloud offering. Apache-2.0 is declared for this code; no cloud-service pricing or permissions are inferred.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "An MCP server exposing mobile accessibility snapshots, screenshots and device interactions across real devices, simulators and emulators.", "topics": ["mobile-navigation", "tool-integration", "automation"], "updated_at": null, "url": "https://github.com/mobile-next/mobile-mcp"}
{"aliases": [], "authors": [], "details": {"availability": "Public source and minimal-experiment documentation are released; default README path is absent.", "repository_revision": "b1a02a52a6c94b96f63016ac12d3a4ab6a410b8e"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:58Z", "role": "primary", "sha256": "74508100843f8db6db6bb151e77ac1cc604a6b7ebd16d3225266d73434465136", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/MobileLLM/AutoDroid-V2/HEAD/README_minimal.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:10Z", "role": "primary", "sha256": "f349963f2ea3f7b81f1c04ee875654382929b1f93d99d0168d6e086155f41d72", "supports": ["identity"], "url": "https://api.github.com/repos/MobileLLM/AutoDroid-V2"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:03Z", "role": "primary", "sha256": "70354b856b75bc4a2aca0a7274321637039ec5d083523ca2f677cd11d475b386", "supports": ["release_status", "licenses.code"], "url": "https://api.github.com/repos/MobileLLM/AutoDroid-V2/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:58Z", "role": "primary", "sha256": "8ba37983a2c2da53d96a52696c07e4bdbea4e987a1a07b557935d614c3be46d2", "supports": ["summary", "details", "relations"], "url": "https://raw.githubusercontent.com/MobileLLM/AutoDroid-V2/HEAD/step_3_training/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:58Z", "role": "primary", "sha256": "cfed016e038d09ce7c9e829273fffa4a6c618bf8e9043312e4690fbe1f40d247", "supports": ["environments", "details"], "url": "https://raw.githubusercontent.com/MobileLLM/AutoDroid-V2/HEAD/step_0_installation/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:58Z", "role": "primary", "sha256": "f5534771f68770ed2e45daf94fbf80304ba2904387571c34c5d80a152bef0a00", "supports": ["summary"], "url": "https://raw.githubusercontent.com/MobileLLM/AutoDroid-V2/HEAD/step_1_doc_generation/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:59Z", "role": "primary", "sha256": "0564f566fde97360496bea4a6eb023c4c84e03cdfdf5201c5b27235c36e32b18", "supports": ["summary", "details"], "url": "https://raw.githubusercontent.com/MobileLLM/AutoDroid-V2/HEAD/step_4_accuracy_validation/README.md"}, {"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": "github:mobilellm/autodroid-v2", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-17T09:14:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/MobileLLM/AutoDroid-V2"}], "name": "AutoDroid V2", "notes": "The repository uses README_minimal.md rather than README.md and includes a documented App Launcher experiment, training setup and DroidTask/LlamaTouch evaluation scripts. Its supplied emulator setup targets x86_64 Linux and Android API 31. No repository-level code license was found; code, APK, data and weight terms are not inferred from one another. Dependencies, external evaluation files and model service are still required; no software was run.", "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:2412.18116", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "An Android automation implementation that builds app documentation from exploration traces and evaluates language-model-generated execution code in emulator workflows.", "topics": ["mobile-navigation", "tool-integration", "data-generation"], "updated_at": null, "url": "https://github.com/MobileLLM/AutoDroid-V2"}
{"aliases": ["Oculix", "SikuliX continuation"], "authors": [], "details": {"availability": "Public source and official stable release assets.", "license_components_note": "Project source declares MIT; the retained third-party dependency notices describe additional component terms.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/oculix-org/Oculix/HEAD/LICENSE"], "reason": "Publisher project-source LICENSE. Third-party dependency notices are separately retained and no blanket MIT assertion is made for the compiled distribution."}}, "release_tag": "v4.0.0", "repository_revision": "02ea8844483a83a2963db8016cd7ad421e15bc91"}, "environments": ["desktop", "windows", "macos", "linux", "android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:14Z", "role": "primary", "sha256": "41d32f40bfb960f38b77ebb62a73ae20454be6c52829c9cbf0b8fedd27de2297", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/oculix-org/Oculix/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:14Z", "role": "primary", "sha256": "151ed1aaf544b9afcfd2d8f13d0fa682d8cf8f8a68496af20c3597b082070801", "supports": ["identity"], "url": "https://api.github.com/repos/oculix-org/Oculix"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:19:48Z", "role": "primary", "sha256": "cba4541e5d89210b570e42cc69c1529ca36e698622c69b4bb0eeffe56339a759", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/oculix-org/Oculix/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:08Z", "role": "primary", "sha256": "1022eb9591a2f01833e88c2bae2869faeac9b92e535901d2c79e93a714278777", "supports": ["release_status", "details.release_tag", "environments", "notes"], "url": "https://api.github.com/repos/oculix-org/Oculix/releases/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:14Z", "role": "primary", "sha256": "8994ff2000b0871aca819ecd7233430bb82cea3f0134a555cebbfa9e02e0f4da", "supports": ["identity", "relations", "notes"], "url": "https://raw.githubusercontent.com/oculix-org/SikuliX1/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:13Z", "role": "primary", "sha256": "102b86e9019fe5e936486b3213d2bbd9e80d6c7ad393284b1a4298b0be555ce7", "supports": ["identity"], "url": "https://api.github.com/repos/RaiMan/SikuliX1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:03Z", "role": "primary", "sha256": "cea4b6292d59625e14968ec666faa7ae6e237b510d846a96875bd6a282677f3a", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/oculix-org/Oculix/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:27:50Z", "role": "primary", "sha256": "f0e0d8aa91c1c60bac2250688248c85503036aeac37ec02b8a75fe9d50b6ae4c", "supports": ["licenses.code", "notes"], "url": "https://raw.githubusercontent.com/oculix-org/Oculix/HEAD/API/src/main/resources/Settings/THIRD-PARTY.txt"}], "id": "github:oculix-org/oculix", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T11:18:14Z", "licenses": {"code": {"basis": "Publisher project-source LICENSE. Third-party dependency notices are separately retained and no blanket MIT assertion is made for the compiled distribution.", "declared": "MIT", "retrieved_at": "2026-10-17T11:23:03Z", "url": "https://raw.githubusercontent.com/oculix-org/Oculix/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/oculix-org/Oculix"}], "name": "OculiX (SikuliX continuation)", "notes": "The official stable-release API lists v4.0.0 with platform-specific API/IDE JARs and an MCP server JAR; no binary was downloaded. The default README still advertises 3.0.3, so release-specific compatibility follows the release notes rather than that older badge. The RaiMan/SikuliX1 API resolves to oculix-org/SikuliX1, whose README calls itself a historical mirror and directs current users here; it is not separately counted. MIT describes the project source. Retained dependency notices also include GPL/LGPL, Apache, BSD and MPL declarations; a bundled artifact is not asserted to be uniformly MIT.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/oculix-org/Oculix/HEAD/README.md", "target_id": "github:sikuli/sikuli", "type": "variant_of"}], "review_status": "curated", "scope": "core", "summary": "A maintained continuation of SikuliX that locates on-screen targets with image matching and OCR, drives desktop or Android interfaces, and exposes scripting and optional MCP interfaces.", "topics": ["automation", "grounding", "tool-integration"], "updated_at": null, "url": "https://github.com/oculix-org/Oculix"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "8f6c33487eca77769798677d32bce15335d0439f129c9a0413e12e0d9e70ab7a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OpenInterpreter/open-interpreter/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:04Z", "role": "primary", "sha256": "d17f227e4df5da1600391338865ce0f3055211760a36688f816941d58232d8dc", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OpenInterpreter/open-interpreter/HEAD/LICENSE"}], "id": "github:openinterpreter/open-interpreter", "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/OpenInterpreter/open-interpreter/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OpenInterpreter/open-interpreter"}], "name": "Open Interpreter", "notes": "The repository has evolved substantially; computer-use functionality is documented in its Computer Use section. Do not infer parity with historical releases.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Assistant framework combining code execution and computer interaction.", "topics": ["tool-integration"], "updated_at": null, "url": "https://github.com/OpenInterpreter/open-interpreter"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["GUI gestures", "foreground-app tool calls in hybrid mode"], "counts": [{"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/bench_env/benchmark/official.json", "population": "official.json", "qualifier": "exact", "unit": "task definitions", "value": 215}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md", "population": "complete simulator installation", "qualifier": "exact", "unit": "simulated clients", "value": 63}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md", "population": "new clients in this release", "qualifier": "exact", "unit": "simulated clients", "value": 35}], "evaluation": "Programmatic checkers inspect changes in application state; both modes share task definitions and initial states.", "evidence_urls": ["https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md", "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/bench_env/benchmark/official.json"], "focus": ["task completion", "cross-application workflows", "GUI and app-native tool comparison"], "notes": "The task count was checked against unique task_id values in the published manifest. The 35 new clients are a subset of 63 total clients. This simulator runs in a browser; Android-emulator apps are a separate part of OpenMobile-2.", "observations": ["screenshots"], "setting": "interactive"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:10Z", "role": "primary", "sha256": "d9f850ec196e715650e679cb81515adaa64ddc5a60a882bfe97b0f6a098ddc83", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:46:06Z", "role": "primary", "sha256": "9b385e9f6aa2f7169f616a76887df3b76df605cb2bea071d99ee7ca3f9b659cc", "supports": ["relations"], "url": "https://os-copilot.github.io/OpenMobile2-Home/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:50:11Z", "role": "primary", "sha256": "35c1641b349201c68917fbe4c4036814487678e9588d9b1c903af26917800316", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/bench_env/benchmark/official.json"}, {"method": "license_header", "retrieved_at": "2026-10-17T01:52:51Z", "role": "primary", "sha256": "7618cd7de039c264759ca8ca0b9130c6de63a4536dcf436baedd4cfbf639073e", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/LICENSE"}], "id": "github:openmobile2/mobilegym-plusplus", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T01:48:10Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T01:52:51Z", "url": "https://raw.githubusercontent.com/OpenMobile2/MobileGym-plusplus/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OpenMobile2/MobileGym-plusplus"}], "name": "MobileGym++ / MobileGym++ Bench", "notes": "These clients simulate named commercial applications and are not those commercial products. The code and simulator data have distinct license files. No benchmark run was performed. The README describes a phone that runs in the browser, and the Android-emulator playground is a separate OpenMobile2 release, so the platform facet is web rather than a mobile operating system.", "published_at": null, "relations": [{"evidence_url": "https://os-copilot.github.io/OpenMobile2-Home/", "target_id": "arxiv:2610.14157", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Browser-rendered mobile simulation with resettable app state and a shared task suite for GUI-only and GUI-plus-tool agents.", "topics": ["mobile-navigation", "evaluation", "cross-app", "tool-integration"], "updated_at": null, "url": "https://github.com/OpenMobile2/MobileGym-plusplus"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:10Z", "role": "primary", "sha256": "485279038266d9454df75e27a80a0a0d71c24e90700126d2dcfa0da921b1273d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T01:52:51Z", "role": "primary", "sha256": "b09fcaf4fbc265223c2479c410beef2b98c5e687c78ecedf3452aa12ffc465f1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/LICENSE"}], "id": "github:openmobile2/openmobile2-code", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T01:48:10Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T01:52:51Z", "url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OpenMobile2/OpenMobile2-Code"}], "name": "OpenMobile-2 code", "notes": "The README marks evaluation code, checkpoints, environments, and data as released; data-construction scripts and other resources remain pending. Each benchmark runtime is started separately. Code terms do not establish dataset permissions.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/README.md", "target_id": "arxiv:2610.14157", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation adapters and training guidance for OpenMobile-2 across Android emulators and browser-rendered mobile simulation.", "topics": ["mobile-navigation", "tool-integration", "evaluation"], "updated_at": null, "url": "https://github.com/OpenMobile2/OpenMobile2-Code"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:02Z", "role": "primary", "sha256": "b840368441695a54137216c2ba6bc99a8d35df75ad4196a4f0f0a1f5b5a70a9d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Copilot/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:04Z", "role": "primary", "sha256": "3704700ddeaef20ec20cad39ed105ea518b316bc80c04dbcf2e69c3dc768a34f", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Copilot/HEAD/LICENSE"}], "id": "github:os-copilot/os-copilot", "identifiers": {}, "kind": "agent", "last_verified_at": "2026-10-16T06:36:02Z", "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/OS-Copilot/OS-Copilot/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OS-Copilot/OS-Copilot"}], "name": "OS-Copilot", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/OS-Copilot/HEAD/README.md", "target_id": "arxiv:2402.07456", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Open-source library for generalist agents acting on an operating system's web, terminals, files and third-party apps, with the FRIDAY assistant, which takes custom tools and self-learns Excel tasks.", "topics": ["planning", "tool-integration"], "updated_at": null, "url": "https://github.com/OS-Copilot/OS-Copilot"}
{"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": "2026-10-16T13:14:29Z", "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 browser-library source and npm package metadata.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/reidbarber/webmarker/HEAD/LICENSE"], "reason": "Publisher source license for WebMarker."}}, "package_version": "0.7.0", "repository_revision": "03cf632eef3f76bfbc4fd9c23b31baa0355e41c7"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:56Z", "role": "primary", "sha256": "b46edb85d41e033f84bdd9bb96122f801364448f28a36fa6deccc110a19b6ffc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/reidbarber/webmarker/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:56Z", "role": "primary", "sha256": "a44253aa866acd38fc3e2275aae61ecd051f956b18cb7b29346c699cf6241d6b", "supports": ["identity"], "url": "https://api.github.com/repos/reidbarber/webmarker"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:06Z", "role": "primary", "sha256": "fc8de18814136ca3684749363727f263908c23e6ae274a0ed589923774aa3a95", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/reidbarber/webmarker/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "1caae04ee16fcc72c48439677f5aba55128879cd99d6d55fd72caa85a7030e69", "supports": ["identity", "release_status", "details.package_version"], "url": "https://registry.npmjs.org/webmarker-js/latest"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "62dcdb525872e3f49d90ff9b354b79341ac2574522bca7c7146f79040491a627", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/reidbarber/webmarker/HEAD/LICENSE"}], "id": "github:reidbarber/webmarker", "identifiers": {}, "kind": "tool", "last_verified_at": "2026-10-17T12:30:56Z", "licenses": {"code": {"basis": "Publisher source license for WebMarker.", "declared": "MIT", "retrieved_at": "2026-10-17T12:34:10Z", "url": "https://raw.githubusercontent.com/reidbarber/webmarker/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/reidbarber/webmarker"}], "name": "WebMarker", "notes": "Source and npm release metadata for webmarker-js 0.7.0 are public. The documented mark/unmark API supports screenshot prompts and Playwright interaction, but this utility is not itself a language model or autonomous agent. The WebVoyager/Set-of-Mark references are usage inspiration and are not asserted as papers introducing this library. No package was installed or executed.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A browser library that labels interactive elements and their boxes for visual prompting, preserving element mappings that automation code can use to execute selected actions.", "topics": ["grounding", "screen-understanding", "tool-integration"], "updated_at": null, "url": "https://github.com/reidbarber/webmarker"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "5559ca58f8ee06d2812c6cbc8c3e49ef112f4ef41fdcf4e95568b268bb24bf14", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["relations", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "5559ca58f8ee06d2812c6cbc8c3e49ef112f4ef41fdcf4e95568b268bb24bf14", "supports": ["notes"], "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "d7a04bf11ce24a7b79cd44b4a78dcc9971eaeb3b565821de437e0ecb7723e634", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/LICENSE"}], "id": "github:saagent/mcpworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/SAAgent/MCPWorld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/SAAgent/MCPWorld"}], "name": "MCPWorld", "notes": "The retained paper reports 201 tasks, whereas the retained repository README describes approximately 170. These are separate source descriptions; their release/version difference has not been established.", "published_at": null, "relations": [{"evidence_url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending", "target_id": "arxiv:2506.07672", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Evaluation environment combining computer interaction with tool access.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://github.com/SAAgent/MCPWorld"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["Python code", "application API calls"], "counts": [{"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50", "population": "original AppWorld benchmark (2024 paper)", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2407.18901", "unit": "tasks", "value": 750}, {"basis": "reported", "evidence_url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50", "population": "original AppWorld environment (2024 paper)", "qualifier": "exact", "reference_url": "https://arxiv.org/abs/2407.18901", "unit": "applications", "value": 9}], "evaluation": "State-based unit tests check task outcomes and unintended changes.", "evidence_urls": ["https://raw.githubusercontent.com/stonybrooknlp/appworld/HEAD/README.md", "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"], "focus": ["interactive coding", "cross-app tool use"], "notes": "This is an adjacent API/coding benchmark, not a screenshot-only task suite. Preserve the named release; task counts here refer to the original paper.", "observations": ["API documentation", "execution feedback"], "setting": "interactive"}, "leaderboard_url": "https://appworld.dev/appworld/leaderboard"}, "environments": ["unspecified"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:45:03Z", "role": "primary", "sha256": "f4049141b6e520217395b7ccd9f27048c8e0f90331690dcbef300d2bc4ff7bdd", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol", "details.leaderboard_url"], "url": "https://raw.githubusercontent.com/stonybrooknlp/appworld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:42Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/StonyBrookNLP/appworld/HEAD/LICENSE"}], "id": "github:stonybrooknlp/appworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:45:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T10:48:42Z", "url": "https://raw.githubusercontent.com/StonyBrookNLP/appworld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/StonyBrookNLP/appworld"}], "name": "AppWorld", "notes": "Primarily API and coding interaction; included as an adjacent comparison for GUI and hybrid agents.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/stonybrooknlp/appworld/HEAD/README.md", "target_id": "arxiv:2407.18901", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Simulated applications, user data, and state-based task checks for evaluating agents that interact through APIs and code.", "topics": ["evaluation", "tool-integration", "cross-app"], "updated_at": null, "url": "https://github.com/StonyBrookNLP/appworld"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://docs.google.com/spreadsheets/d/e/2PACX-1vRR3Wl7wsCgHpwUw1_eUXW_fptAPLL3FkhnW_rua0O1Ji_GIVrpTjY5LaKAhwO-WeARjnY_KNw0SYNJ/pubhtml"}, "environments": ["cross-platform", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "1b7d1f1cd7a74ac9497c639324ec342dd3f8769e9c0c4cd2cd92a4eb6740ac98", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/LICENSE"}], "id": "github:thudm/agentbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T10:42:12Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T10:48:41Z", "url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/AgentBench"}], "name": "AgentBench", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/AgentBench/HEAD/README.md", "target_id": "arxiv:2308.03688", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Agent evaluation suite with containerized tasks and a function-calling interface, including operating-system and shopping environments.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://github.com/THUDM/AgentBench"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/Tongyi-MAI/MAI-UI/HEAD/MAI-UI/LICENSE", "https://raw.githubusercontent.com/Tongyi-MAI/MAI-UI/HEAD/MAI-UI/README.md"], "reason": "The implementation and LICENSE live under MAI-UI/. A root-level missing license or the follow-up Qwen-UI-Agent documentation should not change the scoped declaration."}}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T05:29:14Z", "role": "primary", "sha256": "f644a5c0d0eb29b0ff8d8552dd505c5ecedf7c9cd9ab8cf2c617a1a14ecd0c14", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "licenses.code"], "url": "https://raw.githubusercontent.com/Tongyi-MAI/MAI-UI/HEAD/MAI-UI/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:11:45Z", "role": "primary", "sha256": "fd90bc766d7e73889d14e8d3c5c92d384fa86e163222808f6e492509f00a3656", "supports": ["identity"], "url": "https://api.github.com/repos/Tongyi-MAI/MAI-UI"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:11:46Z", "role": "primary", "sha256": "4340b3d36bc56963fb8173079da8a0f7e91f195994d29fc47a9f7087cb56ffba", "supports": ["release_status"], "url": "https://api.github.com/repos/Tongyi-MAI/MAI-UI/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T05:29:15Z", "role": "primary", "sha256": "bf989559906a7926b77390082e4267ff7b2b494e8f599a9dc5640319eb45c2ce", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Tongyi-MAI/MAI-UI/HEAD/MAI-UI/LICENSE"}], "id": "github:tongyi-mai/mai-ui", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T05:29:14Z", "licenses": {"code": {"basis": "Retained MAI-UI/LICENSE and MAI-UI README; applies to the inspected implementation, subject to third-party components.", "declared": "Apache-2.0 (MAI-UI subdirectory)", "retrieved_at": "2026-10-17T05:29:15Z", "url": "https://raw.githubusercontent.com/Tongyi-MAI/MAI-UI/HEAD/MAI-UI/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Tongyi-MAI/MAI-UI"}], "name": "MAI-UI inference and evaluation code", "notes": "The original implementation is in the MAI-UI/ subdirectory, with an Apache-2.0 LICENSE there. The current root README also promotes Qwen-UI-Agent, but that follow-up directory contains only documentation/display assets in the inspected tree. This entry does not claim the follow-up runtime, training system or weights are released.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Tongyi-MAI/MAI-UI/HEAD/MAI-UI/README.md", "target_id": "arxiv:2512.22047", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Published MAI-UI grounding and navigation-agent implementations with inference demos and evaluation scripts.", "topics": ["mobile-navigation", "grounding", "tool-integration"], "updated_at": null, "url": "https://github.com/Tongyi-MAI/MAI-UI"}
{"aliases": [], "authors": [], "details": {"availability": "Public task definitions, evaluator and container orchestration source.", "benchmark_protocol": {"actions": ["Android GUI operations", "MCP calls when enabled", "agent-user interaction when enabled"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md", "population": "README benchmark overview", "qualifier": "exact", "unit": "tasks", "value": 201}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md", "population": "README benchmark overview", "qualifier": "exact", "unit": "mobile applications", "value": 20}], "evaluation": "The publisher documents text-answer matching, backend database verification, local-storage inspection and application callbacks.", "evidence_urls": ["https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md"], "focus": ["Android task completion", "agent-user interaction", "hybrid GUI and MCP use"], "notes": "Publisher-reported inventory. GUI-only, agent-user and MCP-augmented configurations have different dependencies and task eligibility; preserve enabled modes and step budget with results.", "observations": ["Android application interface", "user-agent interaction when enabled"], "setting": "interactive"}, "leaderboard_url": "https://tongyi-mai.github.io/MobileWorld/#leaderboard", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/LICENSE"], "reason": "Repository code license; self-hosted applications, model APIs and MCP services keep their own terms."}}, "repository_revision": "e7eee8981135ecbe39005a231edf63574defd79c"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:16Z", "role": "primary", "sha256": "fbc306b3414dfdb230171358efd6eb8d67573fbb94e11487a4a7e971e2abf0d9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:15Z", "role": "primary", "sha256": "266e1fe5b5b1a4c1540ab0fd5eba206395084ee283f0079f424930a156cf8e18", "supports": ["identity"], "url": "https://api.github.com/repos/Tongyi-MAI/MobileWorld"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:08Z", "role": "primary", "sha256": "14500ed2dfa14bb5c34211090f0c8827e4a16c3233e35a2d695c0d7261f9358b", "supports": ["release_status"], "url": "https://api.github.com/repos/Tongyi-MAI/MobileWorld/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:09Z", "role": "primary", "sha256": "b09fcaf4fbc265223c2479c410beef2b98c5e687c78ecedf3452aa12ffc465f1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/LICENSE"}], "id": "github:tongyi-mai/mobileworld", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T09:12:16Z", "licenses": {"code": {"basis": "Repository code license; self-hosted applications, model APIs and MCP services keep their own terms.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T09:14:09Z", "url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Tongyi-MAI/MobileWorld"}], "name": "MobileWorld (Tongyi-MAI)", "notes": "This is Tongyi-MAI/MobileWorld, distinct from jacklishufan/MobileWorld (MobileWorldBench). Public source includes task definitions, verification logic and environment orchestration. The README reports 201 tasks across 20 applications; this is not an independent count of the evolving task tree. Linux/KVM or WSL2/KVM is required by the documented host setup, with macOS support in progress. MCP and simulated-user settings can require separate external services.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Tongyi-MAI/MobileWorld/HEAD/README.md", "target_id": "arxiv:2512.19432", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A containerized Android evaluation environment with GUI-only, agent-user interaction and MCP-augmented tasks, plus checks on application and backend state.", "topics": ["evaluation", "mobile-navigation", "human-in-the-loop", "tool-integration"], "updated_at": null, "url": "https://github.com/Tongyi-MAI/MobileWorld"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "cfe9d8614b0b0558c37ba7c50371f6c47f119b914061ae815a9e3d796dd9ccfa", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:01:50Z", "role": "primary", "sha256": "7a4ca3daa6d011ff2d24d993191882d2b5545d8476c7eddf396581c0fef77496", "supports": ["identity"], "url": "https://api.github.com/repos/UbiquitousLearning/DroidCall"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "d0717d706359969376889eac73ea8764b659f41a2b4f3c1a409110267a912526", "supports": ["release_status"], "url": "https://api.github.com/repos/UbiquitousLearning/DroidCall/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:16Z", "role": "primary", "sha256": "2832cc5333f7166924a2823aa38a286d918b39b7fef2ca7af6521766c68a5303", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "08a6c5cfa655e1ecd774a41a9a815e845ad0afd47b289378bc1d6ee66c81dda9", "supports": ["relations"], "url": "https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:02:16Z", "role": "primary", "sha256": "2832cc5333f7166924a2823aa38a286d918b39b7fef2ca7af6521766c68a5303", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/LICENSE"}], "id": "github:ubiquitouslearning/droidcall", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:02:06Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T02:02:16Z", "url": "https://raw.githubusercontent.com/UbiquitousLearning/DroidCall/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/UbiquitousLearning/DroidCall"}], "name": "DroidCall generation and evaluation code", "notes": "Source utilities are published under MIT. This component operates through Android intent functions and is marked adjacent to visual GUI agents. Training data is separately released under its own card declaration; no checkpoint release is inferred from the fine-tuning instructions.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md", "target_id": "arxiv:2412.00402", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Data-generation, fine-tuning and evaluation utilities for teaching small language models to invoke Android intents through predefined functions.", "topics": ["tool-integration", "data-generation", "evaluation"], "updated_at": null, "url": "https://github.com/UbiquitousLearning/DroidCall"}
{"aliases": [], "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": {"disabled": false, "download_url": "https://huggingface.co/Tongyi-MAI/MAI-UI-8B/tree/main", "gated": false, "hub_id": "Tongyi-MAI/MAI-UI-8B"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "a1087d2a0dde6e7ba6d79db546ee6360a5ffd95ffbc3add1478eedec603a4187", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/Tongyi-MAI/MAI-UI-8B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:18Z", "role": "primary", "sha256": "349e6d5753fa94a47feb4b2fc53ecd6b9188b645a0cffabef92fd1aa0976ecad", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/Tongyi-MAI/MAI-UI-8B"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "a1087d2a0dde6e7ba6d79db546ee6360a5ffd95ffbc3add1478eedec603a4187", "supports": ["licenses.weights"], "url": "https://huggingface.co/Tongyi-MAI/MAI-UI-8B/raw/main/README.md"}], "id": "url:00b1a3abf9bcb424b264", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T04:07:19Z", "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": "2026-10-17T04:07:19Z", "url": "https://huggingface.co/Tongyi-MAI/MAI-UI-8B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/Tongyi-MAI/MAI-UI-8B"}], "name": "MAI-UI 8B", "notes": "Four public safetensors shards are present and Apache-2.0 is declared. The card presents family-wide results and device-cloud architecture; those claims are not all assigned to the 8B checkpoint. Other parameter sizes are not duplicated in this proposal batch.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/Tongyi-MAI/MAI-UI-8B/raw/main/README.md", "target_id": "arxiv:2512.22047", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A released MAI-UI checkpoint combining mobile GUI actions, agent-user interaction and tool-use-oriented navigation.", "topics": ["grounding", "mobile-navigation", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/Tongyi-MAI/MAI-UI-8B"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/Hcompany/Holo4-35B-A3B/tree/main", "gated": false, "hub_id": "Hcompany/Holo4-35B-A3B"}, "environments": ["desktop", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:11:45Z", "role": "primary", "sha256": "4f1aa5006aa9adaeb5ac549e5a5f8ea164352714a00528b1afb8602812d45b44", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/Hcompany/Holo4-35B-A3B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:11:45Z", "role": "primary", "sha256": "7a41bdbfb4a300302a7d3633f12962505d56ba874a37292afccbd48b1bec8287", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/Hcompany/Holo4-35B-A3B"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T04:11:45Z", "role": "primary", "sha256": "4f1aa5006aa9adaeb5ac549e5a5f8ea164352714a00528b1afb8602812d45b44", "supports": ["licenses.weights"], "url": "https://huggingface.co/Hcompany/Holo4-35B-A3B/raw/main/README.md"}], "id": "url:16083a0225a60882c704", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T04:11:45Z", "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": "2026-10-17T04:11:45Z", "url": "https://huggingface.co/Hcompany/Holo4-35B-A3B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/Hcompany/Holo4-35B-A3B"}], "name": "Holo4 35B-A3B BF16", "notes": "The public manifest contains 26 safetensors shards for the BF16 checkpoint and the card declares Apache-2.0. Quantized FP8/NVFP4/GGUF variants are not split into duplicate entries here. The card also contains demos and scores for Holo4-27B; those are not assigned to this 35B-A3B checkpoint. No training corpus or runtime was downloaded.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "A Qwen3.6-based computer-use checkpoint designed to request GUI actions, code execution and tool calls through an external agent harness.", "topics": ["grounding", "action-prediction", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/Hcompany/Holo4-35B-A3B"}
{"aliases": ["url:1ab38d1d6a4c46a8eb3e", "OpenMobile-2/OpenMobile-27B", "https://huggingface.co/OpenMobile-2/OpenMobile-27B"], "authors": [], "details": {"artifact_inspection": "Public file-manifest inspection only; weights were not downloaded.", "base_model": "Qwen/Qwen3.6-27B", "gated": false, "hub_revision": "3d844db8c20421524a5012837b151e444ad35ae6", "listed_weight_files": ["model-00001-of-00012.safetensors", "model-00002-of-00012.safetensors", "model-00003-of-00012.safetensors", "model-00004-of-00012.safetensors", "model-00005-of-00012.safetensors", "model-00006-of-00012.safetensors", "model-00007-of-00012.safetensors", "model-00008-of-00012.safetensors", "model-00009-of-00012.safetensors", "model-00010-of-00012.safetensors", "model-00011-of-00012.safetensors", "model-00012-of-00012.safetensors"]}, "environments": ["android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:51Z", "role": "primary", "sha256": "ba760ab75e997941dabf0845b34e5fb83daeb83a5cfa16b13929477cfc61a31c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:49:48Z", "role": "primary", "sha256": "c7e388efe1ecfd415b7b5af7f296e62646a0a1380037f1e57ad1a43c26463b9a", "supports": ["details.gated", "details.hub_revision", "details.listed_weight_files"], "url": "https://huggingface.co/api/models/OpenMobile-2/OpenMobile-2-27B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:46:06Z", "role": "primary", "sha256": "9b385e9f6aa2f7169f616a76887df3b76df605cb2bea071d99ee7ca3f9b659cc", "supports": ["environments"], "url": "https://os-copilot.github.io/OpenMobile2-Home/"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T01:48:51Z", "role": "primary", "sha256": "ba760ab75e997941dabf0845b34e5fb83daeb83a5cfa16b13929477cfc61a31c", "supports": ["licenses.weights"], "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B/raw/main/README.md"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:11Z", "role": "primary", "sha256": "f669181b7f1b92bbbcb8ab71f0d15d8c3cf93b3665d73d20e9b9f0dc3a36e158", "supports": ["aliases"], "url": "https://huggingface.co/api/models/OpenMobile-2/OpenMobile-27B"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:12Z", "role": "primary", "sha256": "f669181b7f1b92bbbcb8ab71f0d15d8c3cf93b3665d73d20e9b9f0dc3a36e158", "supports": ["aliases"], "url": "https://huggingface.co/api/models/OpenMobile-2/OpenMobile-2-27B"}], "id": "url:1611de0e6d606165d12d", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T01:48:51Z", "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": "2026-10-17T01:48:51Z", "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B"}], "name": "OpenMobile-2 27B", "notes": "The public Hub manifest lists weight files; the catalog did not download weights or reproduce inference. License declarations apply to this checkpoint separately from code and training data.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B/raw/main/README.md", "target_id": "github:openmobile2/openmobile2-code", "type": "variant_of"}, {"evidence_url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B/raw/main/README.md", "target_id": "url:546e6c95825d1318c98d", "type": "trained_on"}, {"evidence_url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B/raw/main/README.md", "target_id": "arxiv:2610.14157", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Mobile-agent checkpoint based on Qwen3.6-27B for GUI and app-native tool interaction.", "topics": ["mobile-navigation", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-27B"}
{"aliases": ["url:a778c3411858c38c0f39", "OpenMobile-2/OpenMobile-Data", "https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data"], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data-v2/raw/main/README.md", "population": "SFT collection", "qualifier": "exact", "unit": "trajectories", "value": 11643}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data-v2/raw/main/README.md", "population": "SFT collection", "qualifier": "exact", "unit": "decision steps", "value": 118582}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data-v2/raw/main/README.md", "population": "RL collection", "qualifier": "exact", "unit": "executable tasks", "value": 2475}], "evidence_urls": ["https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data-v2/raw/main/README.md"], "formats": ["JSON", "JSONL", "PNG"], "modalities": ["screenshots", "text", "tool calls"], "notes": "These are publisher-reported totals at three distinct grains, not values to sum. Earlier screenshots may be collapsed; each SFT example keeps at most three images. The card and repository disagree on the singular/plural RL filename, so consumers should inspect the actual file manifest.", "splits": ["sft/train", "rl/train"], "supervision": ["next-action supervision", "verifiable task outcomes"]}}, "environments": ["android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:51Z", "role": "primary", "sha256": "abacf9a10759a9e7a8eeec5a266c96ed83742bd9296a19be95d09c9717e18b7c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data-v2/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:10Z", "role": "primary", "sha256": "485279038266d9454df75e27a80a0a0d71c24e90700126d2dcfa0da921b1273d", "supports": ["relations", "environments"], "url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/README.md"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:18Z", "role": "primary", "sha256": "14b1211ed709147f4022f311c17303ca70faaffa22be143617eb431d966eec1a", "supports": ["aliases"], "url": "https://huggingface.co/api/datasets/OpenMobile-2/OpenMobile-Data"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:20Z", "role": "primary", "sha256": "14b1211ed709147f4022f311c17303ca70faaffa22be143617eb431d966eec1a", "supports": ["aliases"], "url": "https://huggingface.co/api/datasets/OpenMobile-2/OpenMobile-Data-v2"}], "id": "url:546e6c95825d1318c98d", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T01:48:51Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data-v2"}], "name": "OpenMobile-Data-v2", "notes": "The reviewed card contains no explicit license declaration, so data-license status remains unknown. The repository code license is not inherited. The current code README points to the v2 dataset; an older project-page URL was not counted as a separate asset.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OpenMobile2/OpenMobile2-Code/HEAD/README.md", "target_id": "arxiv:2610.14157", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Mobile-agent supervision from emulator and simulator trajectories, plus executable tasks with starting states and success checks for reinforcement learning.", "topics": ["mobile-navigation", "data-generation", "reinforcement-learning", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/datasets/OpenMobile-2/OpenMobile-Data-v2"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/wanlilll/WeaveBench/tree/main", "gated": false, "hub_id": "wanlilll/WeaveBench", "languages": ["en"], "modalities": [], "size_category_reported": ["n<1K"]}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "e65e5793fe5b251ef7793a0618f4e8139fe5c92c2cf8093d0cc6142e8ece9d02", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://huggingface.co/api/datasets/wanlilll/WeaveBench"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:38:27Z", "role": "primary", "sha256": "e65e5793fe5b251ef7793a0618f4e8139fe5c92c2cf8093d0cc6142e8ece9d02", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/wanlilll/WeaveBench"}], "id": "url:6a212bc12f82c824f6b9", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:38:27Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card", "declared": "mit", "retrieved_at": "2026-10-16T06:38:27Z", "url": "https://huggingface.co/datasets/wanlilll/WeaveBench"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/wanlilll/WeaveBench"}], "name": "WeaveBench", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Computer-use benchmark mixing graphical interaction and code execution.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/datasets/wanlilll/WeaveBench"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/mllmTeam/DroidCall/tree/main", "gated": false, "hub_id": "mllmTeam/DroidCall", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md"], "reason": "The Hub dataset card declares Apache-2.0 in its metadata header (license: apache-2.0). The companion repository's MIT code license is a separate declaration."}}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "08a6c5cfa655e1ecd774a41a9a815e845ad0afd47b289378bc1d6ee66c81dda9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.data"], "url": "https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "c7804987d56b03314a4c48e3ddf2e5819fc0d9c83cc65ba54a981a2f4a584dd9", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/mllmTeam/DroidCall"}], "id": "url:ae8f5208eea9e7f1f606", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:02:28Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T02:02:28Z", "url": "https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/mllmTeam/DroidCall"}], "name": "DroidCall data", "notes": "The release provides JSONL data and API annotations. This is Android intent-calling supervision rather than screenshot-and-click trajectories, so it is marked adjacent. The dataset card declares Apache-2.0; the companion repository’s MIT code license is a separate declaration.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/mllmTeam/DroidCall/raw/main/README.md", "target_id": "arxiv:2412.00402", "type": "documented_by"}], "review_status": "curated", "scope": "adjacent", "summary": "Instruction-to-function-call examples for learning Android intent invocation, with training, testing and API-annotation files for mobile task execution.", "topics": ["mobile-navigation", "tool-integration", "data-generation"], "updated_at": null, "url": "https://huggingface.co/datasets/mllmTeam/DroidCall"}
{"aliases": [], "authors": [], "details": {"artifact_inspection": "Public file-manifest inspection only; weights were not downloaded.", "base_model": "Qwen/Qwen3.5-9B", "gated": false, "hub_revision": "68ebbac4d5bdf725103a22d076122b8aa883a0f2", "listed_weight_files": ["model.safetensors"]}, "environments": ["android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:51Z", "role": "primary", "sha256": "09c59e6b35a5fe1eea528595f326546269ea8a2a75aa404f1f7950b9de5f30d1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:49:48Z", "role": "primary", "sha256": "1aab6befd309e54ba71676670039b4098362974ce0ce89b25872f137d50c103f", "supports": ["details.gated", "details.hub_revision", "details.listed_weight_files"], "url": "https://huggingface.co/api/models/OpenMobile-2/OpenMobile-2-9B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:46:06Z", "role": "primary", "sha256": "9b385e9f6aa2f7169f616a76887df3b76df605cb2bea071d99ee7ca3f9b659cc", "supports": ["environments"], "url": "https://os-copilot.github.io/OpenMobile2-Home/"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T01:48:51Z", "role": "primary", "sha256": "09c59e6b35a5fe1eea528595f326546269ea8a2a75aa404f1f7950b9de5f30d1", "supports": ["licenses.weights"], "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B/raw/main/README.md"}], "id": "url:d35224f787e7ec21ce36", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T01:48:51Z", "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": "2026-10-17T01:48:51Z", "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B"}], "name": "OpenMobile-2 9B", "notes": "The public Hub manifest lists weight files; the catalog did not download weights or reproduce inference. License declarations apply to this checkpoint separately from code and training data.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B/raw/main/README.md", "target_id": "github:openmobile2/openmobile2-code", "type": "variant_of"}, {"evidence_url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B/raw/main/README.md", "target_id": "url:546e6c95825d1318c98d", "type": "trained_on"}, {"evidence_url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B/raw/main/README.md", "target_id": "arxiv:2610.14157", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Mobile-agent checkpoint trained on OpenMobile-Data-v2 for screenshot actions and app-native tool calls.", "topics": ["mobile-navigation", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/OpenMobile-2/OpenMobile-2-9B"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://recreation-bench.cc/#leaderboard"}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "79deaa899d085ece77a127d53b12ddb9edea8a796943f35b75e9acec67bcb976", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/Qwen/RecreationBench/raw/main/README.md"}], "id": "url:d9d8e68f19ef0956d189", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:50:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/Qwen/RecreationBench"}], "name": "RecreationBench", "notes": "Dataset card declares multiple upstream licenses; do not treat the package as having one blanket permissive license.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Tasks for recreating reference applications with hybrid computer interaction.", "topics": ["evaluation", "tool-integration"], "updated_at": null, "url": "https://huggingface.co/datasets/Qwen/RecreationBench"}