{"aliases": [], "authors": ["Evan Zheran Liu", "Kelvin Guu", "Panupong Pasupat", "Tianlin Shi", "Percy Liang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "75d19fa97ce85a93bb9665a8ae06607bc0b90fde1c8376cdd60e99b55e90729c", "submitted_at": "2018-02-24T05:32:47Z", "updated_at": "2018-02-24T05:32:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/1802.08802v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:1802.08802", "identifiers": {"arxiv": "1802.08802"}, "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/1802.08802"}], "name": "Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2018-02-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses workflows extracted from demonstrations to constrain reinforcement-learning exploration for interactive web tasks.", "topics": ["reinforcement-learning", "workflow"], "updated_at": "2018-02-24", "url": "https://arxiv.org/abs/1802.08802"}
{"aliases": [], "authors": ["Daniel Toyama", "Philippe Hamel", "Anita Gergely", "Gheorghe Comanici", "Amelia Glaese", "Zafarali Ahmed", "Tyler Jackson", "Shibl Mourad", "Doina Precup"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7f0caf694223bc2569cdd01662cadf54cbc96cad370866a231c15437c5d5d0ba", "submitted_at": "2021-05-27T15:20:14Z", "updated_at": "2021-05-27T15:20:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2105.13231v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2105.13231", "identifiers": {"arxiv": "2105.13231"}, "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/2105.13231"}], "name": "AndroidEnv: A Reinforcement Learning Platform for Android", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2021-05-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents an Android environment for training and evaluating reinforcement-learning agents through touchscreen interaction.", "topics": ["reinforcement-learning", "sandbox"], "updated_at": "2021-05-27", "url": "https://arxiv.org/abs/2105.13231"}
{"aliases": [], "authors": ["Peter C Humphreys", "David Raposo", "Toby Pohlen", "Gregory Thornton", "Rachita Chhaparia", "Alistair Muldal", "Josh Abramson", "Petko Georgiev", "Alex Goldin", "Adam Santoro", "Timothy Lillicrap"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "94129847ac85f4a8aeab74907cc68a31c435b078ab950695d9328b6c3896b46d", "submitted_at": "2022-02-16T15:23:46Z", "updated_at": "2022-11-11T13:45:44Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2202.08137v2"}, "publication": {"authors": ["Peter Humphreys", "David Raposo", "Toby Pohlen", "Gregory Thornton", "Rachita Chhaparia", "Alistair Muldal", "Josh Abramson", "Petko Georgiev", "Adam Santoro", "Timothy Lillicrap"], "date_basis": "Explicit publisher metadata: 2022/06/28; precision=day.", "date_parts": [2022, 6, 28], "doi": null, "event_dates": "17--23 Jul 2022", "evidence_urls": ["https://proceedings.mlr.press/v162/humphreys22a.html", "https://proceedings.mlr.press/v162/humphreys22a/humphreys22a.pdf"], "identity_basis": "Same work: exact normalized landing title, corresponding author list, and matching specific contribution. Publisher and arXiv versions remain distinct.", "issue": null, "notes": "The formal ICML PDF lists 10 authors beginning Peter Humphreys and uses Toby Pohlen. The landing lists Peter C Humphreys and Tobias Pohlen. The retained arXiv snapshot lists 11 authors, including Alex Goldin. These are preserved as distinct version-specific author lists; the publication field uses only the formal PDF list. The publisher issue date comes from citation_publication_date; the conference date range is retained separately and does not replace it. Publisher issue date is 2022-06-28; the July 17–23 conference range is an event date, not a substitute publication day. Publisher-hosted proceedings record; distinct from the retained arXiv preprint snapshot.", "pages": "9466-9482", "pdf_url": "https://proceedings.mlr.press/v162/humphreys22a/humphreys22a.pdf", "publisher": "PMLR", "series": "Proceedings of Machine Learning Research", "title": "A Data-Driven Approach for Learning to Control Computers", "type": "conference-paper", "url": "https://proceedings.mlr.press/v162/humphreys22a.html", "venue": "Proceedings of the 39th International Conference on Machine Learning", "volume": "162"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:47+00:00", "role": "primary", "sha256": "b0ac6ec4bf3aad3773c53a0fffe8688a6287f25f28faf3743603ec2c1ae08a8b", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.event_dates", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.series", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.mlr.press/v162/humphreys22a.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:59:46+00:00", "role": "primary", "sha256": "ba2a9a0a5e46aa5f34cfc720ac08049253917407e2fbd4a8cdbd2c8ca3ec8455", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.title"], "url": "https://proceedings.mlr.press/v162/humphreys22a/humphreys22a.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2202.08137", "identifiers": {"arxiv": "2202.08137"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2202.08137"}, {"rel": "publication", "url": "https://proceedings.mlr.press/v162/humphreys22a.html"}, {"rel": "publication_pdf", "url": "https://proceedings.mlr.press/v162/humphreys22a/humphreys22a.pdf"}], "name": "A data-driven approach for learning to control computers", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2022-02-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines reinforcement learning with behavioral priors from human interactions to learn computer control through keyboard and mouse actions.", "topics": ["reinforcement-learning", "automation"], "updated_at": "2022-11-11", "url": "https://arxiv.org/abs/2202.08137"}
{"aliases": [], "authors": ["Zhizheng Zhang", "Wenxuan Xie", "Xiaoyi Zhang", "Yan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f389de80bb86a759df2cfc7000188286efcd73c99a43ede376c46f190832ef80", "submitted_at": "2023-10-07T07:22:41Z", "updated_at": "2023-10-07T07:22:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2310.04716v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2310.04716", "identifiers": {"arxiv": "2310.04716"}, "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/2310.04716"}], "name": "Reinforced UI Instruction Grounding: Towards a Generic UI Task Automation API", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-10-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains a screenshot-based instruction-grounding model to predict target coordinates, using reinforcement learning to improve spatial decoding.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2023-10-07", "url": "https://arxiv.org/abs/2310.04716"}
{"aliases": [], "authors": ["Hanyu Lai", "Xiao Liu", "Iat Long Iong", "Shuntian Yao", "Yuxuan Chen", "Pengbo Shen", "Hao Yu", "Hanchen Zhang", "Xiaohan Zhang", "Yuxiao Dong", "Jie Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "340206da25df688ff73da617fd66a872586d1e36981c431d18331cee9b9669e0", "submitted_at": "2024-04-04T17:58:40Z", "updated_at": "2024-10-12T05:05:38Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.03648v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.03648", "identifiers": {"arxiv": "2404.03648"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.03648"}], "name": "AutoWebGLM: A Large Language Model-based Web Navigating Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-04-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds a web agent through simplified HTML observations, human-and-AI browsing data, curriculum training, and reinforcement-learning refinement.", "topics": ["web-navigation", "reinforcement-learning", "data-generation"], "updated_at": "2024-10-12", "url": "https://arxiv.org/abs/2404.03648"}
{"aliases": [], "authors": ["Moghis Fereidouni", "Adib Mosharrof", "A. B. Siddique"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "127d037bb462ae4ed7bd72c46a62e2c4427611b30b67193454e40cc5c83cca3f", "submitted_at": "2024-04-16T20:15:32Z", "updated_at": "2025-01-26T19:07:41Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.10887v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.10887", "identifiers": {"arxiv": "2404.10887"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.10887"}], "name": "Grounded Language Agent for Product Search via Intelligent Web Interactions", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-04-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GLAINTEL, a Flan-T5-based product-search web agent studied under unsupervised learning, demonstration-based training, and unsupervised domain adaptation. It compares ways to combine demonstrations with reinforcement learning.", "topics": ["web-navigation", "reinforcement-learning", "imitation-learning"], "updated_at": "2025-01-26", "url": "https://arxiv.org/abs/2404.10887"}
{"aliases": [], "authors": ["Lucas-Andreï Thil", "Mirela Popa", "Gerasimos Spanakis"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3498a7bc289c6537ab69780a56a9fd0a2bb697fb74c077949824c4fc018762c6", "submitted_at": "2024-05-01T13:51:45Z", "updated_at": "2024-05-01T13:51:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2405.00516v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2405.00516", "identifiers": {"arxiv": "2405.00516"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2405.00516"}], "name": "Navigating WebAI: Training Agents to Complete Web Tasks with Large Language Models and Reinforcement Learning", "notes": "Reviewed against the retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2024-05-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines supervised and reinforcement learning for MiniWoB web tasks and investigates whether agents understand HTML structure or memorize target elements.", "topics": ["web-navigation", "reinforcement-learning", "evaluation"], "updated_at": "2024-05-01", "url": "https://arxiv.org/abs/2405.00516"}
{"aliases": [], "authors": ["Hao Bai", "Yifei Zhou", "Mert Cemri", "Jiayi Pan", "Alane Suhr", "Sergey Levine", "Aviral Kumar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2c5d4e75d6ba88574f6f033bb2c1cf44cf04149e94e017cd8a1a227e7a1c8238", "submitted_at": "2024-06-14T17:49:55Z", "updated_at": "2024-06-14T17:49:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2406.11896v1"}, "publication": {"authors": ["Hao Bai", "Yifei Zhou", "Mert Cemri", "Jiayi Pan", "Alane Suhr", "Sergey Levine", "Aviral Kumar"], "date_basis": "Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 12, 16], "doi": "10.52202/079017-0397", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/1704ddd0bb89f159dfe609b32c889995-Abstract-Conference.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "12461-12495", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/1704ddd0bb89f159dfe609b32c889995-Paper-Conference.pdf", "publisher": null, "series": null, "title": "DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/1704ddd0bb89f159dfe609b32c889995-Abstract-Conference.html", "venue": "Advances in Neural Information Processing Systems", "volume": "37"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:03Z", "role": "primary", "sha256": "0fd68a0f088591eb695e205286823c7be4059fba8a14e62030ca1b5b11707e21", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/1704ddd0bb89f159dfe609b32c889995-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2406.11896", "identifiers": {"arxiv": "2406.11896"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.11896"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/1704ddd0bb89f159dfe609b32c889995-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/1704ddd0bb89f159dfe609b32c889995-Paper-Conference.pdf"}], "name": "DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines offline initialization and online reinforcement learning in Android environments, with automated evaluation and curriculum-based sampling.", "topics": ["reinforcement-learning", "mobile-navigation"], "updated_at": "2024-06-14", "url": "https://arxiv.org/abs/2406.11896"}
{"aliases": [], "authors": ["Taiyi Wang", "Zhihao Wu", "Jianheng Liu", "Jianye Hao", "Jun Wang", "Kun Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a9e3d8350cd0be5d4f3e82b3b35a15fe89403cd602e8bb20b7cd783c7a7067a0", "submitted_at": "2024-10-18T18:19:56Z", "updated_at": "2025-02-21T16:23:59Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2410.14803v5"}, "publication": {"authors": ["Taiyi Wang", "Zhihao Wu", "Jianheng Liu", "Jianye HAO", "Jun Wang", "Kun Shao"], "date_basis": "Explicit publisher citation_publication_date: 2025-05-01. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 5, 1], "doi": null, "event_dates": null, "evidence_urls": ["https://proceedings.iclr.cc/paper_files/paper/2025/hash/b9e472cd579c83e2f6aa3459f46aac28-Abstract-Conference.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity. The publisher title uses singular Agent; the preprint uses Agents. The authors and DistRL contribution agree.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "74757-74782", "pdf_url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/b9e472cd579c83e2f6aa3459f46aac28-Paper-Conference.pdf", "publisher": null, "series": null, "title": "DistRL: An Asynchronous Distributed Reinforcement Learning Framework for On-Device Control Agent", "type": "conference-paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/b9e472cd579c83e2f6aa3459f46aac28-Abstract-Conference.html", "venue": "International Conference on Learning Representations", "volume": "2025"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:02Z", "role": "primary", "sha256": "f3c834388e958e41ea7a8419b354435c1c522ef256c00d3af7f72bebb89c7fdc", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/b9e472cd579c83e2f6aa3459f46aac28-Abstract-Conference.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:35Z", "role": "primary", "sha256": "b3e67984f59a1b05ccd946eb54e2d37b0bb78ab0ca796fbb6d544793dafea3e0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2410.00467%2C2410.00689%2C2410.02810%2C2410.02907%2C2410.05243%2C2410.06703%2C2410.08164%2C2410.09006%2C2410.09024%2C2410.11843%2C2410.11871%2C2410.11872%2C2410.13232%2C2410.13757%2C2410.13824%2C2410.13825%2C2410.13886%2C2410.14803%2C2410.15164%2C2410.16464%2C2410.17236%2C2410.17401%2C2410.17520%2C2410.17883%2C2410.18603%2C2410.18963%2C2410.18967%2C2410.19054%2C2410.19100%2C2410.19461%2C2410.19609%2C2410.20745%2C2410.22552%2C2410.23218%2C2410.24024%2C2411.00820%2C2411.02337%2C2411.02391%2C2411.04468%2C2411.04890%2C2411.06559%2C2411.07618%2C2411.10323%2C2411.13451%2C2411.13591%2C2411.15004%2C2411.17465%2C2411.18279%2C2412.00402%2C2412.01268&max_results=50"}], "id": "arxiv:2410.14803", "identifiers": {"arxiv": "2410.14803"}, "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/2410.14803"}, {"rel": "publication", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/b9e472cd579c83e2f6aa3459f46aac28-Abstract-Conference.html"}, {"rel": "publication_pdf", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/file/b9e472cd579c83e2f6aa3459f46aac28-Paper-Conference.pdf"}], "name": "DistRL: An Asynchronous Distributed Reinforcement Learning Framework for On-Device Control Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces asynchronous distributed reinforcement learning with centralized updates and decentralized interaction collection for device-control agents.", "topics": ["reinforcement-learning", "efficiency"], "updated_at": "2025-02-21", "url": "https://arxiv.org/abs/2410.14803"}
{"aliases": [], "authors": ["Xiao Liu", "Bo Qin", "Dongzhu Liang", "Guang Dong", "Hanyu Lai", "Hanchen Zhang", "Hanlin Zhao", "Iat Long Iong", "Jiadai Sun", "Jiaqi Wang", "Junjie Gao", "Junjun Shan", "Kangning Liu", "Shudan Zhang", "Shuntian Yao", "Siyi Cheng", "Wentao Yao", "Wenyi Zhao", "Xinghan Liu", "Xinyi Liu", "Xinying Chen", "Xinyue Yang", "Yang Yang", "Yifan Xu", "Yu Yang", "Yujia Wang", "Yulin Xu", "Zehan Qi", "Yuxiao Dong", "Jie Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7cfe1aa31818f06567aa3d619ad90ac4bdb9f9f9003a1a8a71294cd4f8ccf8dc", "submitted_at": "2024-10-28T17:05:10Z", "updated_at": "2024-10-28T17:05:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2411.00820v1"}}, "environments": ["web", "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:2411.00820", "identifiers": {"arxiv": "2411.00820"}, "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/2411.00820"}], "name": "AutoGLM: Autonomous Foundation Agents for GUIs", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents AutoGLM agents for browser and phone control, separating planning from grounding and training through an evolving interaction curriculum.", "topics": ["automation", "reinforcement-learning", "grounding"], "updated_at": "2024-10-28", "url": "https://arxiv.org/abs/2411.00820"}
{"aliases": [], "authors": ["Zehan Qi", "Xiao Liu", "Iat Long Iong", "Hanyu Lai", "Xueqiao Sun", "Wenyi Zhao", "Yu Yang", "Xinyue Yang", "Jiadai Sun", "Shuntian Yao", "Tianjie Zhang", "Wei Xu", "Jie Tang", "Yuxiao Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db039a2592cd79a1f8f7f1c985c2ddf42447c63d1f58d2815ba1d6ef2369da47", "submitted_at": "2024-11-04T17:59:58Z", "updated_at": "2025-01-27T11:56:15Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2411.02337v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2411.02337", "identifiers": {"arxiv": "2411.02337"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.02337"}], "name": "WebRL: Training LLM Web Agents via Self-Evolving Online Curriculum Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-11-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains web agents with an evolving task curriculum, outcome supervision, and reinforcement-learning updates intended to manage policy drift.", "topics": ["reinforcement-learning", "web-navigation"], "updated_at": "2025-01-27", "url": "https://arxiv.org/abs/2411.02337"}
{"aliases": [], "authors": ["Zhirui Deng", "Zhicheng Dou", "Yutao Zhu", "Ji-Rong Wen", "Ruibin Xiong", "Mang Wang", "Weipeng Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dd631a31f72dabe3a3bebf322031f9d686082073a8e4aa71eb65745498327f4e", "submitted_at": "2024-11-06T10:35:11Z", "updated_at": "2024-12-09T09:20:11Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2411.03817v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:41Z", "role": "primary", "sha256": "0ba893415123ecc1763614ef5fc97e5c8bb7835f10b26b43f1b921ea860484cf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2410.06212%2C2410.06215%2C2410.07283%2C2410.07331%2C2410.07869%2C2410.08209%2C2410.08368%2C2410.10570%2C2410.12481%2C2410.12669%2C2410.15549%2C2410.16950%2C2410.21909%2C2410.22584%2C2410.22770%2C2410.23022%2C2410.23555%2C2411.02006%2C2411.02393%2C2411.02479%2C2411.03312%2C2411.03817%2C2411.05451%2C2411.11145%2C2411.11904%2C2411.11916%2C2411.12990%2C2411.14228%2C2411.15397%2C2412.01857%2C2412.04144%2C2412.04383%2C2412.05819%2C2412.06224%2C2412.06771%2C2412.09530%2C2412.09613%2C2412.10047%2C2412.10380%2C2412.10673%2C2412.12004%2C2412.13178%2C2412.13810%2C2412.14203%2C2412.14222%2C2412.16359%2C2412.19638%2C2412.20206%2C2501.03895%2C2501.05179&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:13:28Z", "role": "primary", "sha256": "b105dc6cc1b32ab71467b2701b6a85506f698e73f04cfbd63ab41eddfbca48b6", "supports": ["name", "kind", "summary", "topics", "notes"], "url": "https://arxiv.org/abs/2411.03817v3"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:15:31Z", "role": "primary", "sha256": "ce801ec21577b4f62d5b0ecec7c50e87744e5b38db612a2aa4017e27fe37102f", "supports": ["summary", "environments", "scope", "notes"], "url": "https://arxiv.org/html/2411.03817v3"}], "id": "arxiv:2411.03817", "identifiers": {"arxiv": "2411.03817"}, "kind": "paper", "last_verified_at": "2026-10-17T09:13:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.03817"}], "name": "From Novice to Expert: LLM Agent Policy Optimization via Step-wise Reinforcement Learning", "notes": "Reviewed against the exact-version official arXiv abstract page; its normalized abstract SHA-256 matches the retained bibliographic snapshot. The original API response body was not locally available for this review. Experimental results and artifact availability have not been independently reproduced. Official full-paper experimental sections explicitly evaluate WebShop and Mind2Web; the overall optimization method also covers embodied and question-answering tasks.", "published_at": "2024-11-06", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Introduces StepAgent, which compares agent and expert actions to supply intermediate rewards for policy learning, with implicit-reward and inverse-RL variants. Evaluation includes WebShop and Mind2Web alongside non-web tasks.", "topics": ["reinforcement-learning", "reward-models", "self-reflection"], "updated_at": "2024-12-09", "url": "https://arxiv.org/abs/2411.03817"}
{"aliases": [], "authors": ["Yifei Zhou", "Qianlan Yang", "Kaixiang Lin", "Min Bai", "Xiong Zhou", "Yu-Xiong Wang", "Sergey Levine", "Erran Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "362d2a470603bece43d2205e56dd633e7310cfad2ba2d91b0b765abcbcc41b2f", "submitted_at": "2024-12-17T18:59:50Z", "updated_at": "2024-12-17T18:59:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2412.13194v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2412.13194", "identifiers": {"arxiv": "2412.13194"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.13194"}], "name": "Proposer-Agent-Evaluator(PAE): Autonomous Skill Discovery For Foundation Model Internet Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-12-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines autonomous task proposal, web interaction, and visual success evaluation to discover skills and improve browser-agent policies.", "topics": ["data-generation", "reinforcement-learning", "web-navigation"], "updated_at": "2024-12-17", "url": "https://arxiv.org/abs/2412.13194"}
{"aliases": [], "authors": ["Qingyuan Wu", "Jianheng Liu", "Jianye Hao", "Jun Wang", "Kun Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "186b7fd8e79db290ea481645d8adab3989a12f787a13ecb11270f4756f5af42a", "submitted_at": "2025-02-11T20:57:46Z", "updated_at": "2025-05-20T19:54:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2502.07949v2"}}, "environments": ["web", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2502.07949", "identifiers": {"arxiv": "2502.07949"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.07949"}], "name": "Advancing Autonomous VLM Agents via Variational Subgoal-Conditioned 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": "2025-02-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses variational subgoal-conditioned reinforcement learning to improve learning efficiency on sparse-reward mobile and web-control tasks.", "topics": ["reinforcement-learning", "planning"], "updated_at": "2025-05-20", "url": "https://arxiv.org/abs/2502.07949"}
{"aliases": [], "authors": ["Hao Bai", "Yifei Zhou", "Li Erran Li", "Sergey Levine", "Aviral Kumar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "99492dab54d9816104d47fca525277b033cadb7a53727dae06ed1a3221fcb416", "submitted_at": "2025-02-13T18:55:14Z", "updated_at": "2025-02-13T18:55:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2502.15760v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}], "id": "arxiv:2502.15760", "identifiers": {"arxiv": "2502.15760"}, "kind": "paper", "last_verified_at": "2026-10-16T06:45:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.15760"}], "name": "Digi-Q: Learning Q-Value Functions for Training Device-Control Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Learns action-value functions from offline device-control experience and uses them to rank candidate actions for policy improvement.", "topics": ["reinforcement-learning", "efficiency"], "updated_at": "2025-02-13", "url": "https://arxiv.org/abs/2502.15760"}
{"aliases": [], "authors": ["Jiani Zheng", "Lu Wang", "Fangkai Yang", "Chaoyun Zhang", "Lingrui Mei", "Wenjie Yin", "Qingwei Lin", "Dongmei Zhang", "Saravan Rajmohan", "Qi Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6d69b006a298a591aacfe7f29df63f508113c0ec824ede82a0bb5bbb9cb66266", "submitted_at": "2025-02-26T07:52:02Z", "updated_at": "2025-02-26T07:52:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2502.18906v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "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:2502.18906", "identifiers": {"arxiv": "2502.18906"}, "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/2502.18906"}], "name": "VEM: Environment-Free Exploration for Training GUI Agent with Value Environment Model", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains GUI agents with a frozen value model learned from offline interactions, guiding action exploration without predicting complete next observations.", "topics": ["reinforcement-learning", "world-models"], "updated_at": "2025-02-26", "url": "https://arxiv.org/abs/2502.18906"}
{"aliases": [], "authors": ["Fanbin Lu", "Zhisheng Zhong", "Ziqin Wei", "Shu Liu", "Chi-Wing Fu", "Jiaya Jia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e883289bafee01759a0f4f3f7051cbbebdcd8059e7520d68c25ba313b8da69a6", "submitted_at": "2025-03-16T14:53:43Z", "updated_at": "2025-03-24T16:33:28Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2503.12532v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2503.12532", "identifiers": {"arxiv": "2503.12532"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.12532"}], "name": "STEVE: A Step Verification Pipeline for Computer-use Agent Training", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-03-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Labels individual computer-use actions by comparing surrounding screenshots, then trains agents from both successful and unsuccessful steps.", "topics": ["data-generation", "evaluation", "reinforcement-learning"], "updated_at": "2025-03-24", "url": "https://arxiv.org/abs/2503.12532"}
{"aliases": [], "authors": ["Zhengxi Lu", "Yuxiang Chai", "Yaxuan Guo", "Xi Yin", "Liang Liu", "Hao Wang", "Han Xiao", "Shuai Ren", "Guanjing Xiong", "Hongsheng Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b794a2f59cbe26a889b242ff47a7cd982f6367f759ddabcf99d96f31e5c122da", "submitted_at": "2025-03-27T15:39:30Z", "updated_at": "2025-05-24T08:46:08Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2503.21620v5"}}, "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:2503.21620", "identifiers": {"arxiv": "2503.21620"}, "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/2503.21620"}], "name": "UI-R1: Enhancing Efficient Action Prediction of GUI Agents by 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": "2025-03-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses rule-based action rewards and policy optimization to train GUI action prediction from a compact set of mobile tasks.", "topics": ["reinforcement-learning", "grounding"], "updated_at": "2025-05-24", "url": "https://arxiv.org/abs/2503.21620"}
{"aliases": [], "authors": ["Run Luo", "Lu Wang", "Wanwei He", "Longze Chen", "Jiaming Li", "Xiaobo Xia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b875617b795e0f8d7d6348e22ec85ccc9ab4dd931e66850182348927b1e5209f", "submitted_at": "2025-04-14T17:45:54Z", "updated_at": "2025-10-01T04:55:20Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2504.10458v4"}}, "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:2504.10458", "identifiers": {"arxiv": "2504.10458"}, "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/2504.10458"}], "name": "GUI-R1 : A Generalist R1-Style Vision-Language Action Model For GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Applies reinforcement fine-tuning with unified action rules and curated cross-platform examples to vision-language GUI agents.", "topics": ["reinforcement-learning", "grounding"], "updated_at": "2025-10-01", "url": "https://arxiv.org/abs/2504.10458"}
{"aliases": [], "authors": ["Yuhang Liu", "Pengxiang Li", "Congkai Xie", "Xavier Hu", "Xiaotian Han", "Shengyu Zhang", "Hongxia Yang", "Fei Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4ce03185cc73c2723f444bd9729c67da9aab7d96b5ef099fb78e47f58790bc18", "submitted_at": "2025-04-19T09:25:55Z", "updated_at": "2025-04-19T09:25:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2504.14239v1"}}, "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:2504.14239", "identifiers": {"arxiv": "2504.14239"}, "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/2504.14239"}], "name": "InfiGUI-R1: Advancing Multimodal GUI Agents from Reactive Actors to Deliberative Reasoners", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses a staged reasoning-focused training framework to move GUI agents from reactive actions toward deliberative planning and recovery.", "topics": ["planning", "reinforcement-learning"], "updated_at": "2025-04-19", "url": "https://arxiv.org/abs/2504.14239"}
{"aliases": [], "authors": ["Jiahao Li", "Kaer Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5040073e7c278734df219d71d94e236468468621f5971ef3ccbc97b94e1792b9", "submitted_at": "2025-04-29T06:55:15Z", "updated_at": "2025-05-13T01:59:29Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.20464v2"}}, "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:2504.20464", "identifiers": {"arxiv": "2504.20464"}, "kind": "survey", "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/2504.20464"}], "name": "A Survey on GUI Agents with Foundation Models Enhanced by Reinforcement Learning", "notes": "", "published_at": "2025-04-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Formalizes GUI tasks as Markov decision processes and traces agent training from prompting through supervised fine-tuning to reinforcement learning.", "topics": ["survey", "reinforcement-learning"], "updated_at": "2025-05-13", "url": "https://arxiv.org/abs/2504.20464"}
{"aliases": [], "authors": ["Kun Huang", "Weikai Xu", "Yuxuan Liu", "Quandong Wang", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Bin Wang", "Bo An"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9bc92029e16b5afb4b54837d2ab2304c7b8176548061c9deafdca101754128dc", "submitted_at": "2025-05-18T08:28:05Z", "updated_at": "2026-03-21T09:25:27Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2505.12299v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2505.12299", "identifiers": {"arxiv": "2505.12299"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.12299"}], "name": "MobileIPL: Enhancing Mobile Agents Thinking Process via Iterative Preference Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileIPL samples trees of action-planning thoughts, propagates rule-based leaf rewards backward, and forms preference pairs for reasoning-level optimization. It also expands screenshot-based instruction data through staged evolution before evaluation on mobile GUI-agent benchmarks.", "topics": ["mobile-navigation", "reinforcement-learning", "planning", "data-generation"], "updated_at": "2026-03-21", "url": "https://arxiv.org/abs/2505.12299"}
{"aliases": [], "authors": ["Xinbin Yuan", "Jian Zhang", "Kaixin Li", "Zhuoxuan Cai", "Lujian Yao", "Jie Chen", "Enguang Wang", "Qibin Hou", "Jinwei Chen", "Peng-Tao Jiang", "Bo Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "755f885b79e9f3eef9b4cbdb8593fb3cde1ac8a92f86df52d5d4203eb6cad039", "submitted_at": "2025-05-18T11:22:04Z", "updated_at": "2025-05-24T03:01:02Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.12370v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.12370", "identifiers": {"arxiv": "2505.12370"}, "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/2505.12370"}], "name": "Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Proposes GUI grounding training that combines curated seed examples, continuous location-dependent policy feedback, and iterative refinement guided by attention maps. The study evaluates the resulting reinforcement-learning procedure on three grounding benchmarks, including ScreenSpot-Pro.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2025-05-24", "url": "https://arxiv.org/abs/2505.12370"}
{"aliases": [], "authors": ["Longxi Gao", "Li Zhang", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Mengwei Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "10d826ea7715ab7054522a9dcba622d5dae3e6c3faee421e4b168e16104d184b", "submitted_at": "2025-05-18T16:34:30Z", "updated_at": "2025-10-10T09:28:53Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2505.12493v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "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:2505.12493", "identifiers": {"arxiv": "2505.12493"}, "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/2505.12493"}], "name": "GUI-Shift: Enhancing VLM-Based GUI Agents through Self-supervised Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-Shift trains an inverse-dynamics task that predicts the initial action connecting two GUI states, using trajectories without natural-language instructions. It combines rule-based reinforcement learning with data filtering and evaluates transfer to Android automation and screenshot grounding.", "topics": ["reinforcement-learning", "data-generation", "grounding", "mobile-navigation"], "updated_at": "2025-10-10", "url": "https://arxiv.org/abs/2505.12493"}
{"aliases": [], "authors": ["Hyunseok Lee", "Jeonghoon Kim", "Beomjun Kim", "Jihoon Tack", "Chansong Jo", "Jaehong Lee", "Cheonbok Park", "Sookyo In", "Jinwoo Shin", "Kang Min Yoo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9acb7a88a0cf07427c1878c896dec920852810879c30bad8692358f2e2016539", "submitted_at": "2025-05-21T08:36:18Z", "updated_at": "2025-05-24T15:08:35Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.15259v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.15259", "identifiers": {"arxiv": "2505.15259"}, "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/2505.15259"}], "name": "ReGUIDE: Data Efficient GUI Grounding via Spatial Reasoning and Search", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines self-generated localization reasoning, spatial consistency checks, and inference-time search to improve data-efficient web grounding.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2025-05-24", "url": "https://arxiv.org/abs/2505.15259"}
{"aliases": [], "authors": ["Hyungjoo Chae", "Sunghwan Kim", "Junhee Cho", "Seungone Kim", "Seungjun Moon", "Gyeom Hwangbo", "Dongha Lim", "Minjin Kim", "Yeonjun Hwang", "Minju Gwak", "Dongwook Choi", "Minseok Kang", "Gwanhoon Im", "ByeongUng Cho", "Hyojun Kim", "Jun Hee Han", "Taeyoon Kwon", "Minju Kim", "Beong-woo Kwak", "Dongjin Kang", "Jinyoung Yeo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "440abdc75c0a51478cfedc6ee0a9196ccff641ec5a23849ba56be43448bab15f", "submitted_at": "2025-05-21T08:56:55Z", "updated_at": "2025-11-25T21:34:57Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.15277v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.15277", "identifiers": {"arxiv": "2505.15277"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.15277"}], "name": "Web-Shepherd: Advancing PRMs for Reinforcing Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Web-Shepherd is a process reward model for assessing individual steps of web-navigation trajectories. It introduces a collection of 40,000 step-level preference pairs with checklists and a separate benchmark for comparing reward models, then evaluates verification during WebArena-lite navigation.", "topics": ["reinforcement-learning", "evaluation", "data-generation", "web-navigation"], "updated_at": "2025-11-25", "url": "https://arxiv.org/abs/2505.15277"}
{"aliases": [], "authors": ["Yuqi Zhou", "Sunhao Dai", "Shuai Wang", "Kaiwen Zhou", "Qinglin Jia", "Jun Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "321f2d627451d60e130c17574994648fae69fc784b0ed03f6560e7f1ef9eb023", "submitted_at": "2025-05-21T17:59:09Z", "updated_at": "2025-05-22T11:15:23Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.15810v2"}}, "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:2505.15810", "identifiers": {"arxiv": "2505.15810"}, "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/2505.15810"}], "name": "GUI-G1: Understanding R1-Zero-Like Training for Visual Grounding in GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "Analyzes reasoning templates, reward design, and policy updates in GUI grounding, then proposes targeted adjustments to the training process.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2025-05-22", "url": "https://arxiv.org/abs/2505.15810"}
{"aliases": [], "authors": ["Fanbin Lu", "Zhisheng Zhong", "Shu Liu", "Chi-Wing Fu", "Jiaya Jia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c092569d4d79026ddbdcc3bc9da1cfc617ff44886dab0b36f311cc81083e0fb5", "submitted_at": "2025-05-22T06:24:32Z", "updated_at": "2025-05-22T06:24:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.16282v1"}}, "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:2505.16282", "identifiers": {"arxiv": "2505.16282"}, "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/2505.16282"}], "name": "ARPO:End-to-End Policy Optimization for GUI Agents with Experience Replay", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "ARPO extends group-relative policy optimization with a replay buffer that reuses successful GUI trajectories across training iterations. It filters tasks using baseline performance and compares end-to-end reinforcement learning with offline preference optimization on OSWorld.", "topics": ["reinforcement-learning", "long-horizon"], "updated_at": "2025-05-22", "url": "https://arxiv.org/abs/2505.16282"}
{"aliases": [], "authors": ["Zhepei Wei", "Wenlin Yao", "Yao Liu", "Weizhi Zhang", "Qin Lu", "Liang Qiu", "Changlong Yu", "Puyang Xu", "Chao Zhang", "Bing Yin", "Hyokun Yun", "Lihong Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f311fb50b36c234787e652b5689246460153a8f442e83b312048cdf6a4fb1502", "submitted_at": "2025-05-22T09:07:43Z", "updated_at": "2025-10-08T21:16:43Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.16421v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.16421", "identifiers": {"arxiv": "2505.16421"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.16421"}], "name": "WebAgent-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains web agents through asynchronous multi-turn environment interaction, using task-success rewards to optimize complete trajectories.", "topics": ["reinforcement-learning", "web-navigation"], "updated_at": "2025-10-08", "url": "https://arxiv.org/abs/2505.16421"}
{"aliases": [], "authors": ["Danyang Zhang", "Situo Zhang", "Ziyue Yang", "Zichen Zhu", "Zihan Zhao", "Ruisheng Cao", "Lu Chen", "Kai Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "40cbdc37216c58acd1b7523c48ad735df9bde485405a4e7b439aec85730c5bc0", "submitted_at": "2025-05-23T17:23:11Z", "updated_at": "2025-05-23T17:23:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.18121v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.18121", "identifiers": {"arxiv": "2505.18121"}, "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/2505.18121"}], "name": "ProgRM: Build Better GUI Agents with Progress Rewards", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "ProgRM predicts per-step task completion progress to provide intermediate rewards during online GUI-agent training. It derives progress labels by using longest common subsequences to identify key trajectory steps, allowing useful partial progress to contribute even when a trajectory ultimately fails.", "topics": ["reinforcement-learning", "data-generation"], "updated_at": "2025-05-23", "url": "https://arxiv.org/abs/2505.18121"}
{"aliases": [], "authors": ["Runliang Niu", "Jinglong Ji", "Yi Chang", "Qi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bb782466eccc7a271c89a4161c7b142f93f8f0d585cad39a1ccad048cf514426", "submitted_at": "2025-05-25T11:13:03Z", "updated_at": "2025-05-25T11:13:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.19095v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:47Z", "role": "primary", "sha256": "27144160e8981bb9f36b60c391689774913ff3f026eeeffc2923154c404a6529", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2504.03206%2C2504.03601%2C2504.06821%2C2504.09723%2C2504.11301%2C2504.11543%2C2504.12609%2C2504.14233%2C2504.14348%2C2504.14603%2C2504.16073%2C2504.16419%2C2504.17892%2C2504.18580%2C2504.20965%2C2505.00019%2C2505.00693%2C2505.01578%2C2505.02094%2C2505.03570%2C2505.03574%2C2505.05026%2C2505.05849%2C2505.06507%2C2505.07003%2C2505.07473%2C2505.10468%2C2505.10887%2C2505.12053%2C2505.12490%2C2505.12728%2C2505.13291%2C2505.14030%2C2505.14454%2C2505.14646%2C2505.15801%2C2505.15816%2C2505.17020%2C2505.18746%2C2505.19095%2C2505.19713%2C2505.20023%2C2505.20732%2C2505.20795%2C2505.21055%2C2505.22304%2C2505.22429%2C2505.23450%2C2505.23559%2C2506.00641&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:24Z", "role": "primary", "sha256": "645bd2ecbc013dab3cbc601d5a42011d8f7306915ea2ae06baf0767b3962e9ed", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2505.19095v1"}], "id": "arxiv:2505.19095", "identifiers": {"arxiv": "2505.19095"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.19095"}], "name": "ScreenExplorer: Training a Vision-Language Model for Diverse Exploration in Open GUI World", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains ScreenExplorer with reinforcement learning in dynamic GUI environments, using a world-model curiosity reward and experience-stream distillation to support sustained exploration.", "topics": ["reinforcement-learning", "world-models", "data-generation"], "updated_at": "2025-05-25", "url": "https://arxiv.org/abs/2505.19095"}
{"aliases": [], "authors": ["Han Xiao", "Guozhi Wang", "Yuxiang Chai", "Zimu Lu", "Weifeng Lin", "Hao He", "Lue Fan", "Liuyang Bian", "Rui Hu", "Liang Liu", "Shuai Ren", "Yafei Wen", "Xiaoxin Chen", "Aojun Zhou", "Hongsheng Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6f19671a527de76b702e68311656e42feea88249716f68911a41ec80be205e4c", "submitted_at": "2025-05-27T17:58:06Z", "updated_at": "2025-05-27T17:58:06Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.21496v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.21496", "identifiers": {"arxiv": "2505.21496"}, "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/2505.21496"}], "name": "UI-Genie: A Self-Improving Approach for Iteratively Boosting MLLM-based 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": "2025-05-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-Genie jointly improves a mobile GUI agent and a reward model that processes screenshots with interaction history and supplies action- and task-level feedback. Its data pipeline uses rule-based verification, corrupted trajectories, and hard negatives to support reward-guided exploration and iterative training.", "topics": ["data-generation", "reinforcement-learning", "mobile-navigation", "interaction-history"], "updated_at": "2025-05-27", "url": "https://arxiv.org/abs/2505.21496"}
{"aliases": [], "authors": ["Yuchen Zhuang", "Di Jin", "Jiaao Chen", "Wenqi Shi", "Hanrui Wang", "Chao Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9d260f52763eac186e6f0ff614c818f17602658b6be27cfecc3fea9d80579c9a", "submitted_at": "2025-05-28T23:45:28Z", "updated_at": "2025-06-08T07:57:08Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.22942v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2505.22942", "identifiers": {"arxiv": "2505.22942"}, "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/2505.22942"}], "name": "WorkForceAgent-R1: Incentivizing Reasoning Capability in LLM-based Web Agents via Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses rule-based rewards for output structure and action correctness to train reasoning for enterprise web navigation.", "topics": ["reinforcement-learning", "workflow"], "updated_at": "2025-06-08", "url": "https://arxiv.org/abs/2505.22942"}
{"aliases": [], "authors": ["Chenyu Yang", "Shiqian Su", "Shi Liu", "Xuan Dong", "Yue Yu", "Weijie Su", "Xuehui Wang", "Zhaoyang Liu", "Jinguo Zhu", "Hao Li", "Wenhai Wang", "Yu Qiao", "Xizhou Zhu", "Jifeng Dai"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "799e7b387f7d281f0aa9b182f37b215e3baca2041a0a0f6d414ddf1cec0de003", "submitted_at": "2025-05-29T17:59:51Z", "updated_at": "2025-05-29T17:59:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.23762v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2505.23762", "identifiers": {"arxiv": "2505.23762"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.23762"}], "name": "ZeroGUI: Automating Online GUI Learning at Zero Human Cost", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines generated tasks, estimated success rewards, and online reinforcement learning to automate GUI-agent interaction training.", "topics": ["reinforcement-learning", "data-generation"], "updated_at": "2025-05-29", "url": "https://arxiv.org/abs/2505.23762"}
{"aliases": [], "authors": ["Zhong Zhang", "Yaxi Lu", "Yikun Fu", "Yupeng Huo", "Shenzhi Yang", "Yesai Wu", "Han Si", "Xin Cong", "Haotian Chen", "Yankai Lin", "Jie Xie", "Wei Zhou", "Wang Xu", "Yuanheng Zhang", "Zhou Su", "Zhongwu Zhai", "Xiaoming Liu", "Yudong Mei", "Jianming Xu", "Hongyan Tian", "Chongyi Wang", "Chi Chen", "Yuan Yao", "Zhiyuan Liu", "Maosong Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3c0b2a065898ba4f73fcb5e5b2e92d723b0b55ac6caa46d5bf02d44eb7a2cdd9", "submitted_at": "2025-06-02T07:30:29Z", "updated_at": "2025-06-17T02:57:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.01391v2"}, "publication": {"authors": ["Zhong Zhang", "Yaxi Lu", "Yikun Fu", "Yupeng Huo", "Shenzhi Yang", "Yesai Wu", "Han Si (司函)", "Xin Cong", "Haotian Chen", "Yankai Lin (林衍凯)", "Jie Xie", "Wei Zhou", "Wang Xu", "Yuanheng Zhang", "Zhou Su", "Zhongwu Zhai", "Xiaoming Liu", "Yudong Mei", "Jianming Xu", "Hongyan Tian", "Chongyi Wang", "Chi Chen", "Yuan Yao", "Zhiyuan Liu", "Maosong Sun (孙茂松)"], "date_basis": "Explicit publisher citation_publication_date: 2025/11. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 11], "doi": "10.18653/v1/2025.emnlp-demos.12", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.emnlp-demos.12/"], "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": "155-180", "pdf_url": "https://aclanthology.org/2025.emnlp-demos.12.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "AgentCPM-GUI: Building Mobile-Use Agents with Reinforcement Fine-Tuning", "type": "conference-paper", "url": "https://aclanthology.org/2025.emnlp-demos.12/", "venue": "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: System Demonstrations", "volume": null}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:59Z", "role": "primary", "sha256": "5de5a828e06a63c1627b8ad9f9a46c1643d8b4b7d8ed6156fd9790b6bd083155", "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.emnlp-demos.12/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2506.01391", "identifiers": {"arxiv": "2506.01391"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.01391"}, {"rel": "publication", "url": "https://aclanthology.org/2025.emnlp-demos.12/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.emnlp-demos.12.pdf"}], "name": "AgentCPM-GUI: Building Mobile-Use Agents with Reinforcement Fine-Tuning", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2025-06-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines grounding pretraining, bilingual interaction trajectories, and reinforcement fine-tuning for mobile GUI actions.", "topics": ["reinforcement-learning", "mobile-navigation"], "updated_at": "2025-06-17", "url": "https://arxiv.org/abs/2506.01391"}
{"aliases": [], "authors": ["Junhong Shen", "Hao Bai", "Lunjun Zhang", "Yifei Zhou", "Amrith Setlur", "Shengbang Tong", "Diego Caples", "Nan Jiang", "Tong Zhang", "Ameet Talwalkar", "Aviral Kumar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e4b6c8b297e88dae689ba2a48724a441698cd5309a3bbe1198f38343e88aceb1", "submitted_at": "2025-06-09T17:50:02Z", "updated_at": "2025-06-10T12:50:18Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.07976v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2506.07976", "identifiers": {"arxiv": "2506.07976"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.07976"}], "name": "Thinking vs. Doing: Agents that Reason by Scaling Test-Time Interaction", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Test-Time Interaction scales the interaction horizon so web agents can explore, backtrack and revise plans within a rollout. Its online reinforcement-learning curriculum adjusts rollout lengths and is evaluated on WebVoyager and WebArena.", "topics": ["reinforcement-learning", "web-navigation", "planning", "long-horizon"], "updated_at": "2025-06-10", "url": "https://arxiv.org/abs/2506.07976"}
{"aliases": [], "authors": ["Jiaqi Tang", "Yu Xia", "Yi-Feng Wu", "Yuwei Hu", "Yuhui Chen", "Qing-Guo Chen", "Xiaogang Xu", "Xiangyu Wu", "Hao Lu", "Yanqing Ma", "Shiyin Lu", "Qifeng Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b00deba927bda03c079621f70decdeb946ec51a6982531a2086c43033ac9925c", "submitted_at": "2025-06-11T03:43:30Z", "updated_at": "2026-04-21T11:50:18Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2506.09373v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.09373", "identifiers": {"arxiv": "2506.09373"}, "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/2506.09373"}], "name": "LPO: Towards Accurate GUI Agent Interaction via Location Preference Optimization", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Location Preference Optimization uses an information-entropy signal to select interaction regions and a distance-based reward to assess click locations. It trains GUI localization through preference optimization and evaluates both offline grounding and online interaction.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2026-04-21", "url": "https://arxiv.org/abs/2506.09373"}
{"aliases": [], "authors": ["Jinjie Wei", "Jiyao Liu", "Lihao Liu", "Ming Hu", "Junzhi Ning", "Mingcheng Li", "Weijie Yin", "Junjun He", "Xiao Liang", "Chao Feng", "Dingkang Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f12d2ae58243b5f89a6a510e8f3ef4073a605c4bafdf63c37fd38db57ea84602", "submitted_at": "2025-06-22T06:30:52Z", "updated_at": "2026-07-21T05:30:48Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.17913v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.17913", "identifiers": {"arxiv": "2506.17913"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.17913"}], "name": "Learning, Reasoning, Refinement: A Framework for Kahneman's Dual-System Intelligence in GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "CogniGUI combines hierarchical visual parsing of interface elements with a GRPO-based grounding agent that compares alternative interaction paths. Its ScreenSeek benchmark evaluates navigation across applications, changing interface states, and consistency between interfaces.", "topics": ["grounding", "planning", "reinforcement-learning", "evaluation"], "updated_at": "2026-07-21", "url": "https://arxiv.org/abs/2506.17913"}
{"aliases": [], "authors": ["Jihao Gu", "Qihang Ai", "Yingyao Wang", "Pi Bu", "Jingxuan Xing", "Zekun Zhu", "Wei Jiang", "Ziming Wang", "Yingxiu Zhao", "Ming-Liang Zhang", "Jun Song", "Yuning Jiang", "Bo Zheng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9620eb7d8fa5f5c2a318ff516d45b8a4cfd35cf0c235b96bab3bba4e5a6aab9b", "submitted_at": "2025-06-25T11:34:43Z", "updated_at": "2026-04-25T13:54:32Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2506.20332v4"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2506.20332", "identifiers": {"arxiv": "2506.20332"}, "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/2506.20332"}], "name": "Mobile-R1: Towards Interactive Capability for VLM-Based Mobile Agent via Systematic Training", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses a staged curriculum from response-format alignment through action feedback to interactive task-level training for mobile agents.", "topics": ["reinforcement-learning", "mobile-navigation"], "updated_at": "2026-04-25", "url": "https://arxiv.org/abs/2506.20332"}
{"aliases": [], "authors": ["Dheeraj Vattikonda", "Santhoshi Ravichandran", "Emiliano Penaloza", "Hadi Nekoei", "Megh Thakkar", "Thibault Le Sellier de Chezelles", "Nicolas Gontier", "Miguel Muñoz-Mármol", "Sahar Omidi Shayegan", "Stefania Raimondo", "Xue Liu", "Alexandre Drouin", "Laurent Charlin", "Alexandre Piché", "Alexandre Lacoste", "Massimo Caccia"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "15fa527a869ab66f4f39900ab07ea8b8db930703a8e950ed1e195e1e66164099", "submitted_at": "2025-07-05T17:12:33Z", "updated_at": "2026-02-13T17:24:17Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2507.04103v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.04103", "identifiers": {"arxiv": "2507.04103"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.04103"}], "name": "How to Train Your LLM Web Agent: A Statistical Diagnosis", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies compute allocation for web-agent post-training with teacher-generated supervised fine-tuning followed by on-policy reinforcement learning. It samples 1,370 training configurations and uses bootstrap analysis to examine hyperparameter sensitivity on WorkArena and MiniWoB++.", "topics": ["imitation-learning", "reinforcement-learning", "efficiency", "evaluation"], "updated_at": "2026-02-13", "url": "https://arxiv.org/abs/2507.04103"}
{"aliases": [], "authors": ["Yucheng Shi", "Wenhao Yu", "Zaitang Li", "Yonglin Wang", "Hongming Zhang", "Ninghao Liu", "Haitao Mi", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "45d36dd32b30bf4faee6d9a72e3cdfe91edff465bbd886bdfc3c27c3da475b82", "submitted_at": "2025-07-08T07:07:53Z", "updated_at": "2025-07-08T07:07:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.05720v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.05720", "identifiers": {"arxiv": "2507.05720"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.05720"}], "name": "MobileGUI-RL: Advancing Mobile GUI Agent through Reinforcement Learning in Online Environment", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileGUI-RL constructs an online training curriculum through agent exploration and task filtering. It adapts GRPO with trajectory-aware advantages and rewards for task completion and execution efficiency, evaluated on three online mobile-agent benchmarks.", "topics": ["mobile-navigation", "reinforcement-learning", "data-generation", "efficiency"], "updated_at": "2025-07-08", "url": "https://arxiv.org/abs/2507.05720"}
{"aliases": [], "authors": ["Yan Yang", "Dongxu Li", "Yutong Dai", "Yuhao Yang", "Ziyang Luo", "Zirui Zhao", "Zhiyuan Hu", "Junzhe Huang", "Amrita Saha", "Zeyuan Chen", "Ran Xu", "Liyuan Pan", "Silvio Savarese", "Caiming Xiong", "Junnan Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3c1f659a30e1a0bd04219b9383927ce72e5f5b0d20783a89b850036d087155a4", "submitted_at": "2025-07-08T08:52:18Z", "updated_at": "2025-10-03T23:50:19Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2507.05791v5"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2507.05791", "identifiers": {"arxiv": "2507.05791"}, "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/2507.05791"}], "name": "GTA1: GUI Test-time Scaling Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "GTA1 samples multiple action proposals at each GUI step and uses a judge to select among them. A separate grounding model is trained with reinforcement learning to map selected proposals to visual targets, linking planning-time search with click localization.", "topics": ["planning", "grounding", "reinforcement-learning"], "updated_at": "2025-10-03", "url": "https://arxiv.org/abs/2507.05791"}
{"aliases": [], "authors": ["Fei Tang", "Zhangxuan Gu", "Zhengxi Lu", "Xuyang Liu", "Shuheng Shen", "Changhua Meng", "Wen Wang", "Wenqi Zhang", "Yongliang Shen", "Weiming Lu", "Jun Xiao", "Yueting Zhuang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d2a2a49373259cb5903fad8c2fb48a7380732642bedd81a644e735af4f028c63", "submitted_at": "2025-07-21T17:53:42Z", "updated_at": "2025-07-28T16:54:13Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2507.15846v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.15846", "identifiers": {"arxiv": "2507.15846"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.15846"}], "name": "GUI-G$^2$: Gaussian Reward Modeling for GUI Grounding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-21", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-G2 replaces binary grounding rewards with Gaussian point rewards and spatial-coverage rewards. An adaptive variance tied to element dimensions supplies continuous training feedback for targets of different sizes, evaluated on the ScreenSpot benchmark family.", "topics": ["grounding", "reinforcement-learning", "reward-models"], "updated_at": "2025-07-28", "url": "https://arxiv.org/abs/2507.15846"}
{"aliases": [], "authors": ["Shuquan Lian", "Yuhang Wu", "Jia Ma", "Yifan Ding", "Zihan Song", "Bingqi Chen", "Xiawu Zheng", "Hui Li", "Rongrong Ji"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "741cce21b25e9b29b99e9804a399695bf8a0ae89df6307d9f9bff68e3e856e85", "submitted_at": "2025-07-29T17:22:07Z", "updated_at": "2026-04-08T16:03:48Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2507.22025v4"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2507.22025", "identifiers": {"arxiv": "2507.22025"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2507.22025"}], "name": "UI-AGILE: Advancing GUI Agents with Effective Reinforcement Learning and Precise Inference-Time Grounding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines grounding-oriented reward and sampling improvements with inference-time image decomposition and target selection.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2026-04-08", "url": "https://arxiv.org/abs/2507.22025"}
{"aliases": [], "authors": ["Zhihao Luo", "Wentao Yan", "Jingyu Gong", "Min Wang", "Zhizhong Zhang", "Xuhong Wang", "Yuan Xie", "Xin Tan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e892ab93511f4164f37ef79ac0473e3c1eed7d6ec11fb6c8ec82092f26c5ca68", "submitted_at": "2025-08-04T04:28:18Z", "updated_at": "2026-05-02T12:32:44Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2508.02046v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2508.02046", "identifiers": {"arxiv": "2508.02046"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.02046"}], "name": "NaviMaster: Learning a Unified Policy for GUI and Embodied Navigation Tasks", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-08-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "NaviMaster trains a shared policy for GUI and embodied navigation using a common visual-trajectory formulation. It combines trajectory collection, reinforcement learning and distance-aware rewards, with evaluation across both task families.", "topics": ["reinforcement-learning", "data-generation", "robustness"], "updated_at": "2026-05-02", "url": "https://arxiv.org/abs/2508.02046"}
{"aliases": [], "authors": ["Liujian Tang", "Shaokang Dong", "Yijia Huang", "Minqi Xiang", "Hongtao Ruan", "Bin Wang", "Shuo Li", "Zhiheng Xi", "Zhihui Cao", "Hailiang Pang", "Heng Kong", "He Yang", "Mingxu Chai", "Zhilin Gao", "Xingyu Liu", "Yingnan Fu", "Jiaming Liu", "Xuanjing Huang", "Yu-Gang Jiang", "Tao Gui", "Qi Zhang", "Kang Wang", "Yunke Zhang", "Yuran Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d5175e5ed0061660fa950512300daa878eb90e5c98bdac8f80d478414b0a2931", "submitted_at": "2025-07-19T12:33:43Z", "updated_at": "2025-09-11T14:28:11Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2508.03700v5"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.03700", "identifiers": {"arxiv": "2508.03700"}, "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/2508.03700"}], "name": "MagicGUI: A Foundational Mobile GUI Agent with Scalable Data Pipeline and Reinforcement Fine-tuning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-07-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines a mobile GUI data pipeline, perception and planning capabilities, and reinforcement fine-tuning for interactive task execution.", "topics": ["mobile-navigation", "data-generation", "reinforcement-learning"], "updated_at": "2025-09-11", "url": "https://arxiv.org/abs/2508.03700"}
{"aliases": [], "authors": ["Yuhao Cheng", "Liang Tang", "Shuxian Li", "Yukang Huo", "Tiaonan Duan", "Kaer Huang", "Yanzhe Jing", "Yiqiang Yan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d54ad74695a60cb040d88fcb85549912d7d4ecf6ca2ed78e636d05f8f17d1795", "submitted_at": "2025-08-06T02:57:22Z", "updated_at": "2026-01-22T05:22:03Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.04037v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.04037", "identifiers": {"arxiv": "2508.04037"}, "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/2508.04037"}], "name": "Evolving in Tasks: Empowering the Multi-modality Large Language Model as the Computer Use Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "The Self-Evolution Agent combines automatic generation of verifiable computer-use trajectories with step-wise reinforcement learning for long-horizon training. It also describes a model-enhancement procedure that integrates grounding and planning capabilities without an additional training stage.", "topics": ["data-generation", "reinforcement-learning", "grounding", "planning"], "updated_at": "2026-01-22", "url": "https://arxiv.org/abs/2508.04037"}
{"aliases": [], "authors": ["Weitai Kang", "Bin Lei", "Gaowen Liu", "Caiwen Ding", "Yan Yan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "87b5336c1850c15228880e0625252331b15b79d1cf1e3929abbc01ead70d5f25", "submitted_at": "2025-08-06T12:35:24Z", "updated_at": "2025-08-06T12:35:24Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.04389v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:51Z", "role": "primary", "sha256": "b91987360225499418fcd3645d7d5effac91e8202bf7c7e17c9f9e31d25b5be8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.01423%2C2506.03231%2C2506.04301%2C2506.05336%2C2506.07385%2C2506.07398%2C2506.08136%2C2506.08837%2C2506.10953%2C2506.11425%2C2506.11763%2C2506.13538%2C2506.14683%2C2506.15461%2C2506.18959%2C2506.19502%2C2506.20668%2C2506.21252%2C2506.21495%2C2506.21614%2C2507.01599%2C2507.02279%2C2507.02825%2C2507.04036%2C2507.07400%2C2507.09481%2C2507.09884%2C2507.13348%2C2507.14447%2C2507.16704%2C2507.19902%2C2507.21678%2C2507.22925%2C2508.01548%2C2508.02866%2C2508.04002%2C2508.04389%2C2508.05193%2C2508.05294%2C2508.05748%2C2508.07871%2C2508.07999%2C2508.08228%2C2508.09124%2C2508.09981%2C2508.15804%2C2508.16153%2C2508.17398%2C2508.18489%2C2508.19611&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:37Z", "role": "primary", "sha256": "afdfecfc117506609a5bff392254299d88cc84c971469efe0601cbcb02660ecf", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2508.04389v1"}], "id": "arxiv:2508.04389", "identifiers": {"arxiv": "2508.04389"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:37Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.04389"}], "name": "GuirlVG: Incentivize GUI Visual Grounding via Empirical Exploration on Reinforcement Learning", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GuirlVG, an empirical study of reinforcement fine-tuning for GUI visual grounding. It analyzes training components and adds an adaptive KL-based mechanism to limit reward over-optimization.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2025-08-06", "url": "https://arxiv.org/abs/2508.04389"}
{"aliases": [], "authors": ["Zeyi Sun", "Ziyu Liu", "Yuhang Zang", "Yuhang Cao", "Xiaoyi Dong", "Tong Wu", "Dahua Lin", "Jiaqi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "268c38e6f11f11771a1251fc522112ff052557b5753dff6cadf2ba49c0ec04b2", "submitted_at": "2025-08-06T17:58:46Z", "updated_at": "2025-08-12T15:11:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.04700v2"}}, "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:2508.04700", "identifiers": {"arxiv": "2508.04700"}, "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/2508.04700"}], "name": "SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "SEAgent explores unfamiliar software using a world-state assessor and an automatically generated task curriculum. It learns from failed and successful interactions through adversarial imitation and GRPO, then combines specialist experience into a generalist agent evaluated on five OSWorld software environments.", "topics": ["reinforcement-learning", "data-generation", "world-models", "imitation-learning"], "updated_at": "2025-08-12", "url": "https://arxiv.org/abs/2508.04700"}
{"aliases": [], "authors": ["Yong Du", "Yuchen Yan", "Fei Tang", "Zhengxi Lu", "Chang Zong", "Weiming Lu", "Shengpei Jiang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c7deba3ec19a878cfd402c5dac61e4dca284d478429bc9204264448062299e99", "submitted_at": "2025-08-07T17:54:27Z", "updated_at": "2025-11-13T12:54:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.05615v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.05615", "identifiers": {"arxiv": "2508.05615"}, "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/2508.05615"}], "name": "Test-Time Reinforcement Learning for GUI Grounding via Region Consistency", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-RC aggregates multiple grounding predictions into spatial voting grids to choose a consensus region. GUI-RCPO then turns agreement with that region into a reward for test-time policy updates on unlabeled GUI screenshots.", "topics": ["grounding", "reinforcement-learning", "training-free"], "updated_at": "2025-11-13", "url": "https://arxiv.org/abs/2508.05615"}
{"aliases": [], "authors": ["Yuhang Liu", "Zeyu Liu", "Shuanghe Zhu", "Pengxiang Li", "Congkai Xie", "Jiasheng Wang", "Xavier Hu", "Xiaotian Han", "Jianbo Yuan", "Xinyao Wang", "Shengyu Zhang", "Hongxia Yang", "Fei Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a69d2bdb4d01e6e5c9f08da33281c8247f59af52f3bc08dfed6b7e24da88e194", "submitted_at": "2025-08-07T17:49:56Z", "updated_at": "2025-12-08T10:49:41Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.05731v2"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.05731", "identifiers": {"arxiv": "2508.05731"}, "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/2508.05731"}], "name": "InfiGUI-G1: Advancing GUI Grounding with Adaptive Exploration Policy Optimization", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses multiple candidate answers and adaptive exploration rewards to improve spatial and semantic alignment in GUI grounding.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2025-12-08", "url": "https://arxiv.org/abs/2508.05731"}
{"aliases": [], "authors": ["Zhangxuan Gu", "Zhengwen Zeng", "Zhenyu Xu", "Xingran Zhou", "Shuheng Shen", "Yunfei Liu", "Beitong Zhou", "Changhua Meng", "Tianyu Xia", "Weizhi Chen", "Yue Wen", "Jingya Dou", "Fei Tang", "Jinzhen Lin", "Yulin Liu", "Zhenlin Guo", "Yichen Gong", "Heng Jia", "Changlong Gao", "Yuan Guo", "Yong Deng", "Zhenyu Guo", "Liang Chen", "Weiqiang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db88c4ba15c48210f0b618a8b5d2b9c02184f2e4375d45adf22b9a0a395abc45", "submitted_at": "2025-08-14T16:58:07Z", "updated_at": "2025-08-15T14:49:07Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.10833v2"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2508.10833", "identifiers": {"arxiv": "2508.10833"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.10833"}], "name": "UI-Venus Technical Report: Building High-performance UI Agents with RFT", "notes": "Summary and topic tags reviewed against the primary abstract. Reported experiments, task counts, and artifact availability are not independently reproduced.", "published_at": "2025-08-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies screenshot-based GUI grounding and navigation using curated interaction data, reward design, and reinforcement fine-tuning.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2025-08-15", "url": "https://arxiv.org/abs/2508.10833"}
{"aliases": [], "authors": ["Songqin Nong", "Xiaoxuan Tang", "Jingxuan Xu", "Sheng Zhou", "Jianfeng Chen", "Tao Jiang", "Wenhao Xu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ea849d8e7640d7a40d582431233418cf1cbd29585ca6c78638919b8832265445", "submitted_at": "2025-08-15T09:55:02Z", "updated_at": "2026-03-13T11:10:50Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.11360v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2508.11360", "identifiers": {"arxiv": "2508.11360"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.11360"}], "name": "CRAFT-GUI: Curriculum-Reinforced Agent For GUI Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "CRAFT-GUI organizes GUI trajectories by difficulty for curriculum-based GRPO training. Its reward combines rule-based signals with model judgments, with evaluation on Android Control and an internal online benchmark.", "topics": ["reinforcement-learning", "reward-models", "mobile-navigation"], "updated_at": "2026-03-13", "url": "https://arxiv.org/abs/2508.11360"}
{"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": ["Qihang Ai", "Pi Bu", "Yue Cao", "Yingyao Wang", "Jihao Gu", "Jingxuan Xing", "Zekun Zhu", "Wei Jiang", "Zhicheng Zheng", "Jun Song", "Yuning Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "385ed1b6e60c27660f0cbf281b0f2bbde052691801f7faae99cc053dfbc0d127", "submitted_at": "2025-08-27T08:40:05Z", "updated_at": "2026-04-25T13:29:08Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.19679v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2508.19679", "identifiers": {"arxiv": "2508.19679"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.19679"}], "name": "InquireMobile: Teaching VLM-based Mobile Agent to Request Human Assistance via Reinforcement Fine-Tuning", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "InquireMobile trains mobile agents to seek human confirmation at critical decisions using staged reinforcement fine-tuning and pre-action reasoning. InquireBench evaluates safe interaction and proactive inquiry, distinguishing asking for help from completing the underlying task.", "topics": ["human-in-the-loop", "reinforcement-learning", "evaluation", "mobile-navigation"], "updated_at": "2026-04-25", "url": "https://arxiv.org/abs/2508.19679"}
{"aliases": [], "authors": ["Zeyi Sun", "Yuhang Cao", "Jianze Liang", "Qiushi Sun", "Ziyu Liu", "Zhixiong Zhang", "Yuhang Zang", "Xiaoyi Dong", "Kai Chen", "Dahua Lin", "Jiaqi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ac401f8d5c56050e72839f3bffba4f2c78bbdf34a00cab05fe504f03efd96a19", "submitted_at": "2025-08-27T17:59:50Z", "updated_at": "2025-08-27T17:59:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.20096v1"}}, "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:2508.20096", "identifiers": {"arxiv": "2508.20096"}, "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/2508.20096"}], "name": "CODA: Coordinating the Cerebrum and Cerebellum for a Dual-Brain Computer Use Agent with Decoupled 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": "2025-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "CODA combines a generalist planner with a specialist GUI executor and trains application-specific planners through decoupled GRPO. Successful specialist trajectories then support supervised training of a consolidated planner, evaluated across four ScienceBoard applications.", "topics": ["reinforcement-learning", "planning", "science", "imitation-learning"], "updated_at": "2025-08-27", "url": "https://arxiv.org/abs/2508.20096"}
{"aliases": [], "authors": ["Zhixiong Zeng", "Jing Huang", "Liming Zheng", "Wenkang Han", "Yufeng Zhong", "Lei Chen", "Longrong Yang", "Yingjie Chu", "Yuzhi He", "Lin Ma"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b2a14819c7d7e8258947c9db8cc89012f4d133eaced4e4b904c3bfd1761181e5", "submitted_at": "2025-08-29T16:40:57Z", "updated_at": "2025-08-29T16:40:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.21767v1"}}, "environments": ["desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.21767", "identifiers": {"arxiv": "2508.21767"}, "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/2508.21767"}], "name": "UItron: Foundational GUI Agent with Advanced Perception and 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": "2025-08-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "UItron combines GUI perception and planning data engineering with supervised training and curriculum reinforcement learning in connected mobile and PC environments. It also collects more than one million operation steps across 100 Chinese apps and builds offline and online evaluation environments.", "topics": ["data-generation", "reinforcement-learning", "planning", "grounding"], "updated_at": "2025-08-29", "url": "https://arxiv.org/abs/2508.21767"}
{"aliases": [], "authors": ["Georgios Papoudakis", "Thomas Coste", "Jianye Hao", "Jun Wang", "Kun Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "44c49412776ec9837a08d3d8347214495ffbda14a391c7fa67ad780a0eac9577", "submitted_at": "2025-09-01T18:55:27Z", "updated_at": "2025-11-13T01:26:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.01720v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2509.01720", "identifiers": {"arxiv": "2509.01720"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.01720"}], "name": "Succeed or Learn Slowly: Sample Efficient Off-Policy Reinforcement Learning for Mobile App Control", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Succeed or Learn Slowly applies direct off-policy updates to successful interactions and more conservative regularized updates to negative examples. Successful Transition Replay prioritizes successful experience during policy learning, with evaluation on AndroidWorld.", "topics": ["reinforcement-learning", "mobile-navigation", "efficiency"], "updated_at": "2025-11-13", "url": "https://arxiv.org/abs/2509.01720"}
{"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": ["Zheng Wu", "Heyuan Huang", "Xingyu Lou", "Xiangmou Qu", "Pengzhou Cheng", "Zongru Wu", "Weiwen Liu", "Weinan Zhang", "Jun Wang", "Zhaoxiang Wang", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3bd6b323a376d7606b8c5ec0615b992f47a388cbd3604191241f6a0a9b2cce6e", "submitted_at": "2025-09-09T09:46:01Z", "updated_at": "2026-04-03T02:36:28Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2509.07553v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:56Z", "role": "primary", "sha256": "3429cc2fc7737050cac964c976293fc4793e52c5913636c1f67d04243fbefb6a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.20410%2C2508.20890%2C2508.21456%2C2509.00088%2C2509.00531%2C2509.00969%2C2509.01684%2C2509.03536%2C2509.04018%2C2509.04908%2C2509.05368%2C2509.05831%2C2509.06917%2C2509.07098%2C2509.07553%2C2509.08820%2C2509.10345%2C2509.10540%2C2509.11253%2C2509.12543%2C2509.12987%2C2509.13615%2C2509.14547%2C2509.14744%2C2509.15061%2C2509.15237%2C2509.17336%2C2509.19185%2C2509.19199%2C2509.21072%2C2509.21291%2C2509.21766%2C2509.21799%2C2509.22715%2C2509.22830%2C2509.23738%2C2509.24704%2C2509.25455%2C2509.25693%2C2510.00172%2C2510.00515%2C2510.02190%2C2510.04491%2C2510.05096%2C2510.05156%2C2510.05788%2C2510.07284%2C2510.07915%2C2510.08787%2C2510.09517&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:50Z", "role": "primary", "sha256": "776192081826494b05077d71757f3006575ed50a8cc244004e2de7053104d7eb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2509.07553v3"}], "id": "arxiv:2509.07553", "identifiers": {"arxiv": "2509.07553"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.07553"}], "name": "VeriOS: Query-Driven Proactive Human-Agent-GUI Interaction for Trustworthy OS Agents", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-09-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces VeriOS-Agent, which learns when to execute GUI actions autonomously and when to query a user under untrustworthy conditions. Its staged training combines supervised learning and group-relative policy optimization.", "topics": ["human-in-the-loop", "safety-security", "reinforcement-learning"], "updated_at": "2026-04-03", "url": "https://arxiv.org/abs/2509.07553"}
{"aliases": [], "authors": ["Zhengxi Lu", "Jiabo Ye", "Fei Tang", "Yongliang Shen", "Haiyang Xu", "Ziwei Zheng", "Weiming Lu", "Ming Yan", "Fei Huang", "Jun Xiao", "Yueting Zhuang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1e31638e531aff49fe59bef6e80b35069b1f851ea0cbf1901130557148df4783", "submitted_at": "2025-09-15T03:24:08Z", "updated_at": "2025-09-24T15:05:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.11543v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2509.11543", "identifiers": {"arxiv": "2509.11543"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.11543"}], "name": "UI-S1: Advancing GUI Automation via Semi-online Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-S1 simulates multi-turn online learning over offline trajectories, using a patch module to recover divergences while retaining the model’s own outputs in context. It combines discounted future returns with step- and episode-level advantages and introduces a semi-online performance proxy.", "topics": ["reinforcement-learning", "evaluation", "mobile-navigation"], "updated_at": "2025-09-24", "url": "https://arxiv.org/abs/2509.11543"}
{"aliases": [], "authors": ["Ziyuan Chen", "Zhenghui Zhao", "Zhangye Han", "Miancan Liu", "Xianhang Ye", "Yiqing Li", "Hongbo Min", "Jinkui Ren", "Xiantao Zhang", "Guitao Cao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cc1e4f78f1d3ffdb1b67f7c0bb645561a76db19dfa5f442fa9aa0a0d4d1ae0de", "submitted_at": "2025-09-17T16:58:44Z", "updated_at": "2025-09-19T02:13:09Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.14172v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.14172", "identifiers": {"arxiv": "2509.14172"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.14172"}], "name": "TGPO: Tree-Guided Preference Optimization for Robust Web Agent Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "TGPO merges semantically equivalent web states into tree-structured trajectories for offline preference optimization. A process reward model scores subgoal progress, redundant behavior, and action validity, while dynamic weighting emphasizes consequential decisions.", "topics": ["reinforcement-learning", "reward-models", "web-navigation"], "updated_at": "2025-09-19", "url": "https://arxiv.org/abs/2509.14172"}
{"aliases": [], "authors": ["Xianhang Ye", "Yiqing Li", "Wei Dai", "Miancan Liu", "Ziyuan Chen", "Zhangye Han", "Hongbo Min", "Jinkui Ren", "Xiantao Zhang", "Wen Yang", "Zhi Jin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e415074c6a4b4995350d6936d338682b9a4469f1cbed3a83e0ac44eee6f02bef", "submitted_at": "2025-09-19T02:34:50Z", "updated_at": "2025-09-19T02:34:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.15532v1"}}, "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:2509.15532", "identifiers": {"arxiv": "2509.15532"}, "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/2509.15532"}], "name": "GUI-ARP: Enhancing Grounding with Adaptive Region Perception for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-ARP uses visual attention to crop relevant screenshot regions and decides whether a target needs single-stage or multi-stage analysis. Supervised fine-tuning followed by GRPO trains this adaptive procedure for high-resolution GUI grounding.", "topics": ["grounding", "reinforcement-learning", "screen-understanding"], "updated_at": "2025-09-19", "url": "https://arxiv.org/abs/2509.15532"}
{"aliases": [], "authors": ["Shaojie Zhang", "Ruoceng Zhang", "Pei Fu", "Shaokang Wang", "Jiahui Yang", "Xin Du", "Shiqi Cui", "Bin Qin", "Ying Huang", "Zhenbo Luo", "Jian Luan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d190eefae9309d87f4ea4780d3483ae0ebd834e0bf70e7f2af204c3788d87e7e", "submitted_at": "2025-09-19T04:03:44Z", "updated_at": "2025-10-27T11:34:58Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2509.15566v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.15566", "identifiers": {"arxiv": "2509.15566"}, "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/2509.15566"}], "name": "BTL-UI: Blink-Think-Link Reasoning Model for GUI Agent", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "BTL-UI separates GUI interaction into screen-region attention, planning and executable-command generation. It adds an automated annotation pipeline and rule-based process/outcome rewards for training, with evaluation on both static understanding and dynamic interaction.", "topics": ["grounding", "planning", "data-generation", "reinforcement-learning"], "updated_at": "2025-10-27", "url": "https://arxiv.org/abs/2509.15566"}
{"aliases": [], "authors": ["Junyu Lu", "Songxin Zhang", "Zejian Xie", "Zhuoyang Song", "Jiaxing Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7b85c9bae829bad343dd6f57c866ea100859fa0ad67f5c1a3ac9b3e53d617486", "submitted_at": "2025-09-22T15:40:31Z", "updated_at": "2025-09-22T15:40:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.17917v1"}}, "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:2509.17917", "identifiers": {"arxiv": "2509.17917"}, "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/2509.17917"}], "name": "Orcust: Stepwise-Feedback Reinforcement Learning for GUI Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Orcust combines principle-constrained reward modeling with GUI trajectory collection in instrumented virtual machines. Environment-checkable and model-derived principles guide stepwise feedback for grounding, basic operations, and complete task execution.", "topics": ["reinforcement-learning", "reward-models", "data-generation", "sandbox"], "updated_at": "2025-09-22", "url": "https://arxiv.org/abs/2509.17917"}
{"aliases": [], "authors": ["Yifan Xu", "Xiao Liu", "Xinghan Liu", "Jiaqi Fu", "Hanchen Zhang", "Bohao Jing", "Shudan Zhang", "Yuting Wang", "Wenyi Zhao", "Yuxiao Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "86bce7c1a120e0e953a239f2a9b39a04a844e6f9710525245d643932ec3c3233", "submitted_at": "2025-09-10T13:09:27Z", "updated_at": "2025-10-24T17:13:05Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.18119v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "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:2509.18119", "identifiers": {"arxiv": "2509.18119"}, "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/2509.18119"}], "name": "MobileRL: Online Agentic Reinforcement Learning for Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileRL adapts GRPO to mobile GUI training with difficulty-dependent replay of successes and filtering of failure curricula. A shortest-path reward adjustment accounts for task length, with evaluations on AndroidWorld and AndroidLab.", "topics": ["reinforcement-learning", "mobile-navigation", "efficiency"], "updated_at": "2025-10-24", "url": "https://arxiv.org/abs/2509.18119"}
{"aliases": [], "authors": ["Yu Zhao", "Wei-Ning Chen", "Huseyin Atahan Inan", "Samuel Kessler", "Lu Wang", "Lukas Wutschitz", "Fangkai Yang", "Chaoyun Zhang", "Pasquale Minervini", "Saravan Rajmohan", "Robert Sim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "83fa51d3642fa7786f7f6c44fcf5d5ab5876d7ae26e5f6184ddff4d5e2009181", "submitted_at": "2025-09-25T20:38:01Z", "updated_at": "2026-05-25T20:58:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.21552v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.21552", "identifiers": {"arxiv": "2509.21552"}, "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/2509.21552"}], "name": "Learning GUI Grounding with Spatial Reasoning from Visual Feedback", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-Cursor treats GUI grounding as a sequence of cursor movements guided by visible feedback and movement history. Online reinforcement learning with trajectory rewards trains the model to refine the target location across multiple steps.", "topics": ["grounding", "reinforcement-learning", "action-prediction", "robustness"], "updated_at": "2026-05-25", "url": "https://arxiv.org/abs/2509.21552"}
{"aliases": [], "authors": ["Renqi Chen", "Zeyin Tao", "Jianming Guo", "Jingzhe Zhu", "Yiheng Peng", "Qingqing Sun", "Tianyi Zhang", "Shuai Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d23b9f33282f77b9d4965078ca993bcf21163dc1d3210aaa4e554d4fc12b37c5", "submitted_at": "2025-09-26T07:05:01Z", "updated_at": "2026-04-13T03:14:30Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.21982v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.21982", "identifiers": {"arxiv": "2509.21982"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.21982"}], "name": "RISK: A Framework for GUI Agents in E-commerce Risk Management", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "RISK combines browser-interaction training data, a benchmark, and reinforcement fine-tuning for e-commerce risk-management workflows. Its trajectories and rewards distinguish individual actions from multi-step stateful tasks involving deeply embedded web content.", "topics": ["industry", "data-generation", "evaluation", "reinforcement-learning"], "updated_at": "2026-04-13", "url": "https://arxiv.org/abs/2509.21982"}
{"aliases": [], "authors": ["Pengxiang Li", "Zechen Hu", "Zirui Shang", "Jingrong Wu", "Yang Liu", "Hui Liu", "Zhi Gao", "Chenrui Shi", "Bofei Zhang", "Zihao Zhang", "Xiaochuan Shi", "Zedong YU", "Yuwei Wu", "Xinxiao Wu", "Yunde Jia", "Liuyu Xiang", "Zhaofeng He", "Qing Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a96e774920489ef366ad462972a938863a3e714672e9c1f4b784d57eee34c4ca", "submitted_at": "2025-09-28T13:19:20Z", "updated_at": "2025-09-28T13:19:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.23866v1"}}, "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:2509.23866", "identifiers": {"arxiv": "2509.23866"}, "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/2509.23866"}], "name": "Efficient Multi-turn RL for GUI Agents via Decoupled Training and Adaptive Data Curation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "DART separates GUI reinforcement learning into asynchronous environment, rollout, data-management, and training services. It curates experience by task difficulty, successful trajectories, and step uncertainty while correcting rollout-policy mismatch, with evaluation on OSWorld.", "topics": ["reinforcement-learning", "efficiency", "data-generation", "sandbox"], "updated_at": "2025-09-28", "url": "https://arxiv.org/abs/2509.23866"}
{"aliases": [], "authors": ["Ram Ramrakhya", "Andrew Szot", "Omar Attia", "Yuhao Yang", "Anh Nguyen", "Bogdan Mazoure", "Zhe Gan", "Harsh Agrawal", "Alexander Toshev"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d41da9264d664abbeacccae889f9f14f403fc4edbd7453ed4ff276689592fe68", "submitted_at": "2025-09-29T17:00:02Z", "updated_at": "2025-09-29T17:00:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.25047v1"}}, "environments": ["android", "linux", "desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2509.25047", "identifiers": {"arxiv": "2509.25047"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.25047"}], "name": "Scaling Synthetic Task Generation for Agents via Exploration", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "AutoPlay explores application states and functions to generate tasks, then uses a model executor and verifier to construct demonstrations. It produces 20,000 tasks across 20 Android apps and 10,000 tasks across 13 Ubuntu apps for supervised and reinforcement learning of UI agents.", "topics": ["data-generation", "reinforcement-learning", "imitation-learning"], "updated_at": "2025-09-29", "url": "https://arxiv.org/abs/2509.25047"}
{"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": ["Sanjari Srivastava", "Gang Li", "Cheng Chang", "Rishu Garg", "Manpreet Kaur", "Charlene Y. Lee", "Yuezhang Li", "Yining Mao", "Ignacio Cases", "Yanan Xie", "Peng Qi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3af3fb75d012ebcf4ad082ab7b83b0553596bbbb8598751d0a70c86d929a22ee", "submitted_at": "2025-10-10T21:20:51Z", "updated_at": "2026-05-18T22:06:26Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.09872v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:52Z", "role": "primary", "sha256": "0b730323b2ce85c39aa7d534665901a480ff06540c73f1f80832fcc0735b8789", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2310.08588%2C2402.09205%2C2406.05720%2C2407.21075%2C2410.07166%2C2503.03743%2C2508.04037%2C2508.10833%2C2508.11360%2C2508.13024%2C2508.13186%2C2508.14104%2C2509.01619%2C2509.01720%2C2509.02322%2C2509.04198%2C2509.07764%2C2509.21823%2C2509.22256%2C2509.23263%2C2509.23866%2C2509.25047%2C2509.26539%2C2510.00413%2C2510.00536%2C2510.01524%2C2510.02204%2C2510.02250%2C2510.02418%2C2510.03204%2C2510.03612%2C2510.04039%2C2510.04373%2C2510.04607%2C2510.06223%2C2510.06607%2C2510.09872%2C2510.10666%2C2510.11035%2C2510.11221%2C2510.11892%2C2510.13543%2C2510.15455%2C2510.15863%2C2510.16252%2C2510.18113%2C2510.18314%2C2510.19205%2C2510.22732%2C2511.06101&max_results=50"}], "id": "arxiv:2510.09872", "identifiers": {"arxiv": "2510.09872"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.09872"}], "name": "WARC-Bench: Web Archive Based Benchmark for GUI Subtask Executions", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "WARC-Bench packages dynamic web pages as replayable archives and evaluates short navigation subtasks involving components such as date pickers and scrollable containers. It also studies supervised training followed by reinforcement learning with verifiable rewards on these browser interactions.", "topics": ["evaluation", "web-navigation", "sandbox", "reinforcement-learning"], "updated_at": "2026-05-18", "url": "https://arxiv.org/abs/2510.09872"}
{"aliases": [], "authors": ["Zhe Wu", "Hongjin Lu", "Junliang Xing", "Changhao Zhang", "Yuxuan Li", "Yin Zhu", "Yuhao Yang", "Yuheng Jing", "Kai Li", "Kun Shao", "Jianye Hao", "Jun Wang", "Yuanchun Shi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "95fd3e1286ed49d569b05644d77f822ea9747cfb4769cb7cb93f39c9271cf57e", "submitted_at": "2025-10-16T07:38:21Z", "updated_at": "2026-02-06T02:04:25Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.14388v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:59Z", "role": "primary", "sha256": "240e3d6528900f4a3e87f12f5e55cb7654a4d25d4478e291e9bbf725e7013118", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.10587%2C2510.10909%2C2510.11144%2C2510.12798%2C2510.14388%2C2510.14513%2C2510.14969%2C2510.16156%2C2510.17017%2C2510.17800%2C2510.19207%2C2510.19247%2C2510.19386%2C2510.19864%2C2510.20333%2C2510.21652%2C2510.22443%2C2510.22936%2C2510.23730%2C2510.23883%2C2510.24702%2C2510.25726%2C2510.26328%2C2510.27135%2C2510.27410%2C2511.00153%2C2511.02650%2C2511.02687%2C2511.03690%2C2511.04220%2C2511.05797%2C2511.06251%2C2511.08319%2C2511.08487%2C2511.10110%2C2511.13725%2C2511.14210%2C2511.14945%2C2511.15705%2C2511.16842%2C2511.17649%2C2511.17775%2C2511.19496%2C2511.19516%2C2511.20636%2C2511.21990%2C2511.22074%2C2512.04302%2C2512.04529%2C2512.06726&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:56Z", "role": "primary", "sha256": "fecd42f8ca75572410c354c892734cb8eefbd3fa622064f6bc6db0f535cf5607", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2510.14388v2"}], "id": "arxiv:2510.14388", "identifiers": {"arxiv": "2510.14388"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.14388"}], "name": "Hi-Agent: Hierarchical Vision-Language Agents for Mobile Device Control", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced. This Hi-Agent has a distinct author team and method from the earlier HiAgent working-memory paper; the shared project-name prefix is not treated as an identity match.", "published_at": "2025-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Hi-Agent, jointly training a high-level reasoning model and a low-level mobile action model. Execution feedback supplies a foresight advantage for optimizing subgoal decisions, with evaluations on Android control benchmarks.", "topics": ["planning", "reinforcement-learning", "mobile-navigation"], "updated_at": "2026-02-06", "url": "https://arxiv.org/abs/2510.14388"}
{"aliases": [], "authors": ["Yuxuan Lu", "Ziyi Wang", "Jing Huang", "Hui Liu", "Jiri Gesi", "Yan Han", "Shihan Fu", "Tianqi Zheng", "Xianfeng Tang", "Chen Luo", "Yisi Sang", "Jin Lai", "Dakuo Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a9a1f8ebdf70ff2fa7b0c807160cd583c0f017a536457ff5915e589fca5fa751", "submitted_at": "2025-10-17T22:54:33Z", "updated_at": "2026-05-17T17:26:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.16252v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.16252", "identifiers": {"arxiv": "2510.16252"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.16252"}], "name": "WEBSERV: A Full-Stack and RL-Ready Web Environment for Training Web Agents at Scale", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebServ combines copy-on-write container environments with a compact DOM-derived browser interface for training web agents. Network-aware waiting supports action execution in single-page applications, with evaluations and reinforcement-learning experiments on WebArena-Lite.", "topics": ["sandbox", "reinforcement-learning", "efficiency", "web-navigation"], "updated_at": "2026-05-17", "url": "https://arxiv.org/abs/2510.16252"}
{"aliases": [], "authors": ["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": ["Liangyu Chen", "Hanzhang Zhou", "Chenglin Cai", "Jianan Zhang", "Panrong Tong", "Quyu Kong", "Xu Zhang", "Chen Liu", "Yuqi Liu", "Wenxuan Wang", "Yue Wang", "Qin Jin", "Steven Hoi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3378dd9187f8c8df983d97b10b2ca5b64b2e2332760a734a09db8ce34a398a0b", "submitted_at": "2025-10-23T07:18:32Z", "updated_at": "2025-10-23T07:18:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.20286v1"}}, "environments": ["cross-platform", "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", "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:2510.20286", "identifiers": {"arxiv": "2510.20286"}, "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/2510.20286"}], "name": "UI-Ins: Enhancing GUI Grounding with Multi-Perspective Instruction-as-Reasoning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Treats alternative instruction formulations as reasoning paths and trains GUI-grounding models to select and combine useful perspectives.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2025-10-23", "url": "https://arxiv.org/abs/2510.20286"}
{"aliases": [], "authors": ["Yangqin Jiang", "Chao Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "09bd73dff01641593f31f7d40effd60df42e3d9460ffa010f76c00a32be3593b", "submitted_at": "2025-10-24T20:23:12Z", "updated_at": "2026-10-05T01:17:12Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2510.22009v3"}}, "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"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2510.22009", "identifiers": {"arxiv": "2510.22009"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.22009"}], "name": "OpenPhone: Mobile Agentic Foundation Models", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "OpenPhone trains a compact mobile policy with supervised learning followed by GRPO on synthetic GUI data, incorporating reasoning and memory management. Its device-cloud design handles routine work locally and delegates complex subtasks, with online evaluation on AndroidLab.", "topics": ["reinforcement-learning", "memory", "efficiency", "mobile-navigation"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2510.22009"}
{"aliases": [], "authors": ["Tao Liu", "Chongyu Wang", "Rongjie Li", "Yingchen Yu", "Xuming He", "Bai Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "283b1a1de2a4e1d4ec8175f0c15af66c10097a7f5ad9fd2793e93b888526ff7e", "submitted_at": "2025-10-31T06:10:57Z", "updated_at": "2025-10-31T06:10:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.27210v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:54:12Z", "role": "primary", "sha256": "5f5bae19420f79231516d4bcf604ea18d4f930fc7790cde5a3e7fcb8d41a1126", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2508.19461%2C2508.19493%2C2509.21782%2C2509.25140%2C2510.03485%2C2510.07809%2C2510.14276%2C2510.27140%2C2510.27210%2C2511.13248%2C2512.08417%2C2601.07263%2C2602.13516%2C2603.00476%2C2603.11862%2C2603.18762%2C2603.23791%2C2604.00986%2C2604.16515%2C2604.16548%2C2604.24348%2C2605.05509%2C2605.08876%2C2605.20833%2C2605.24785%2C2607.00333%2C2607.18659%2C2607.25255%2C2608.00202%2C2608.08939%2C2608.24913%2C2608.29596&max_results=32"}], "id": "arxiv:2510.27210", "identifiers": {"arxiv": "2510.27210"}, "kind": "paper", "last_verified_at": "2026-10-18T10:54:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.27210"}], "name": "GUI-Rise: Structured Reasoning and History Summarization for GUI Navigation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-10-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines structured progress-and-decision reasoning with history summarization to improve cross-domain generalization in GUI navigation agents.", "topics": ["planning", "interaction-history", "reinforcement-learning"], "updated_at": "2025-10-31", "url": "https://arxiv.org/abs/2510.27210"}
{"aliases": [], "authors": ["Shaojie Zhang", "Pei Fu", "Ruoceng Zhang", "Jiahui Yang", "Anan Du", "Xiuwen Xi", "Shaokang Wang", "Ying Huang", "Bin Qin", "Zhenbo Luo", "Jian Luan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6869e86d7cd4242ba8dcf942bb1cef3c157dff4f0ef37f7138ff5025d082c7b8", "submitted_at": "2025-10-31T08:07:02Z", "updated_at": "2026-05-27T02:32:58Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.27266v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.27266", "identifiers": {"arxiv": "2510.27266"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.27266"}], "name": "Enhancing Trustworthy GUI Grounding via Self-Critiqued Reinforcement Learning", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-31", "relations": [], "review_status": "curated", "scope": "core", "summary": "HyperClick trains GUI grounders to produce a click and an accompanying confidence estimate using rewards for correctness and confidence alignment. Its evaluation examines localization, calibration, and abstention behavior.", "topics": ["grounding", "reinforcement-learning", "robustness", "evaluation"], "updated_at": "2026-05-27", "url": "https://arxiv.org/abs/2510.27266"}
{"aliases": [], "authors": ["Aarash Feizi", "Shravan Nayak", "Xiangru Jian", "Kevin Qinghong Lin", "Kaixin Li", "Rabiul Awal", "Xing Han Lù", "Johan Obando-Ceron", "Juan A. Rodriguez", "Nicolas Chapados", "David Vazquez", "Adriana Romero-Soriano", "Reihaneh Rabbany", "Perouz Taslakian", "Christopher Pal", "Spandana Gella", "Sai Rajeswar"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0f2b80f294ddf5e3331dcb47a836864032f88f2fdbe35588bfef000cd9ceddc8", "submitted_at": "2025-11-10T17:35:21Z", "updated_at": "2026-06-09T21:30:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2511.07332v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.07332", "identifiers": {"arxiv": "2511.07332"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.07332"}], "name": "Grounding Computer Use Agents on Human Demonstrations", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "GroundCUA turns expert desktop demonstrations into densely annotated screenshots and grounding instructions. The GroundNext model family uses this supervision and subsequent reinforcement learning, with both grounding benchmarks and an OSWorld agent configuration used for evaluation.", "topics": ["grounding", "data-generation", "evaluation", "reinforcement-learning"], "updated_at": "2026-06-09", "url": "https://arxiv.org/abs/2511.07332"}
{"aliases": [], "authors": ["Ziwei Wang", "Leyang Yang", "Xiaoxuan Tang", "Sheng Zhou", "Dajun Chen", "Wei Jiang", "Yong Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d0f61f09ce52b1e66fa81479483ea419aaa16e6b27ce331d3c1039927075fd91", "submitted_at": "2025-11-12T09:06:25Z", "updated_at": "2025-11-12T09:06:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.09127v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.09127", "identifiers": {"arxiv": "2511.09127"}, "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/2511.09127"}], "name": "History-Aware Reasoning for GUI Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "HAR-GUI combines episodic reasoning and short-term memory to use earlier interactions when selecting GUI actions. Its training constructs reflective scenarios and correction guidance and applies a hybrid reinforcement-learning reward to the 3B policy.", "topics": ["interaction-history", "memory", "self-reflection", "reinforcement-learning"], "updated_at": "2025-11-12", "url": "https://arxiv.org/abs/2511.09127"}
{"aliases": [], "authors": ["Yuan Zhao", "Hualei Zhu", "Tingyu Jiang", "Shen Li", "Xiaohang Xu", "Hao Henry Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "101d38ed439c41ebfa826e990697b502b9126728956295d25b0ea9d539ff23e3", "submitted_at": "2025-11-13T03:41:02Z", "updated_at": "2025-11-13T03:41:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2511.10705v1"}}, "environments": ["web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.10705", "identifiers": {"arxiv": "2511.10705"}, "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/2511.10705"}], "name": "Co-EPG: A Framework for Co-Evolution of Planning and Grounding in Autonomous GUI Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Co-EPG alternates reinforcement learning for a planner with updates to a grounding model using generated training examples. The evolving grounder provides feedback to the planner, and evaluation uses Multimodal Mind2Web and AndroidControl.", "topics": ["grounding", "planning", "reinforcement-learning", "data-generation"], "updated_at": "2025-11-13", "url": "https://arxiv.org/abs/2511.10705"}
{"aliases": [], "authors": ["Zengyi Qin", "Jinyuan Chen", "Yunze Man", "Shengcao Cao", "Ziqi Pang", "Zhuoyuan Wang", "Han Fang", "Ling Zhu", "Zixin Xie", "Zibu Wei", "Tianshu Ran", "Haoran Geng", "Ray Pan", "Qizhen Sun", "Zachary Bright", "Yuyang Cai", "Chongye Yang", "Jiace Zhao", "Tianrui Liu", "Han Cao", "Yeyang Zhou", "Rui Wang", "Song Wang", "Xiang Ren", "Bo Zhang", "Yutong Ban", "Pieter Abbeel", "Brian Anthony"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "755d41575ae8ee5d07be0022c8f078c742da4628f12ecd1cad984269441021cf", "submitted_at": "2025-11-11T20:03:38Z", "updated_at": "2026-04-02T02:42:36Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2511.11672v5"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2511.11672", "identifiers": {"arxiv": "2511.11672"}, "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/2511.11672"}], "name": "OSGym: Scalable OS Infra 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": "2025-11-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "OSGym supplies full operating-system sandboxes with GUIs for parallel computer-use data collection and training. It combines per-replica state management, hardware-aware orchestration, copy-on-write virtual-machine disks, and layered failure recovery.", "topics": ["sandbox", "efficiency", "reinforcement-learning", "data-generation"], "updated_at": "2026-04-02", "url": "https://arxiv.org/abs/2511.11672"}
{"aliases": [], "authors": ["Zehao Deng", "Tianjie Ju", "Zheng Wu", "Zhuosheng Zhang", "Gongshen Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9acb395dfaad18b84ea4b52d97be194a42eb42d6ac922c55f2179e38cc258a31", "submitted_at": "2025-11-27T09:01:38Z", "updated_at": "2026-03-04T12:15:06Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2511.22235v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}], "id": "arxiv:2511.22235", "identifiers": {"arxiv": "2511.22235"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:55Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.22235"}], "name": "Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-11-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "This work trains high-level schedulers with staged reinforcement learning driven by execution feedback. Its CES framework separates a Coordinator for planning and decomposition, a StateTracker for compressed context, and a low-level Executor.", "topics": ["planning", "state-tracking", "reinforcement-learning", "long-horizon"], "updated_at": "2026-03-04", "url": "https://arxiv.org/abs/2511.22235"}
{"aliases": [], "authors": ["Haolong Yan", "Yeqing Shen", "Xin Huang", "Jia Wang", "Kaijun Tan", "Zhixuan Liang", "Hongxin Li", "Zheng Ge", "Osamu Yoshie", "Si Li", "Xiangyu Zhang", "Daxin Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fda844ac6767e6d9dfd16919042469e2070ec7f092e0b01fb5a243ed6eb4c282", "submitted_at": "2025-12-02T05:11:23Z", "updated_at": "2025-12-02T05:11:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.02423v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:55Z", "role": "primary", "sha256": "e6061c8dd9c00a70fcf1d4325b671c50a6912063ecbe2b6359175ae4622983a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2511.13091%2C2511.21398%2C2511.22235%2C2512.00756%2C2512.00846%2C2512.01174%2C2512.01763%2C2512.02423%2C2512.04105%2C2512.05941%2C2512.07725%2C2512.09396%2C2512.11611%2C2512.12692%2C2512.17221%2C2512.19396%2C2512.21302%2C2512.22009%2C2512.22047%2C2512.22322%2C2512.23128%2C2601.02371%2C2601.02439%2C2601.03641%2C2601.04035%2C2601.04126%2C2601.04500%2C2601.05787%2C2601.06899%2C2601.08605%2C2601.09636%2C2601.09770%2C2601.11631%2C2601.12349%2C2601.16356%2C2601.17418%2C2601.18197%2C2601.18305%2C2601.19199%2C2601.20732%2C2601.21352%2C2601.21872%2C2601.21961%2C2601.22149%2C2601.22491%2C2601.22528%2C2602.02468%2C2602.05354%2C2602.06090%2C2602.06351&max_results=50"}], "id": "arxiv:2512.02423", "identifiers": {"arxiv": "2512.02423"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:55Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.02423"}], "name": "GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-12-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUI Exploration Lab, a simulation engine for composing screens, interface elements, and navigation graphs. It studies how supervised learning, single-turn RL, and multiturn RL affect GUI navigation and exploration.", "topics": ["reinforcement-learning", "evaluation", "data-generation"], "updated_at": "2025-12-02", "url": "https://arxiv.org/abs/2512.02423"}
{"aliases": [], "authors": ["Zhenyu Wu", "Jingjing Xie", "Zehao Li", "Bowen Yang", "Qiushi Sun", "Zhaoyang Liu", "Zhoumianze Liu", "Yu Qiao", "Xiangyu Yue", "Zun Wang", "Zichen Ding"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "67a1bd40b117b34e5fb994b1c2c93cd285714097d1bc9730baeeaba6c31a7b8a", "submitted_at": "2025-12-18T08:29:50Z", "updated_at": "2025-12-18T08:29:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.16295v1"}}, "environments": ["web", "desktop", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.16295", "identifiers": {"arxiv": "2512.16295"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.16295"}], "name": "OS-Oracle: A Comprehensive Framework for Cross-Platform GUI Critic Models", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-12-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "OS-Oracle synthesizes GUI critic examples and trains a pre-action critic with supervised learning followed by consistency-preserving GRPO. It introduces OS-Critic Bench across mobile, web, and desktop settings and tests critic-assisted agents in OSWorld and AndroidWorld.", "topics": ["reward-models", "reinforcement-learning", "data-generation", "evaluation"], "updated_at": "2025-12-18", "url": "https://arxiv.org/abs/2512.16295"}
{"aliases": [], "authors": ["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": ["Hao Bai", "Alexey Taymanov", "Tong Zhang", "Aviral Kumar", "Spencer Whitehead"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ae8751c4fae3b4f838eba59ffa8be37a1a5f4fa497dfd69fa3075f8581547f51", "submitted_at": "2026-01-05T09:35:11Z", "updated_at": "2026-05-02T18:45:09Z", "version": 6, "versioned_url": "https://arxiv.org/abs/2601.02439v6"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.02439", "identifiers": {"arxiv": "2601.02439"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.02439"}], "name": "WebGym: Scaling Training Environments for Visual Web Agents with Realistic Tasks", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebGym supplies realistic web tasks with rubric-based evaluation and an asynchronous rollout system for reinforcement learning. It studies how training on an agent’s own interaction traces scales with task coverage and transfers to unseen websites.", "topics": ["web-navigation", "reinforcement-learning", "data-generation", "evaluation", "efficiency"], "updated_at": "2026-05-02", "url": "https://arxiv.org/abs/2601.02439"}
{"aliases": [], "authors": ["Ziyun Zhang", "Zezhou Wang", "Xiaoyi Zhang", "Zongyu Guo", "Jiahao Li", "Bin Li", "Yan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d4d0a59f1eb8197b4d389f5f40957fb8aeb1e22d0739f9877b26d57ed70fe59b", "submitted_at": "2026-01-07T17:40:08Z", "updated_at": "2026-06-30T13:54:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.04126v3"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.04126", "identifiers": {"arxiv": "2601.04126"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.04126"}], "name": "InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "InfiniteWeb generates functional, interconnected websites and task evaluators using shared specifications and task-driven development. It evaluates website construction and the transfer of GUI policies trained in these synthetic web environments to OSWorld and Online-Mind2Web.", "topics": ["data-generation", "reinforcement-learning", "evaluation", "sandbox"], "updated_at": "2026-06-30", "url": "https://arxiv.org/abs/2601.04126"}
{"aliases": [], "authors": ["Zezhou Wang", "Ziyun Zhang", "Xiaoyi Zhang", "Zhuzhong Qian", "Yan Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e342723078b50e612df749df80eadfd16e0fe29f5875777c1b3189d8b5981512", "submitted_at": "2026-01-09T13:26:38Z", "updated_at": "2026-02-10T15:12:17Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.05787v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.05787", "identifiers": {"arxiv": "2601.05787"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.05787"}], "name": "From Off-Policy to On-Policy: Enhancing GUI Agents via Bi-level Expert-to-Policy Assimilation", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "BEPA adapts a small set of expert GUI traces into policy-aligned guidance through trajectories reachable by the learner and a dynamically updated task cache. It evaluates this two-level approach to reinforcement learning on desktop and web benchmarks.", "topics": ["reinforcement-learning", "data-generation", "evaluation"], "updated_at": "2026-02-10", "url": "https://arxiv.org/abs/2601.05787"}
{"aliases": [], "authors": ["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": ["Yurun Song", "Jiong Yin", "Rongjunchen Zhang", "Ian G. Harris"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "600f2a44695886fc527b04e1a8b1ad07cec90cf5c58dee834605d5133a760b1e", "submitted_at": "2026-01-14T03:52:57Z", "updated_at": "2026-01-14T03:52:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.11631v1"}}, "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:2601.11631", "identifiers": {"arxiv": "2601.11631"}, "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/2601.11631"}], "name": "Compress to Focus: Efficient Coordinate Compression for Policy Optimization in Multi-Turn 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": "CCPO couples GUI policy optimization with coordinate-guided compression of past visual observations. It aggregates targets across rollouts to focus retained context and uses distance-based advantages to train grounding and compression together.", "topics": ["grounding", "reinforcement-learning", "efficiency", "interaction-history"], "updated_at": "2026-01-14", "url": "https://arxiv.org/abs/2601.11631"}
{"aliases": [], "authors": ["Le Zhang", "Yixiong Xiao", "Xinjiang Lu", "Jingjia Cao", "Yusai Zhao", "Jingbo Zhou", "Lang An", "Zikan Feng", "Wanxiang Sha", "Yu Shi", "Congxi Xiao", "Jian Xiong", "Yankai Zhang", "Hua Wu", "Haifeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ad764cf86a2ecab81f3f730d6d26ea582544a355a90a8be55be220c98aabd7bf", "submitted_at": "2026-01-28T08:45:17Z", "updated_at": "2026-01-28T08:45:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.20380v1"}}, "environments": ["android", "desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.20380", "identifiers": {"arxiv": "2601.20380"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.20380"}], "name": "OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Execution", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "OmegaUse combines exploratory and taxonomy-guided data synthesis with supervised fine-tuning followed by GRPO for GUI control. The report introduces OS-Nav for offline Android and Ubuntu evaluation alongside established grounding and action-prediction benchmarks.", "topics": ["data-generation", "reinforcement-learning", "grounding", "evaluation"], "updated_at": "2026-01-28", "url": "https://arxiv.org/abs/2601.20380"}
{"aliases": [], "authors": ["Ziwei Liu", "Borui Kang", "Hangjie Yuan", "Zixiang Zhao", "Wei Li", "Yifan Zhu", "Tao Feng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1083cf063ccf1781f2cdfe5e266b75b52079fd12ad16d05755577b880c010aec", "submitted_at": "2026-01-28T16:06:31Z", "updated_at": "2026-03-25T10:25:50Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2601.20732v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.20732", "identifiers": {"arxiv": "2601.20732"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.20732"}], "name": "Continual GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Continual GUI Agents defines grounding under changing application domains and screen resolutions. Its GUI-AiF training framework uses separate point- and region-alignment rewards to adapt grounding as these distributions shift.", "topics": ["grounding", "reinforcement-learning", "robustness", "evaluation"], "updated_at": "2026-03-25", "url": "https://arxiv.org/abs/2601.20732"}
{"aliases": [], "authors": ["Yao Zhang", "Shijie Tang", "Zeyu Li", "Zhen Han", "Volker Tresp"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "68ae67856eb871d9a8f259dfb180bde02d29ee97fd7abf0e60719fc2c7abd7ab", "submitted_at": "2026-01-29T15:39:50Z", "updated_at": "2026-04-09T17:48:11Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.21872v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.21872", "identifiers": {"arxiv": "2601.21872"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.21872"}], "name": "WebArbiter: A Principle-Guided Reasoning Process Reward Model for Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebArbiter generates reasoned action-preference verdicts for web navigation, using reasoning distillation followed by reinforcement learning. It introduces WebPRMBench and evaluates both preference judgments and reward-guided trajectory search.", "topics": ["reward-models", "reinforcement-learning", "evaluation", "web-navigation"], "updated_at": "2026-04-09", "url": "https://arxiv.org/abs/2601.21872"}
{"aliases": [], "authors": ["Hang Ding", "Peidong Liu", "Junqiao Wang", "Ziwei Ji", "Meng Cao", "Rongzhao Zhang", "Lynn Ai", "Eric Yang", "Tianyu Shi", "Lei Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aedb5ed90c91d72f62f5ab60be5374b6f431f887425bc760883ce3fb85a46d43", "submitted_at": "2026-01-29T18:59:07Z", "updated_at": "2026-04-17T18:27:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.22149v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.22149", "identifiers": {"arxiv": "2601.22149"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.22149"}], "name": "DynaWeb: Model-Based Reinforcement Learning of Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "DynaWeb learns a world model that predicts web-page representations after actions and uses it to generate policy-training rollouts. It interleaves these rollouts with expert trajectories and evaluates the trained web agents on WebArena and WebVoyager.", "topics": ["world-models", "reinforcement-learning", "data-generation", "web-navigation"], "updated_at": "2026-04-17", "url": "https://arxiv.org/abs/2601.22149"}
{"aliases": [], "authors": ["Han Xiao", "Guozhi Wang", "Hao Wang", "Shilong Liu", "Yuxiang Chai", "Yue Pan", "Yufeng Zhou", "Xiaoxin Chen", "Yafei Wen", "Hongsheng Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f51eb0b7102a37fcc705c53640c5a1d30fcc115cedcdca4a00493ed45b9d5507", "submitted_at": "2026-02-05T16:21:43Z", "updated_at": "2026-02-05T16:21:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.05832v1"}}, "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:2602.05832", "identifiers": {"arxiv": "2602.05832"}, "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/2602.05832"}], "name": "UI-Mem: Self-Evolving Experience Memory for Online Reinforcement Learning in Mobile GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-Mem stores workflows, skills and failure patterns as reusable templates to guide online GUI reinforcement learning. It varies guidance within rollout groups and continually updates the memory so an unguided policy can learn from guided experiences.", "topics": ["memory", "reinforcement-learning", "cross-app"], "updated_at": "2026-02-05", "url": "https://arxiv.org/abs/2602.05832"}
{"aliases": [], "authors": ["Zhongyin Zhao", "Yuan Liu", "Yikun Liu", "Haicheng Wang", "Le Tian", "Xiao Zhou", "Yangxiu You", "Zilin Yu", "Yang Yu", "Jie Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "852813305299178baeda3540cbad539dc30b840681e7bbdaed34dac1341dc0de", "submitted_at": "2026-02-06T05:14:11Z", "updated_at": "2026-02-06T05:14:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.06391v1"}}, "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:2602.06391", "identifiers": {"arxiv": "2602.06391"}, "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/2602.06391"}], "name": "POINTS-GUI-G: GUI-Grounding Journey", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies GUI-grounding development from a weakly grounded base model through data preparation, visual training, and verifiable rewards.", "topics": ["grounding", "reinforcement-learning"], "updated_at": "2026-02-06", "url": "https://arxiv.org/abs/2602.06391"}
{"aliases": [], "authors": ["Venus Team", "Changlong Gao", "Zhangxuan Gu", "Yulin Liu", "Xinyu Qiu", "Shuheng Shen", "Yue Wen", "Tianyu Xia", "Zhenyu Xu", "Zhengwen Zeng", "Beitong Zhou", "Xingran Zhou", "Weizhi Chen", "Sunhao Dai", "Jingya Dou", "Yichen Gong", "Yuan Guo", "Zhenlin Guo", "Feng Li", "Qian Li", "Jinzhen Lin", "Yuqi Zhou", "Linchao Zhu", "Liang Chen", "Zhenyu Guo", "Changhua Meng", "Weiqiang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "efc4ad8f13d55939abcc783bdd652f276c3ac0164ea35232a3fb9366642d4558", "submitted_at": "2026-02-09T18:43:40Z", "updated_at": "2026-02-24T15:17:55Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.09082v2"}}, "environments": ["cross-platform", "web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}], "id": "arxiv:2602.09082", "identifiers": {"arxiv": "2602.09082"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.09082"}], "name": "UI-Venus-1.5 Technical Report", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-02-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a GUI-agent family combining broad intermediate training, full-trajectory online reinforcement learning, and merging of specialized models.", "topics": ["reinforcement-learning", "grounding", "mobile-navigation"], "updated_at": "2026-02-24", "url": "https://arxiv.org/abs/2602.09082"}
{"aliases": [], "authors": ["Yuhao Zheng", "Li'an Zhong", "Yi Wang", "Rui Dai", "Kaikui Liu", "Xiangxiang Chu", "Linyuan Lv", "Philip Torr", "Kevin Qinghong Lin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "55287e2fe8a5e94671e4aab8e9d4a907ba9b6c83306d9b3f61b74adccd5c5059", "submitted_at": "2026-02-10T14:56:19Z", "updated_at": "2026-02-10T14:56:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.09856v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2602.09856", "identifiers": {"arxiv": "2602.09856"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.09856"}], "name": "Code2World: A GUI World Model via Renderable Code Generation", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Code2World predicts future GUI screens through renderable code, trained on AndroidCode trajectories with supervised learning and render-aware reinforcement learning. It evaluates visual prediction and downstream AndroidWorld navigation with the learned world model.", "topics": ["world-models", "data-generation", "reinforcement-learning", "mobile-navigation"], "updated_at": "2026-02-10", "url": "https://arxiv.org/abs/2602.09856"}
{"aliases": [], "authors": ["Tianci Xue", "Zeyi Liao", "Tianneng Shi", "Zilu Wang", "Kai Zhang", "Dawn Song", "Yu Su", "Huan Sun"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d3385966e7b483b33738b3c8e6c9a4d9cf17afa671e1f2e49de4f946cbc5f251", "submitted_at": "2026-02-10T23:06:02Z", "updated_at": "2026-05-11T00:11:46Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.10356v2"}}, "environments": ["desktop", "linux", "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", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:29:29+00:00", "role": "primary", "sha256": "fba520c76ca9268af9239fbeaf5836e21932ab41ce53ccba71f47e3fc767028a", "supports": ["summary", "environments", "scope"], "url": "https://arxiv.org/html/2602.10356v2"}], "id": "arxiv:2602.10356", "identifiers": {"arxiv": "2602.10356"}, "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/2602.10356"}], "name": "Autonomous Continual Learning for Environment Adaptation of Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract and the specified sections of the same recorded paper version. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "ACuRL uses autonomous interface exploration and an adaptive task curriculum to continually train computer-use policies with reinforcement learning. CUAJudge supplies outcome feedback, and experiments study adaptation across desktop applications and environment changes.", "topics": ["reinforcement-learning", "data-generation", "reward-models", "robustness"], "updated_at": "2026-05-11", "url": "https://arxiv.org/abs/2602.10356"}
{"aliases": [], "authors": ["Congmin Zheng", "Xiaoyun Mo", "Xinbei Ma", "Qiqiang Lin", "Yin Zhao", "Jiachen Zhu", "Xingyu Lou", "Jun Wang", "Zhaoxiang Wang", "Weiwen Liu", "Zhuosheng Zhang", "Yong Yu", "Weinan Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ffec5a47e31b541cfdf0fb3c143b55cbb28119463731d9a6f3add0fe116caea9", "submitted_at": "2026-02-12T03:31:40Z", "updated_at": "2026-02-12T03:31:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.11524v1"}}, "environments": ["android", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.11524", "identifiers": {"arxiv": "2602.11524"}, "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/2602.11524"}], "name": "Adaptive Milestone Reward for GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "ADMIRE extracts milestones from successful GUI explorations to provide adaptive intermediate rewards. It assigns credit differently to successful and failed trajectories and evaluates training on AndroidWorld, with additional transfer studies in web and embodied settings.", "topics": ["reinforcement-learning", "reward-models", "mobile-navigation"], "updated_at": "2026-02-12", "url": "https://arxiv.org/abs/2602.11524"}
{"aliases": [], "authors": ["Yuyu Guo", "Wenjie Yang", "Siyuan Yang", "Ziyang Liu", "Cheng Chen", "Yuan Wei", "Yun Hu", "Yang Huang", "Guoliang Hao", "Dongsheng Yuan", "Jianming Wang", "Xin Chen", "Hang Yu", "Lei Lei", "Peng Di"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "04298ed8a4bc0f2811f52eeee6da0a13faa265bb3c5cbc5ea6be573a77c21f27", "submitted_at": "2026-02-14T02:33:55Z", "updated_at": "2026-04-30T02:05:10Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.13559v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2602.13559", "identifiers": {"arxiv": "2602.13559"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.13559"}], "name": "OpAgent: Operator Agent for Web Navigation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-02-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "OpAgent combines multi-task GUI fine-tuning with online web reinforcement learning using outcome judgments and rule-based progress rewards. A modular planner, grounder, reflector and summarizer support navigation and recovery, evaluated separately from the underlying trained policy.", "topics": ["web-navigation", "reinforcement-learning", "planning", "self-reflection"], "updated_at": "2026-04-30", "url": "https://arxiv.org/abs/2602.13559"}
{"aliases": [], "authors": ["Yibo Wang", "Guangda Huzhang", "Yuwei Hu", "Yu Xia", "Shiyin Lu", "Qing-Guo Chen", "Zhao Xu", "Weihua Luo", "Kaifu Zhang", "Lijun Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "573d859133ec9986af92e1943d0ec913d914acf4e54a030a4b103181ba505ae0", "submitted_at": "2026-02-14T07:44:47Z", "updated_at": "2026-02-14T07:44:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.13653v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2602.13653", "identifiers": {"arxiv": "2602.13653"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.13653"}], "name": "Building Autonomous GUI Navigation via Agentic-Q Estimation and Step-Wise Policy Optimization", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "This framework learns an agentic-Q model to estimate individual actions’ contributions to GUI task completion. It then optimizes a policy with step-level samples from its own trajectories, decoupling policy updates from environment interaction.", "topics": ["reinforcement-learning", "reward-models", "grounding"], "updated_at": "2026-02-14", "url": "https://arxiv.org/abs/2602.13653"}
{"aliases": [], "authors": ["Yuan Cao", "Dezhi Ran", "Mengzhou Wu", "Yuzhe Guo", "Xin Chen", "Ang Li", "Gang Cao", "Gong Zhi", "Hao Yu", "Linyi Li", "Wei Yang", "Tao Xie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "29b4c8a5bda1df237fad33ce483018b3476684f904841ed0598ec4286234b858", "submitted_at": "2026-02-15T10:58:01Z", "updated_at": "2026-02-15T10:58:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.14093v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.14093", "identifiers": {"arxiv": "2602.14093"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.14093"}], "name": "GUI-GENESIS: Automated Synthesis of Efficient Environments with Verifiable Rewards for GUI Agent Post-Training", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-GENESIS reconstructs applications as lightweight web environments with executable assertions for reward computation. It evaluates the cost of these training environments and transfer from synthetic interaction training to held-out real tasks.", "topics": ["data-generation", "sandbox", "reinforcement-learning", "evaluation"], "updated_at": "2026-02-15", "url": "https://arxiv.org/abs/2602.14093"}
{"aliases": [], "authors": ["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": ["Zhenquan Yao", "Zitong Huang", "Yihan Zeng", "Jianhua Han", "Hang Xu", "Chun-Mei Feng", "Jianwei Ma", "Wangmeng Zuo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "75f8eb625e969c2b180eeb3e90aa4e513ab5a1dc15657d3849c68b69bc5edf7c", "submitted_at": "2026-03-03T13:02:20Z", "updated_at": "2026-03-07T10:11:08Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.02951v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2603.02951", "identifiers": {"arxiv": "2603.02951"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.02951"}], "name": "CGL: Advancing Continual GUI Learning via Reinforcement Fine-Tuning", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The abstract describes artifact release as future work; this annotation does not assert current availability.", "published_at": "2026-03-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines supervised and reinforcement fine-tuning for continual GUI learning, using policy entropy to adjust their proportions and gradient projection to limit conflicting updates. AndroidControl-CL groups applications into sequential tasks to evaluate adaptation and retention.", "topics": ["reinforcement-learning", "robustness", "evaluation"], "updated_at": "2026-03-07", "url": "https://arxiv.org/abs/2603.02951"}
{"aliases": [], "authors": ["Haoyu Liu", "Dingcheng Li", "Lukas Rutishauser", "Zeyu Zheng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "971770e61bf0af91ad1f17b197f09fc166b26bfb4953d3f2c74d457a87e25f3a", "submitted_at": "2026-03-04T18:29:54Z", "updated_at": "2026-10-05T23:01:49Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.04364v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.04364", "identifiers": {"arxiv": "2603.04364"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.04364"}], "name": "Dual-Modality Multi-Stage Adversarial Safety Training: Robustifying Multimodal Web Agents Against Cross-Modal Attacks", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies webpage injections that affect both screenshots and accessibility trees, then trains web agents through teacher imitation, supervised adversarial examples, and adversarial reinforcement learning. The evaluation tests defenses on MiniWoB++ and held-out tasks.", "topics": ["safety-security", "reinforcement-learning", "evaluation"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2603.04364"}
{"aliases": [], "authors": ["Sicheng Fan", "Qingyun Shi", "Shengze Xu", "Shengbo Cai", "Tieyong Zeng", "Li Ling", "Yanyi Shang", "Dehan Kong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fcca36d9a28ace835ede3fa5a93cb9c54608f164240c18bfe007bc945273b460", "submitted_at": "2026-03-05T10:51:34Z", "updated_at": "2026-04-14T08:14:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.05044v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.05044", "identifiers": {"arxiv": "2603.05044"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.05044"}], "name": "WebFactory: Automated Compression of Foundational Language Intelligence into Grounded Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebFactory, a GUI-agent training pipeline that synthesizes websites, generates tasks and trajectories, and trains with decomposed rewards. Offline and online transfer evaluations examine agents trained on the generated environments.", "topics": ["data-generation", "reinforcement-learning", "web-navigation"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2603.05044"}
{"aliases": [], "authors": ["Li Gu", "Zihuan Jiang", "Zhixiang Chi", "Huan Liu", "Ziqiang Wang", "Yuanhao Yu", "Glen Berseth", "Yang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "68350312fecc36630ecaeec18a203e692e072d31ff80f442662d202b43ead788", "submitted_at": "2026-03-08T03:08:11Z", "updated_at": "2026-03-08T03:08:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.07432v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2603.07432", "identifiers": {"arxiv": "2603.07432"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.07432"}], "name": "Generalization in Online Reinforcement Learning for Mobile Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AndroidWorld-Generalization with separate tests for unseen task instances, templates, and applications. A GRPO training system combines containerized environments with asynchronous trajectory collection to study these generalization regimes.", "topics": ["reinforcement-learning", "evaluation", "robustness"], "updated_at": "2026-03-08", "url": "https://arxiv.org/abs/2603.07432"}
{"aliases": [], "authors": ["Yiping Xie", "Song Chen", "Jingxuan Xing", "Wei Jiang", "Zekun Zhu", "Yingyao Wang", "Pi Bu", "Jun Song", "Yuning Jiang", "Bo Zheng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db31c6ca303f0c7dc43f24dd1eea69621b1e557af33122386945154e8236ab78", "submitted_at": "2026-03-09T16:04:08Z", "updated_at": "2026-03-31T03:08:51Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2603.08533v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.08533", "identifiers": {"arxiv": "2603.08533"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.08533"}], "name": "SecAgent: Efficient Mobile GUI Agent with Semantic Context", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. Mobile use is explicit, but the abstract does not identify a mobile operating system.", "published_at": "2026-03-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents a Chinese mobile GUI dataset and navigation benchmark together with SecAgent, which compresses previous screenshots and actions into semantic summaries. The agent is trained with supervised and reinforcement fine-tuning to use this compact interaction context.", "topics": ["mobile-navigation", "memory", "reinforcement-learning"], "updated_at": "2026-03-31", "url": "https://arxiv.org/abs/2603.08533"}
{"aliases": [], "authors": ["Hyungjoo Chae", "Jungsoo Park", "Alan Ritter"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "178b2140d0c70ded39fede20bc91d1459d0a4bf65c5cb6a63c09c9db73828513", "submitted_at": "2026-03-11T07:58:53Z", "updated_at": "2026-03-11T07:58:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.10505v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.10505", "identifiers": {"arxiv": "2603.10505"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.10505"}], "name": "Safe and Scalable Web Agent Learning via Recreated Websites", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The abstract promises a release upon acceptance; this annotation does not verify artifact availability.", "published_at": "2026-03-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces VeriEnv, which recreates websites as executable training environments and exposes internal state through a Python SDK for task generation and deterministic reward checks. Experiments examine web-agent learning across new and expanded site environments.", "topics": ["data-generation", "sandbox", "reinforcement-learning"], "updated_at": "2026-03-11", "url": "https://arxiv.org/abs/2603.10505"}
{"aliases": [], "authors": ["Siqi Pei", "Liang Tang", "Tiaonan Duan", "Long Chen", "Shuxian Li", "Kaer Huang", "Yanzhe Jing", "Yiqiang Yan", "Bo Zhang", "Chenghao Jiang", "Borui Zhang", "Jiwen Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2101a32746d6c668bff79461983b2e12138959f49ab3cfcb542ecbd629317f1d", "submitted_at": "2026-03-18T07:26:18Z", "updated_at": "2026-03-18T07:26:18Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.17441v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.17441", "identifiers": {"arxiv": "2603.17441"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.17441"}], "name": "AdaZoom-GUI: Adaptive Zoom-based GUI Grounding with Instruction Refinement", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines instruction rewriting with conditional second-stage zooming for small GUI targets. A grounding dataset and GRPO training support joint prediction of click coordinates and element bounding boxes.", "topics": ["grounding", "reinforcement-learning", "screen-understanding"], "updated_at": "2026-03-18", "url": "https://arxiv.org/abs/2603.17441"}
{"aliases": [], "authors": ["Zehao Li", "Zhenyu Wu", "Yibo Zhao", "Bowen Yang", "Jingjing Xie", "Zhaoyang Liu", "Zhoumianze Liu", "Kaiming Jin", "Jianze Liang", "Zonglin Li", "Feng Wu", "Bowen Zhou", "Zun Wang", "Zichen Ding"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ff31371580af6a0c43a6c26f076d5060ee06f974d3a6c1ade67348dc0af4859d", "submitted_at": "2026-03-19T17:47:47Z", "updated_at": "2026-03-19T17:47:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.19191v1"}}, "environments": ["android", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2603.19191", "identifiers": {"arxiv": "2603.19191"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.19191"}], "name": "OS-Themis: A Scalable Critic Framework for Generalist GUI Rewards", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents OS-Themis, a multi-agent critic that divides GUI trajectories into verifiable milestones and reviews their evidence before assigning an outcome. OGRBench evaluates GUI rewards across platforms, with AndroidWorld experiments examining training and trajectory filtering.", "topics": ["reward-models", "reinforcement-learning", "evaluation"], "updated_at": "2026-03-19", "url": "https://arxiv.org/abs/2603.19191"}
{"aliases": [], "authors": ["Taiyi Wang", "Sian Gooding", "Florian Hartmann", "Oriana Riva", "Edward Grefenstette"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5538933a0e3d11df6ff4d767849979703d83412599fbbb0f8b432c4d9850633b", "submitted_at": "2026-03-20T06:32:16Z", "updated_at": "2026-03-20T06:32:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.19685v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}], "id": "arxiv:2603.19685", "identifiers": {"arxiv": "2603.19685"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:16Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.19685"}], "name": "A Subgoal-driven Framework for Improving Long-Horizon LLM Agents", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-03-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "The framework separates inference-time subgoal planning from MiRA, a reinforcement-learning method that supplies milestone rewards. WebArena-Lite experiments evaluate these mechanisms for long-horizon web navigation.", "topics": ["planning", "reinforcement-learning", "reward-models", "long-horizon", "evaluation"], "updated_at": "2026-03-20", "url": "https://arxiv.org/abs/2603.19685"}
{"aliases": [], "authors": ["Yuzhe Zhang", "Xianwei Xue", "Xingyong Wu", "Mengke Chen", "Chen Liu", "Xinran He", "Run Shao", "Feiran Liu", "Huanmin Xu", "Qiutong Pan", "Haiwei Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7c9ef656cbee786bec0d37ac4915bfd04adb36410eda8e647fab0d8605f90753", "submitted_at": "2026-04-07T06:16:23Z", "updated_at": "2026-04-07T06:16:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.05477v1"}}, "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:2604.05477", "identifiers": {"arxiv": "2604.05477"}, "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/2604.05477"}], "name": "Don't Act Blindly: Robust GUI Automation via Action-Effect Verification and Self-Correction", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces VeriGUI, which checks expected action effects and reasons about recovery before subsequent GUI actions. Training combines synthetic failure trajectories with verification-weighted GRPO, evaluated on an AndroidControl-based robustness benchmark.", "topics": ["robustness", "reinforcement-learning", "state-tracking"], "updated_at": "2026-04-07", "url": "https://arxiv.org/abs/2604.05477"}
{"aliases": [], "authors": ["Guo Gan", "Yuxuan Ding", "Cong Chen", "Yuwei Ren", "Yin Huang", "Hong Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f4d0dbeac950760f4f676ab7c651a881fd275f4adb69d46a94a11c4f10a03980", "submitted_at": "2026-04-08T16:40:26Z", "updated_at": "2026-05-08T03:06:31Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.07277v2"}}, "environments": ["android"], "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:2604.07277", "identifiers": {"arxiv": "2604.07277"}, "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/2604.07277"}], "name": "Android Coach: Improve Online Agentic Training Efficiency with Single State Multiple Actions", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Android Coach, which uses a learned critic to evaluate multiple candidate actions at each observed emulator state. Process rewards and a group-wise advantage estimator support online training, evaluated on AndroidLab and AndroidWorld.", "topics": ["reinforcement-learning", "reward-models", "efficiency"], "updated_at": "2026-05-08", "url": "https://arxiv.org/abs/2604.07277"}
{"aliases": [], "authors": ["Fei Tang", "Zhiqiong Lu", "Boxuan Zhang", "Weiming Lu", "Jun Xiao", "Yueting Zhuang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "df56b38203bc5ec8d7e871eb2ad9fca9891a049df62b7d72796fe400eb8bce5e", "submitted_at": "2026-04-13T17:52:04Z", "updated_at": "2026-04-13T17:52:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.11784v1"}}, "environments": ["android", "harmonyos", "ios"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.11784", "identifiers": {"arxiv": "2604.11784"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.11784"}], "name": "ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents ClawGUI, a shared harness for GUI-agent reinforcement learning, benchmark evaluation and deployment on mobile devices. It combines parallel virtual and physical environments, process rewards, and hybrid CLI-GUI execution with persistent memory.", "topics": ["reinforcement-learning", "evaluation", "tool-integration", "mobile-navigation"], "updated_at": "2026-04-13", "url": "https://arxiv.org/abs/2604.11784"}
{"aliases": [], "authors": ["Yulin Chen", "Tri Cao", "Haoran Li", "Yue Liu", "Yibo Li", "Yufei He", "Le Minh Khoi", "Yangqiu Song", "Shuicheng Yan", "Bryan Hooi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "64d4323050a5ad0f7265231337eb22a4c0823cee41587099c8bdc5db6b320dea", "submitted_at": "2026-04-14T04:50:35Z", "updated_at": "2026-04-14T04:50:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.12284v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.12284", "identifiers": {"arxiv": "2604.12284"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.12284"}], "name": "WebAgentGuard: A Reasoning-Driven Guard Model for Detecting Prompt Injection Attacks in Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebAgentGuard separates prompt-injection detection from the operating web agent through a parallel multimodal guard. The guard is trained on synthetic visual/textual examples using reasoning-focused supervised fine-tuning followed by reinforcement learning.", "topics": ["safety-security", "data-generation", "imitation-learning", "reinforcement-learning", "evaluation"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2604.12284"}
{"aliases": [], "authors": ["Wenkai Wang", "Xiyun Li", "Hongcan Guo", "Wenhao Yu", "Tianqing Fang", "Haitao Mi", "Dong Yu", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5e15661746ab05f2894e23d28b4675e13919bbcd71407fbc33260d84f330ea5d", "submitted_at": "2026-04-23T04:23:31Z", "updated_at": "2026-04-23T04:23:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.21268v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.21268", "identifiers": {"arxiv": "2604.21268"}, "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/2604.21268"}], "name": "Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding", "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": "2026-04-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains a GUI grounding proposer together with a visual critic that evaluates candidate locations rendered on screenshots. An adaptive reinforcement-learning schedule balances their development, using proposal diversity and critic feedback during training.", "topics": ["grounding", "reinforcement-learning", "self-reflection"], "updated_at": "2026-04-23", "url": "https://arxiv.org/abs/2604.21268"}
{"aliases": [], "authors": ["Haowen Hu", "Pengzhou Cheng", "Zheng Wu", "Lingzhong Dong", "Gongshen Liu", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8738f2584f096c1d54555adcf8bdbbb283f0d3178eb1f5e7b7c8a61b2965c15e", "submitted_at": "2026-05-02T02:54:36Z", "updated_at": "2026-05-02T02:54:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.01208v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.01208", "identifiers": {"arxiv": "2605.01208"}, "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/2605.01208"}], "name": "Faithful Mobile GUI Agents with Guided Advantage Estimator", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-05-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Faithful-Agent, which first teaches abstention under perturbed evidence and then refines GUI policies with an anchor-based, variance-adaptive advantage estimator. A thought-action consistency reward encourages decisions grounded in displayed evidence and instructions.", "topics": ["reinforcement-learning", "robustness", "mobile-navigation"], "updated_at": "2026-05-02", "url": "https://arxiv.org/abs/2605.01208"}
{"aliases": [], "authors": ["Oğuzhan Fatih Kar", "Roman Bachmann", "Yuanzheng Gong", "Anders Boesen Lindbo Larsen", "Afshin Dehghan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2e1fb158dad88d32f8754ac27f12346ba8fbe5f9a5e736fba3e81390071895a3", "submitted_at": "2026-05-07T17:17:10Z", "updated_at": "2026-05-07T17:17:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.06761v1"}}, "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:2605.06761", "identifiers": {"arxiv": "2605.06761"}, "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/2605.06761"}], "name": "Weblica: Scalable and Reproducible Training Environments for Visual Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-05-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Weblica, which combines HTTP-level capture and replay with website-grounded environment synthesis. The resulting interactive web environments support reinforcement learning and tests of navigation transfer and inference-time scaling.", "topics": ["sandbox", "data-generation", "reinforcement-learning"], "updated_at": "2026-05-07", "url": "https://arxiv.org/abs/2605.06761"}
{"aliases": [], "authors": ["Yilin Zhang", "Yingkai Hua", "Chunyu Wei", "Xin Wang", "Yueguo Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b57d784c70b50d2c25055c6d139c0a5acfc97da3a4effbd63536b72d6df38a69", "submitted_at": "2026-05-10T12:11:52Z", "updated_at": "2026-05-10T12:11:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.09497v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.09497", "identifiers": {"arxiv": "2605.09497"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.09497"}], "name": "Don't Click That: Teaching Web Agents to Resist Deceptive Interfaces", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "DUDE combines asymmetric hybrid-reward learning with summaries of deceptive-interface failures to guide safer web actions. Its RUC benchmark evaluates deception susceptibility alongside ordinary task performance across interface scenarios.", "topics": ["safety-security", "reinforcement-learning", "evaluation", "robustness"], "updated_at": "2026-05-10", "url": "https://arxiv.org/abs/2605.09497"}
{"aliases": [], "authors": ["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": ["Bowen Wang", "Dunjie Lu", "Junli Wang", "Tianyi Bai", "Shixuan Liu", "Zhipeng Zhang", "Haiquan Wang", "Hao Hu", "Tianbao Xie", "Shuai Bai", "Dayiheng Liu", "Que Shen", "Junyang Lin", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ecc0712f55ee90c237f9a48e013e396b1082f932a485c22d8249d1da7acbf307", "submitted_at": "2026-05-25T09:28:03Z", "updated_at": "2026-06-08T08:32:35Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.25624v2"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.25624", "identifiers": {"arxiv": "2605.25624"}, "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/2605.25624"}], "name": "CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The abstract promises future open sourcing; this annotation does not independently verify artifact release or licensing.", "published_at": "2026-05-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents CUA-Gym, which co-generates instructions, initial and target states, and executable reward checks through generator-discriminator iterations. Mock web applications expand training environments, with learned agents evaluated on OSWorld-Verified and WebArena.", "topics": ["data-generation", "reinforcement-learning", "sandbox"], "updated_at": "2026-06-08", "url": "https://arxiv.org/abs/2605.25624"}
{"aliases": [], "authors": ["Dingbang Wu", "Rui Hao", "Haiyang Wang", "Shuzhe Wu", "Han Xiao", "Zhenghong Li", "Bojiang Zhou", "Zheng Ju", "Zichen Liu", "Lue Fan", "Zhaoxiang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "145c7012fb15a85b203924ecd559c8b20d18b51bf79805a6751fec75d6f68dec", "submitted_at": "2026-05-25T17:59:49Z", "updated_at": "2026-09-23T17:59:57Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2605.26114v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.26114", "identifiers": {"arxiv": "2605.26114"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.26114"}], "name": "MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced. The simulator is browser-hosted; the abstract does not identify a native mobile operating system.", "published_at": "2026-05-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MobileGym, a browser-hosted simulation of mobile app interactions whose configurable JSON state supports deterministic judging and parallel rollouts. MobileGym-Bench adds parameterized tasks, with experiments examining online training and transfer to real devices.", "topics": ["sandbox", "evaluation", "reinforcement-learning"], "updated_at": "2026-09-23", "url": "https://arxiv.org/abs/2605.26114"}
{"aliases": [], "authors": ["Yuxuan Liu", "Xin Lai", "Junyi Li", "Pengyuan Lyu", "Jason", "Yiduo Guo", "Zhengyao Fang", "Yang Ding", "Yi Zhang", "Weinong Wang", "Huawen Shen", "Xingran Zhou", "Liang Wu", "Fei Tang", "Sunqi Fan", "Shangpin Peng", "Zheng Ruan", "Anran Zhang", "Chengquan Zhang", "Han Hu", "Benyou Wang", "Ji-Rong Wen", "Rui Yan", "Zhengyang Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a41cb1a93e6aa6ba8fbc7dca55415577ab5880a6592e9b0bee5b12b5a7732cbf", "submitted_at": "2026-05-28T07:14:15Z", "updated_at": "2026-09-28T15:27:12Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2605.29486v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:03Z", "role": "primary", "sha256": "2adad07a9ec7189aea25e48c76fa28206eda97f8974d5745cadda9529103bb1a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2605.06761%2C2605.07110%2C2605.07134%2C2605.09497%2C2605.11212%2C2605.12481%2C2605.12501%2C2605.12549%2C2605.13631%2C2605.14290%2C2605.14747%2C2605.15030%2C2605.15542%2C2605.15777%2C2605.16024%2C2605.16116%2C2605.16402%2C2605.17320%2C2605.18048%2C2605.18652%2C2605.18758%2C2605.19260%2C2605.19769%2C2605.21470%2C2605.25160%2C2605.25624%2C2605.25707%2C2605.26114%2C2605.27761%2C2605.28116%2C2605.28534%2C2605.28646%2C2605.28775%2C2605.29119%2C2605.29324%2C2605.29447%2C2605.29486%2C2605.29534%2C2605.29615%2C2606.06322%2C2606.09426%2C2606.11042%2C2606.22557%2C2606.23189%2C2606.24551%2C2608.17659%2C2609.22000%2C2609.34139&max_results=48"}], "id": "arxiv:2605.29486", "identifiers": {"arxiv": "2605.29486"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:03Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.29486"}], "name": "PhoneWorld: From Real-App Trajectories to Dynamic and Verifiable Environments for Phone-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-05-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces PhoneWorld, which converts observed app pages, transitions and state changes into executable Android environments through a build-inspect-repair process. Generated tasks and programmatic verifiers support benchmarking, trajectory collection and online training.", "topics": ["data-generation", "sandbox", "reinforcement-learning", "evaluation"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2605.29486"}
{"aliases": [], "authors": ["Hao Bai", "Rui Yang", "Chenlu Ye", "Spencer Whitehead", "Aviral Kumar", "Tong Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8a8afe44ffa73739505070c5f8004d8b4bbe98455fc0f5f3cef4b450bf441c74", "submitted_at": "2026-06-04T02:18:44Z", "updated_at": "2026-08-07T17:58:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2606.05597v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.05597", "identifiers": {"arxiv": "2606.05597"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.05597"}], "name": "AsyncWebRL: Efficient Asynchronous Reinforcement Learning for Multi-Step Visual Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "AsyncWebRL overlaps web-agent rollouts, updates and policy refresh while managing screenshot handling and a persistent rollout pool. It also replaces trajectory-length normalization in multi-step GRPO to reduce verbose failed trajectories, with throughput and out-of-distribution evaluation.", "topics": ["reinforcement-learning", "efficiency", "web-navigation"], "updated_at": "2026-08-07", "url": "https://arxiv.org/abs/2606.05597"}
{"aliases": [], "authors": ["Haojie Hao", "Longkun Hao", "Yihang Lou", "Yan Bai", "Zhenyang Li", "Zhichao Yang", "Dongshuo Huang", "Hongyu Lin", "Lanqing Hong", "Jiakai Wang", "Xianglong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2d530152b210a69cb458fecb036c10621c72daccfbcac7c8306c7e248c87cf14", "submitted_at": "2026-06-05T08:17:28Z", "updated_at": "2026-06-12T07:28:19Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.07027v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.07027", "identifiers": {"arxiv": "2606.07027"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.07027"}], "name": "StainFlow: Entity-Stain Tracking and Evidence Linking for Process Rewards in GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "StainFlow tracks visually verifiable task entities through GUI trajectories to identify phases and retrieve focused evidence windows. Its process-reward model is tested for both online policy learning and trajectory-completion judgments.", "topics": ["reward-models", "state-tracking", "reinforcement-learning", "evaluation"], "updated_at": "2026-06-12", "url": "https://arxiv.org/abs/2606.07027"}
{"aliases": [], "authors": ["Bojie Rong", "Zheyu Shen", "Qiaoping Wang", "Pengfei Kang", "Yang Xu", "Yawen Wei", "Hanyu Wu", "Zhi Zhao", "Leihao Pei", "Linquan Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0a852c77650fa54f234794c6cd27777f4e7219b8389b509d807e27db16138a1b", "submitted_at": "2026-06-08T12:55:42Z", "updated_at": "2026-06-08T12:55:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.09447v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.09447", "identifiers": {"arxiv": "2606.09447"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.09447"}], "name": "AliyunConsoleAgent: Training Web Agents in Real-World Cloud Environments via Distillation and Reinforcement Learning", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "AliyunConsoleAgent trains browser agents to verify documented procedures against changing cloud-console UIs through trajectory distillation and online GRPO. Provisioned rollout environments and backend audit-log checks support training and outcome evaluation.", "topics": ["web-navigation", "imitation-learning", "reinforcement-learning", "evaluation"], "updated_at": "2026-06-08", "url": "https://arxiv.org/abs/2606.09447"}
{"aliases": [], "authors": ["Can Lin", "Tao Feng", "Hangjie Yuan", "Dan Zhang", "Yifan Zhu", "Zhonghong Ou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7523d3901755d8feeecbaf2f02dcc00fd0a40102e1498d49b9f789b0a7c02d75", "submitted_at": "2026-06-09T07:52:10Z", "updated_at": "2026-07-06T06:23:24Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.10522v2"}}, "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:2606.10522", "identifiers": {"arxiv": "2606.10522"}, "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/2606.10522"}], "name": "GUI-AC: Enhancing Continual Learning in GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-AC uses grounding certainty to adjust advantage weighting and policy-clipping bounds during continual reinforcement fine-tuning. Its mechanisms aim to suppress noisy confidence updates while retaining exploration under shifting GUI domains and resolutions.", "topics": ["reinforcement-learning", "grounding", "robustness"], "updated_at": "2026-07-06", "url": "https://arxiv.org/abs/2606.10522"}
{"aliases": [], "authors": ["Xinyu Qiu", "Yunzhu Zhang", "Heng Jia", "Shuheng Shen", "Changhua Meng", "Linchao Zhu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1a5e949a760f236c08ae3cebc32d8a52220b1101f5324770d40bd7d054351f6a", "submitted_at": "2026-06-12T15:58:18Z", "updated_at": "2026-06-12T15:58:18Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.14579v1"}}, "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:2606.14579", "identifiers": {"arxiv": "2606.14579"}, "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/2606.14579"}], "name": "VISTA: View-Consistent Self-Verified Training for GUI Grounding", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "VISTA forms GRPO comparison groups from target-preserving crops of the same GUI instance with exactly remapped target boxes. A conditional, advantage-weighted cross-view anchor supplies additional supervision without entering the group baseline, and evaluation includes view-consistency robustness.", "topics": ["grounding", "reinforcement-learning", "robustness", "data-generation"], "updated_at": "2026-06-12", "url": "https://arxiv.org/abs/2606.14579"}
{"aliases": [], "authors": ["Guangyi Liu", "Pengxiang Zhao", "Gao Wu", "Yiwen Yin", "Mading Li", "Liang Liu", "Congxiao Liu", "Zhang Qi", "Mengyan Wang", "Liang Guo", "Jiangning Zhang", "Yong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cab809618ea4e218da26b4ae11bf937859c3a4bc240c91dd27b3a0d333737069", "submitted_at": "2026-06-18T08:29:33Z", "updated_at": "2026-08-07T03:54:59Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.19930v2"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.19930", "identifiers": {"arxiv": "2606.19930"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.19930"}], "name": "MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileForge couples real-app exploration and generated tasks with hierarchical feedback for mobile-agent adaptation. Its optimizer turns trajectory outcomes, step feedback and corrective hints into contextualized GRPO updates, with evaluation on AndroidWorld and a GUI-only MobileWorld split.", "topics": ["mobile-navigation", "reinforcement-learning", "data-generation", "evaluation"], "updated_at": "2026-08-07", "url": "https://arxiv.org/abs/2606.19930"}
{"aliases": [], "authors": ["Yuexing Hao", "Xiaomin Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b370491b37d5bf25c624f2bac654e438459c8084b5da292fb18a0bf9142adaa2", "submitted_at": "2026-06-18T15:25:42Z", "updated_at": "2026-06-18T15:25:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.20363v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.20363", "identifiers": {"arxiv": "2606.20363"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.20363"}], "name": "Automating SKILL.md Generation for Computer-Using Agents via Interaction Trajectory Mining", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Segments and clusters GUI trajectories into inspectable skills, then trains a skill-aware policy from those annotations. Its diagnostic results distinguish readable source-domain clusters from reliable transfer and report limited or absent downstream gains.", "topics": ["memory", "data-generation", "reinforcement-learning", "evaluation"], "updated_at": "2026-06-18", "url": "https://arxiv.org/abs/2606.20363"}
{"aliases": [], "authors": ["Daniel Beechey", "Derek Yuen", "Jianheng Liu", "Dezhao Luo", "Tiantian He", "Weilin Luo", "Jun Wang", "Kun Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9e761b20caa6882605ecfd6cd001aeda4a2d36b2e702eea9a6d70451fc0ea9eb", "submitted_at": "2026-05-26T13:32:14Z", "updated_at": "2026-05-26T13:32:14Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.20622v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:00:51.952206+00:00", "role": "primary", "sha256": "5cd7fb95d5ee8e8c473e0c9d5b892bf50ee425604c6a320e8e9a12eff17120a1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:00:51.952206+00:00", "role": "primary", "sha256": "5cd7fb95d5ee8e8c473e0c9d5b892bf50ee425604c6a320e8e9a12eff17120a1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22UI+agent%22+OR+ti%3A%22desktop+agent%22+OR+ti%3A%22screen+grounding%22%29+AND+submittedDate%3A%5B200001010000+TO+202610180000%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:14:00.181958+00:00", "role": "primary", "sha256": "e4711466886372760daa83384cb4d129d54fdcfe320be6e91fb7b4971d1a61de", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/pdf/2606.20622v1"}], "id": "arxiv:2606.20622", "identifiers": {"arxiv": "2606.20622"}, "kind": "paper", "last_verified_at": "2026-10-18T10:00:51.952206+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.20622"}], "name": "Darwin Mobile Agent: A Roadmap for Self-Evolution", "notes": "The implemented stage uses selected tasks, an external judge and heuristic history management. Autonomous curricula, verification and memory are roadmap directions, not completed capabilities established by this record. No runtime or results were reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (2 Agent-Environment Interface and Policy Optimisation; 3 Scalable Asynchronous Mobile Environment; 3.2 Persistent State Management; 3.3 Task Lifecycle; 4.2 Rollout and Optimisation; 4.3 Anatomy of the Self-Evolving Agent); not an entire-paper reading or reproduction claim.", "published_at": "2026-05-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes asynchronous reinforcement-learning infrastructure using parallel cloud-hosted Android instances, explicit task lifecycles, and persistent-state management for mobile GUI interaction.", "topics": ["reinforcement-learning", "mobile-navigation", "evaluation"], "updated_at": "2026-05-26", "url": "https://arxiv.org/abs/2606.20622"}
{"aliases": [], "authors": ["Yincheng Zhou", "Athena Zhuoming Zhong", "Shijie Zhang", "Kevin Zhang", "Teresa Xiaotao Shang", "Shanghang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fbd8892e06a5e847c16aa3101b98b3d669a783b84ee8e3d4d4d810754914028b", "submitted_at": "2026-06-22T07:28:55Z", "updated_at": "2026-06-22T07:28:55Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.22948v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.22948", "identifiers": {"arxiv": "2606.22948"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.22948"}], "name": "ENVS: Environment-Native Verified Search for Long-Horizon GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "ENVS branches over GUI actions in live desktop VMs, verifies successful leaves, and converts them into balanced step-level training supervision. It also introduces OSWorld-Noisy to test recovery from dynamic interruptions while retaining the underlying tasks.", "topics": ["data-generation", "reinforcement-learning", "long-horizon", "robustness", "evaluation"], "updated_at": "2026-06-22", "url": "https://arxiv.org/abs/2606.22948"}
{"aliases": [], "authors": ["Marta Sumyk", "Oleksandr Kosovan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0aeb8c28b570133f575c2f336b6dab0dcbecdfaec78549115407b21722816e8a", "submitted_at": "2026-06-23T12:46:29Z", "updated_at": "2026-06-23T12:46:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.24515v1"}}, "environments": ["desktop", "macos", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.24515", "identifiers": {"arxiv": "2606.24515"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.24515"}], "name": "Reinforcement Learning for Computer-Use Agents with Autonomous Evaluation", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "The framework uses a vision-language judge’s final-screen assessment as a terminal reward for GUI reinforcement learning. It models evaluator noise and corrects the reward estimator used by PPO across desktop benchmarks.", "topics": ["reinforcement-learning", "reward-models", "evaluation", "robustness"], "updated_at": "2026-06-23", "url": "https://arxiv.org/abs/2606.24515"}
{"aliases": [], "authors": ["Chengheng Li-Chen", "Zhiqian Zhou", "Hao Chen", "Nicolas Chauvin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7a8efe4b9476d7f55ca4da17621ece653be2e33bf02be0183a339a12f5fac3ca", "submitted_at": "2026-06-24T15:35:29Z", "updated_at": "2026-06-24T15:35:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.25964v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.25964", "identifiers": {"arxiv": "2606.25964"}, "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/2606.25964"}], "name": "WinDOM: Self-Family Distillation for Small-Model GUI Grounding", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "WinDOM obtains bounding-box supervision from a Windows 11 web reimplementation driven through Playwright, and studies rejection-sampling distillation for small GUI grounders. It treats the depth of the supervised cold start as a GRPO choice, comparing same-size EMA and larger same-family teachers.", "topics": ["grounding", "data-generation", "imitation-learning", "reinforcement-learning"], "updated_at": "2026-06-24", "url": "https://arxiv.org/abs/2606.25964"}
{"aliases": [], "authors": ["Sunqi Fan", "Lingshan Chen", "Runqi Yin", "Qingle Liu", "Yongming Rao", "Meng-Hao Guo", "Shi-Min Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7beb70ac91463b77a8babaf2b077c46c5072e8f86d22f4d9afc9c1e77680f309", "submitted_at": "2026-06-29T02:16:21Z", "updated_at": "2026-06-29T02:16:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.29705v1"}}, "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:2606.29705", "identifiers": {"arxiv": "2606.29705"}, "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/2606.29705"}], "name": "GUICrafter: Weakly-Supervised GUI Agent Leveraging Massive Unannotated Screenshots", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUICrafter first learns grounding from large collections of unannotated screenshots and webpages, then uses a smaller high-quality dataset for reinforcement-learning calibration. The curriculum targets reduced dependence on manually annotated GUI training data.", "topics": ["grounding", "data-generation", "reinforcement-learning", "screen-understanding"], "updated_at": "2026-06-29", "url": "https://arxiv.org/abs/2606.29705"}
{"aliases": [], "authors": ["Wanxia Cao", "Chengzhen Duan", "Pei Fu", "Pengzhi Gao", "Niu Lian", "Fazhan Liu", "Hui Liu", "Heng Qu", "Qinzhuo Wu", "Zhehao Yu", "Tongbo Chen", "Shiqi Cui", "Anan Du", "Shukai Jia", "Yuanfa Li", "Wei Liu", "Yike Liu", "Wenchao Lu", "Zhenbo Luo", "Haoyuan Sun", "Jiatong Sun", "Cheng Tan", "Yajie Wang", "Changqiao Wu", "Tao Xiong", "Jiahui Yang", "Yuxuan Yuan", "Ruoceng Zhang", "Shaojie Zhang", "Jian Zhu", "Jian Luan", "Cong Zou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "974d471d25373502914bf0d146d5d4a54a5751281c04921a9d331d83bce64b33", "submitted_at": "2026-06-30T09:36:35Z", "updated_at": "2026-07-01T03:59:21Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.31410v2"}}, "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", "relations"], "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:2606.31410", "identifiers": {"arxiv": "2606.31410"}, "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/2606.31410"}], "name": "Xiaomi-GUI-0 Technical Report", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a mobile GUI agent trained through real-device interaction, corrected failure trajectories, supervised learning, and reinforcement learning.", "topics": ["mobile-navigation", "reinforcement-learning", "robustness"], "updated_at": "2026-07-01", "url": "https://arxiv.org/abs/2606.31410"}
{"aliases": [], "authors": ["Mianqiu Huang", "Taofeng Xue", "Chong Peng", "Jinrui Ding", "Jie Yang", "Sicheng Fan", "Jiale Hong", "Yufei Gao", "Xiaocheng Zhang", "Linsen Guo", "Xin Yang", "Dengchang Zhao", "Yuchen Xie", "Peng Pei", "Xunliang Xie", "Xipeng Qiu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "53c254c53a834d4f28bb4b65317121f5c46f5dd6985d03a30f1cc1f0fa78707d", "submitted_at": "2026-07-07T16:36:25Z", "updated_at": "2026-09-04T15:47:10Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.09773v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.09773", "identifiers": {"arxiv": "2607.09773"}, "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/2607.09773"}], "name": "EvoCUA-1.5: Online Reinforcement Learning for Multi-turn 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": "2026-07-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "EvoCUA-1.5 trains multi-turn desktop agents through online reinforcement learning in executable sandboxes. It combines step-level policy optimization, task filtering and calibration, an adaptive curriculum, and asynchronous rollout handling with staleness control.", "topics": ["reinforcement-learning", "long-horizon", "sandbox", "data-generation"], "updated_at": "2026-09-04", "url": "https://arxiv.org/abs/2607.09773"}
{"aliases": [], "authors": ["Chengguang Gan", "Hanjun Wei", "Yunhao Liang", "Zhixi Cai", "Qinghao Zhang", "Shiwen Ni"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "37cfa436d182ed8fbc8823213b9173fbfcf9fb421e8d2ae89a093a5393c374ba", "submitted_at": "2026-07-11T02:08:46Z", "updated_at": "2026-07-16T03:28:43Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2607.10079v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.10079", "identifiers": {"arxiv": "2607.10079"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.10079"}], "name": "MAG: A Web-Agent Benchmark and Harness for Multimodal Action and Guide Generation", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAG jointly evaluates browser task actions and user-facing guides grounded in screenshots through element marks or pixel coordinates. Its harness covers annotation, live evaluation and joint metrics, with an expert-trajectory-augmented GRPO training study.", "topics": ["evaluation", "grounding", "reinforcement-learning", "action-prediction"], "updated_at": "2026-07-16", "url": "https://arxiv.org/abs/2607.10079"}
{"aliases": [], "authors": ["Bowen Lv", "Xiao Liu", "Yanyu Ren", "Hanyu Lai", "Bohao Jing", "Hanchen Zhang", "Yanxiao Zhao", "Shuntian Yao", "Jie Tang", "Yuxiao Dong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "554844f23cddbdaadc3c52e93d07d06c756faadbd0136ffef0984b25423f84f8", "submitted_at": "2026-07-13T07:32:35Z", "updated_at": "2026-07-13T07:32:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.11185v1"}}, "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:2607.11185", "identifiers": {"arxiv": "2607.11185"}, "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/2607.11185"}], "name": "SCALECUA: Scaling Computer Use Agents with Verifiable Task Synthesis and Efficient Online RL", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Scales computer-agent reinforcement learning through verifiable task generation, capability-aware rollout allocation, and segmented visual context.", "topics": ["data-generation", "reinforcement-learning", "efficiency"], "updated_at": "2026-07-13", "url": "https://arxiv.org/abs/2607.11185"}
{"aliases": [], "authors": ["Chengguang Gan", "Zhixi Cai", "Yunhao Liang", "Hanjun Wei", "Shiwen Ni", "Qinghao Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e96ea0cf71ff705c064db1bcd1d2cbf8b8cd712bb4e8030931cafbb31ff29b10", "submitted_at": "2026-07-14T11:17:35Z", "updated_at": "2026-07-14T11:17:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.12640v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.12640", "identifiers": {"arxiv": "2607.12640"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.12640"}], "name": "A Learning-Rate-Gated Failure of GRPO in a Small Language and Vision-Language Model Web Agent: A Controlled Null and Its Mechanism", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Studies a controlled setting where GRPO fails to improve a strong small web agent and can reduce text-track performance. Matched reachable-reward tasks and component grafting distinguish limited optimization headroom from learning-rate-dependent failure mechanisms.", "topics": ["reinforcement-learning", "evaluation", "robustness"], "updated_at": "2026-07-14", "url": "https://arxiv.org/abs/2607.12640"}
{"aliases": [], "authors": ["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": ["Yash Pandya", "Sahil Gupta", "Sarthak Harne", "Archana Yadav", "Kavyansh Chourasia", "Hussein Mozannar", "Vibhav Vineet", "Sara Abdali", "Corby Rosset", "Yash Lara", "Ahmed Awadallah", "Ece Kamar", "Akshay Nambi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6f28e1ee8e101000f3cfe9215555251656fcb40b921b5c41c895fbc958f867e3", "submitted_at": "2026-07-30T11:48:22Z", "updated_at": "2026-07-30T11:48:22Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.28074v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.28074", "identifiers": {"arxiv": "2607.28074"}, "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/2607.28074"}], "name": "Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Echoverse, which compiles specifications into stateful applications with database-grounded graders and uses rollouts to improve both the environments and agent policies. It examines behavioral depth, targeted interface practice, and environment repair as sources of training value.", "topics": ["sandbox", "data-generation", "reinforcement-learning"], "updated_at": "2026-07-30", "url": "https://arxiv.org/abs/2607.28074"}
{"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": ["Dunjie Lu", "Shuai Bai", "Tianyi Bai", "Sicheng Fan", "Chang Gao", "Jian Guan", "Feng Hu", "Mianqiu Huang", "Xingyang Huang", "Yizhen Jiang", "Yuheng Jing", "Dehui Kong", "Ning Li", "Dayiheng Liu", "Shixuan Liu", "Zheng Liu", "Que Shen", "Bowen Wang", "Junli Wang", "Chencan Wu", "Rui Xie", "Tianbao Xie", "Zhihui Xie", "Haiyang Xu", "An Yang", "Tao Yu", "Wenzhen Yuan", "Xi Zhang", "Zhenru Zhang", "Mingkang Zhu", "Zhaoqing Zhu", "Yizhong Cao", "Kai Dang", "Binyuan Hui", "Kaixin Li", "Junyang Lin", "Haiquan Wang", "Zekun Wang", "Yiheng Xu", "Fan Yan", "Mengqi Yuan", "Danyang Zhang", "Jiajun Zhang", "Zhipeng Zhang", "Fan Zhou", "Fan Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "26a61ccfdddb5cdc2da2b14d29ae8d4098544fea4e9458126f2a49c28418c4f0", "submitted_at": "2026-08-03T15:04:20Z", "updated_at": "2026-08-03T15:04:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.02352v1"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.02352", "identifiers": {"arxiv": "2608.02352"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.02352"}], "name": "Qwen-CUA: Native Computer Use for (almost) Everything", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Qwen-CUA operates through screenshots and keyboard/mouse events while compressing older visual history into reusable context blocks. Its report combines verifiable task generation, trajectory-based training and separate browser and Bash-augmented studies.", "topics": ["reinforcement-learning", "imitation-learning", "data-generation", "interaction-history", "long-horizon", "evaluation"], "updated_at": "2026-08-03", "url": "https://arxiv.org/abs/2608.02352"}
{"aliases": [], "authors": ["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": ["Weikai Xu", "Yunren Feng", "Haoxiang Lei", "Kun Huang", "Yuxuan Liu", "Kang Zhao", "Xiaolin Hu", "Shuo Shang", "Bo An"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ce37dd8217afbaa07ddea25c0daf78cdbfb449851ca6ab17b7f3ecacd451eb2f", "submitted_at": "2026-08-06T11:15:41Z", "updated_at": "2026-08-06T11:15:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.05891v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.05891", "identifiers": {"arxiv": "2608.05891"}, "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/2608.05891"}], "name": "AppDeltaWorld: Transition-Grounded Delta Code World Model for Mobile GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "AppDeltaWorld predicts executable code updates for the next mobile UI state, using transition-constrained retrieval and rendered visual assets. It evaluates reconstruction fidelity separately from policy training and adaptation using generated environments.", "topics": ["world-models", "data-generation", "imitation-learning", "reinforcement-learning", "mobile-navigation"], "updated_at": "2026-08-06", "url": "https://arxiv.org/abs/2608.05891"}
{"aliases": [], "authors": ["Renshan Zhang", "Haoyang Meng", "Yixiao He", "Rui Shao", "April Hua Liu", "Liqiang Nie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "87e9b3463232f606aea60df7d86626d5aa188e774b0fe515af43e1be49b1c66b", "submitted_at": "2026-08-10T15:23:52Z", "updated_at": "2026-08-10T15:23:52Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.09723v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.09723", "identifiers": {"arxiv": "2608.09723"}, "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/2608.09723"}], "name": "LookAgain: Closed-Loop GUI Grounding with Visually Grounded Reflection", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-08-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "LookAgain treats a GUI coordinate as a revisable hypothesis, rendering its marker and a local screenshot patch before confirmation. Reflective supervised trajectories provide a cold start, followed by GRPO using terminal grounding correctness.", "topics": ["grounding", "self-reflection", "reinforcement-learning", "imitation-learning"], "updated_at": "2026-08-10", "url": "https://arxiv.org/abs/2608.09723"}
{"aliases": [], "authors": ["AIMAE Team", "Tianxiang Chen", "Yan Cheng", "Zhangye Han", "Xiaowei Li", "Chang Liu", "Cheng Liu", "Zhongqiang Ma", "Long Peng", "Xiaobing Tu", "Yinggui Wang", "Hongliang Wei", "Chen Wu", "Daiping Xin", "Kunyu Zhou", "Pengyang Zhou", "Peiyuan Chen", "Ziyuan Chen", "Yutao Deng", "Chunyu Dong", "Xiangyu Fu", "Yicheng Feng", "Ruian He", "Haochen Li", "Miancan Liu", "Zhengqin Liu", "Wei Peng", "Jinkui Ren", "Haoyu Tan", "Dong Xiao", "Rongkun Xue", "Shujian Yang", "Xianhang Ye", "Ziqi Yuan", "Ziyang Yu", "Linghan Zhang", "Xiantao Zhang", "Xuanpu Zhao", "Yinan Zhao", "Zhenghui Zhao", "Bin Zhu", "Likai Zou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "474187aaa79165dc8652eec1e2af6b1733b2ee37b3cd125c89ff0defcaf2e193", "submitted_at": "2026-08-18T03:23:31Z", "updated_at": "2026-08-18T03:23:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.17319v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.17319", "identifiers": {"arxiv": "2608.17319"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.17319"}], "name": "Wuying-Browser-Agent: Real-World Centric Fundamental Long-Horizon Browser Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines a browser execution framework, recovery-oriented supervision, and online reinforcement learning, with a bilingual live-web task benchmark.", "topics": ["web-navigation", "long-horizon", "reinforcement-learning"], "updated_at": "2026-08-18", "url": "https://arxiv.org/abs/2608.17319"}
{"aliases": [], "authors": ["Chengyang Gu", "Le Zhang", "Jingbo Zhou", "Yize Chen", "Yu Shi", "Siqi Bao", "Zheng-Fan Wu", "Hua Wu", "Hui Xiong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9434f8da41679177f6978c039bd42fcacce543de2bd0e4ce05b62ab4d61ed761", "submitted_at": "2026-08-22T08:01:47Z", "updated_at": "2026-08-22T08:01:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.21830v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.21830", "identifiers": {"arxiv": "2608.21830"}, "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/2608.21830"}], "name": "Beyond Success and Failure: Length-Aware Contrastive Learning for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces LACL-GUI, a contrastive learning objective that distinguishes trajectory quality within both successful and failed GUI executions. It favors concise successes and uses divergence from successful trajectories to differentiate failures during policy optimization.", "topics": ["reinforcement-learning", "efficiency", "action-prediction"], "updated_at": "2026-08-22", "url": "https://arxiv.org/abs/2608.21830"}
{"aliases": [], "authors": ["Chenghao Zhang", "Yuxi Cheng", "Saisai Hu", "Canran Xiao", "Yikai Mao", "Dan Roth"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7d862afa969c0b4d1c6edc8d126017749844158f786b29c3b0710afecdec2008", "submitted_at": "2026-08-22T10:18:45Z", "updated_at": "2026-09-28T15:54:31Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.21898v2"}}, "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:2608.21898", "identifiers": {"arxiv": "2608.21898"}, "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/2608.21898"}], "name": "Training Needs Trustworthy Worlds: Verified Synthetic Web Environments for Agent 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": "2026-08-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Builds synthetic websites with verified navigation, backend state, and feasible tasks to provide reliable supervision for web-agent training.", "topics": ["data-generation", "sandbox", "reinforcement-learning"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2608.21898"}
{"aliases": [], "authors": ["Long Zhang", "Yuhan Chen", "Chaoran Zhang", "Wanxia Cao", "Kun Huang", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Chenliang Li", "Lixin Zou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4c1198ce65d5a2fcffc51b37bc4553b35b9e0d809757dc4181747c57c7ae443d", "submitted_at": "2026-08-24T06:27:47Z", "updated_at": "2026-08-24T06:27:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.22847v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.22847", "identifiers": {"arxiv": "2608.22847"}, "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/2608.22847"}], "name": "GSAR: Goal-State-Anchor Rewards for Mobile GUI Agents with Self-Evolving Data Synthesis", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GSAR, which synthesizes GUI tasks and successful goal states, then annotates task-relevant interface elements as reward anchors. These anchors support offline trajectory verification and online policy training.", "topics": ["data-generation", "reward-models", "reinforcement-learning"], "updated_at": "2026-08-24", "url": "https://arxiv.org/abs/2608.22847"}
{"aliases": [], "authors": ["Guo Gan", "Yilun Zhao", "Cong Chen", "Jinbiao Wei", "Tingyu Song", "Zheyuan Yang", "Lin Fu", "Hong Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "51901009de5898c52a1b9d4958b698808a8f8b850d19afa32dd7a3f32d42a851", "submitted_at": "2026-08-25T05:57:25Z", "updated_at": "2026-08-25T05:57:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.24099v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.24099", "identifiers": {"arxiv": "2608.24099"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.24099"}], "name": "Are Android GUI Agents Robust Against Runtime Anomalies? AnTrap: Evaluating Agents in Dynamic Adversarial Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AnTrap, which injects task-solvable runtime anomalies into Android GUI-agent trajectories using a layered failure taxonomy. It compares model robustness and adversarial reinforcement learning to distinguish recoverable interface traps from persistent contextual failures.", "topics": ["evaluation", "robustness", "reinforcement-learning"], "updated_at": "2026-08-25", "url": "https://arxiv.org/abs/2608.24099"}
{"aliases": [], "authors": ["Tao Xiong", "Xavier Hu", "Wenkai Wang", "Qinzhuo Wu", "Changqiao Wu", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "76cd09f70bfec0d2a49cddae8adcf32ca37fb39bcade7c1d36d5f3dc3fa6bfa7", "submitted_at": "2026-08-25T07:40:12Z", "updated_at": "2026-08-25T07:40:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.24174v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2608.24174", "identifiers": {"arxiv": "2608.24174"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.24174"}], "name": "Task-Adaptive Rubrics for GUI Reward Modeling", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AdaptRubric, which retrieves reusable criteria for a GUI task family and then generates instruction-specific checks for concrete values and constraints. It evaluates these adaptive rubrics in offline reward assessment and online agent optimization.", "topics": ["reward-models", "evaluation", "reinforcement-learning"], "updated_at": "2026-08-25", "url": "https://arxiv.org/abs/2608.24174"}
{"aliases": [], "authors": ["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": ["Yu Han", "Tianwen Qian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "576702fae164cbcb4adc86a81115487aae7e1c40cf07a1ecf3ef7c4c7e822c3b", "submitted_at": "2026-08-27T07:09:28Z", "updated_at": "2026-08-27T07:09:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.27508v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.27508", "identifiers": {"arxiv": "2608.27508"}, "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/2608.27508"}], "name": "WM-R1: Training GUI Agents to Reason and leverage World Models with 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": "2026-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses learned world models for mobile-agent training rollouts and for reasoning about the consequences of candidate GUI actions.", "topics": ["world-models", "reinforcement-learning"], "updated_at": "2026-08-27", "url": "https://arxiv.org/abs/2608.27508"}
{"aliases": [], "authors": ["Jiahe Ying", "Wendong Bu", "Kaihang Pan", "Bingchen Miao", "Siyu Chen", "Wen Wang", "Xueming Jiang", "Juncheng Li", "Siliang Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "352699b04f3a9406d7cab95ce9b5fb000e3e98a28996777e75b7399a6290357a", "submitted_at": "2026-08-28T03:15:11Z", "updated_at": "2026-08-28T03:15:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.27866v1"}}, "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:2608.27866", "identifiers": {"arxiv": "2608.27866"}, "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/2608.27866"}], "name": "Iron: Intent-Aligned and Retrospective Dual Learning Framework for Enhancing Generalist Virtual Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Iron, a GUI-agent training framework that aligns low-level actions with high-level intent using cycle-consistency rewards. It also repurposes failed trajectories through hindsight reconstruction and evaluates transfer to other digital tasks.", "topics": ["reinforcement-learning", "data-generation", "grounding"], "updated_at": "2026-08-28", "url": "https://arxiv.org/abs/2608.27866"}
{"aliases": [], "authors": ["Venus Team", "Zhuohan Cai", "Haoxing Chen", "Jiaxuan Chen", "Weizhi Chen", "Changlong Gao", "Zhangxuan Gu", "Yuan Guo", "Yusong Hu", "Jianrong Jiang", "Jianguo Li", "Runze Li", "Jinzhen Lin", "Zhenyu Ma", "Changhua Meng", "Han Peng", "Xinyu Qiu", "Shuheng Shen", "Zhongyi Shui", "Weiqiang Wang", "Ming Wen", "Zhuoer Xu", "Hang Yan", "Kaiwen Yang", "Ruilin Yao", "Nanjun Yu", "Zhengwen Zeng", "Lianrui Zhang", "Yunzhu Zhang", "Zhe Zhao", "Beitong Zhou"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "34383fa6c6e5cca059c53fd0452d93eec9ac6a96271ddc14f81ddb96fdeaceec", "submitted_at": "2026-08-27T17:15:33Z", "updated_at": "2026-08-27T17:15:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.00028v1"}}, "environments": ["web", "desktop", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2609.00028", "identifiers": {"arxiv": "2609.00028"}, "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/2609.00028"}], "name": "UI-Venus-2 Technical Report", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a GUI foundation agent trained through expanded environments, function-grounded tasks, and trajectory verification with safety-aware execution mechanisms.", "topics": ["automation", "reinforcement-learning", "safety-security"], "updated_at": "2026-08-27", "url": "https://arxiv.org/abs/2609.00028"}
{"aliases": [], "authors": ["Tong Ye", "Kunyang Han", "Guozhi Wang", "Longqiang Luo", "Zhifeng Ding", "Yongxiang Zhang", "Xiaolei Shen", "Yuxuan Zhang", "Zhuping Zhang", "Tao Xu", "Yue Pan", "Yucheng Zhao", "Yupei Hu", "Yuanjiang Ouyang", "Danfeng Shen", "Runqi Lin", "Hongda Cai", "Zhaoxiong Wang", "Mengjia Yan", "Yingjie Zhong", "Chen Zhou", "Zeyu Zhang", "Xuwen Zhu", "Penggang Shi", "Mingcheng Luo", "Ziyang Wu", "Min Jin", "Mingfu Shen", "Zairong Xu", "Fan Zhang", "Hao Wang", "Liang Liu", "Zhulin Xie", "Lijun Yao", "Xiao Liang", "Liangmin Wen", "Liqiang Feng", "Feilong Wu", "Min Hu", "Min Chen", "Guanjing Xiong", "Xiaohu Ruan", "Xiaoxin Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cc15573bca49d453001ec22061f190e429704a5318751b041550e3610c469f4e", "submitted_at": "2026-09-11T03:36:35Z", "updated_at": "2026-09-15T06:36:04Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2609.12394v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.12394", "identifiers": {"arxiv": "2609.12394"}, "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/2609.12394"}], "name": "BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a mobile-agent training pipeline centered on real-device rollouts, recovery of useful supervision from failures, and evolving evaluation tasks.", "topics": ["mobile-navigation", "reinforcement-learning", "data-generation"], "updated_at": "2026-09-15", "url": "https://arxiv.org/abs/2609.12394"}
{"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": ["Zeyu Kang", "Zhenyun Yin", "Yang Zhang", "Shan He", "Shanzhe Lei", "Yanjiu Zhong", "Xinquan Chen", "Xuhong Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a88d287d5433009962a905dbdbf03359e1183e66e4a6015441ec28405f64a6ee", "submitted_at": "2026-08-27T05:43:50Z", "updated_at": "2026-09-22T06:39:18Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.22178v2"}}, "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:2609.22178", "identifiers": {"arxiv": "2609.22178"}, "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/2609.22178"}], "name": "Beyond Task Completion: Training Capable and Safe Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SCOPE, which trains computer-use agents on ordinary task demonstrations, safe continuations around hazards, and explicit refusals. Its SCOPE-Gen pipeline pairs verifiable tasks with environmental-risk variants to support joint capability and safety training.", "topics": ["safety-security", "reinforcement-learning", "data-generation"], "updated_at": "2026-09-22", "url": "https://arxiv.org/abs/2609.22178"}
{"aliases": [], "authors": ["Tingyu Qu", "Weigao Sun", "Yuecheng Liu", "Yucheng Zhao", "Yi Zhu", "Yifeng Ding", "Qiyi Wang", "Sihan Cao", "Pengkun Jiao", "Hanlei Xie", "Xiongwei Wu", "Qichao Wang", "Haodong Zhang", "Jiajun Liu", "Yuhao Wang", "Yuqing Xie", "Junpeng Zhao", "Long Chen", "Ming Ma", "Sihan Yang", "Ziwang Zhao", "Yanhao Jia", "Liangquan Gong", "Feida Zhu", "Yiran Zhong", "Steven Hoi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "572ba5397fc1e4bb138febc3e7004e6e4f244e0ba8b628fb91debef4a854561f", "submitted_at": "2026-09-24T14:38:08Z", "updated_at": "2026-09-24T14:38:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.29892v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2609.29892v1"], "reason": "Record the paper-specific arXiv document declaration; it does not establish code, data or model-weight permissions."}}}, "environments": ["unspecified"], "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": "2026-10-20T15:43:20.402537Z", "role": "primary", "sha256": "d4385231cea2965a48895eeebd2dd6000265dd7b43731a0ad88b0373a21d0308", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2609.29892v1"}], "id": "arxiv:2609.29892", "identifiers": {"arxiv": "2609.29892"}, "kind": "paper", "last_verified_at": "2026-10-20T15:43:20.402537Z", "licenses": {"code": null, "data": null, "document": {"basis": "The exact-v1 arXiv HTML infobox (#license-tr) declares License: CC BY 4.0 for this paper. This declaration is document-only.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T15:43:20.402537Z", "url": "https://arxiv.org/html/2609.29892v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.29892"}], "name": "Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents", "notes": "Primarily typed tool actions and structured feedback, not pixel-coordinate GUI control; backends combine programmatic sandboxes, LLM simulation and selected real-device sessions. No specific OS is established. Model/harness updates are reviewed offline, not autonomous serving-time changes. Co-evolution scores use an evolve/development set; output-cost estimates exclude input, tools, devices and harness processing. Memory comparisons differ in history access and processing cost; user authorization relies on trusted upstream services. No independent reproduction.", "published_at": "2026-09-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains a mobile planner with supervised trajectories and competence-adaptive RL, then couples typed tool execution with memory, skills and offline model-harness refinement.", "topics": ["planning", "data-generation", "reinforcement-learning", "memory", "mobile-navigation", "efficiency"], "updated_at": "2026-09-24", "url": "https://arxiv.org/abs/2609.29892"}
{"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": ["Zhenhao Zhang", "Zhaoyu Fan", "Haohan Ying", "Jingwen Hu", "Hancen Fan", "Junhao Zhou", "Zitian Chen", "Linchao Zhu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fa2bcfa05fcbd6937af6ba3d23e892e5a105248939538dcdc9d7b015daf38593", "submitted_at": "2026-09-25T20:04:24Z", "updated_at": "2026-09-25T20:04:24Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.31957v1"}}, "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:2609.31957", "identifiers": {"arxiv": "2609.31957"}, "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/2609.31957"}], "name": "CaptchaArena: A Large-Scale, Fine-Grained Dataset for Training Computer-Use Agents on Interactive CAPTCHAs", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CaptchaArena, a dataset of execution-verified interactive CAPTCHA solutions with screenshot–action trajectories, reasoning annotations, and pixel masks for irregular targets. The work trains a single policy across CAPTCHA types using supervised learning followed by verifier-reward reinforcement learning.", "topics": ["data-generation", "grounding", "reinforcement-learning"], "updated_at": "2026-09-25", "url": "https://arxiv.org/abs/2609.31957"}
{"aliases": [], "authors": ["Xin Yan", "Zhengbo Jiao", "Jiaqi Liu", "Zhenglin Wan", "SiYuan Ma", "Xuliang Yu", "Tianyi Jiang", "Chubin Zhang", "Pengfei Zhou", "Wangbo Zhao", "Xingrui Yu", "Bo An", "Yang You", "Ivor Tsang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d4083bbe61c846d93663cd41d6335a2d6a54e361df9c3bdcd636b312c436d1a4", "submitted_at": "2026-09-26T16:16:08Z", "updated_at": "2026-09-26T16:16:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32750v1"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.32750", "identifiers": {"arxiv": "2609.32750"}, "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/2609.32750"}], "name": "CUA-Sandbox: Efficient Environments for Computer-Use Agent Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CUA-Sandbox, which separates each rollout’s mutable state from shared initialized software runtimes. State-scoped execution and transactional resets or branches aim to reduce training-environment costs while retaining original interfaces and evaluators.", "topics": ["sandbox", "reinforcement-learning", "efficiency"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32750"}
{"aliases": [], "authors": ["Shengtian Yang", "Ziyu Xiong", "Kaibing Yang", "Guangfeng Cai", "Yewen Li", "Peng Jiang", "Gai Kun", "Qingpeng Cai", "Lei Feng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d45cee8117e302aa190d38c8298bf479936a5a27aa9b85716a76a2a90499069c", "submitted_at": "2026-09-29T07:51:57Z", "updated_at": "2026-09-29T07:51:57Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.36939v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.36939", "identifiers": {"arxiv": "2609.36939"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.36939"}], "name": "SCA: Spatial Credit Assignment for Reinforcement Learning of GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Spatial Credit Assignment, which uses click coordinates to refine group-relative reinforcement-learning updates for GUI grounding. It distinguishes failed clicks by spatial evidence while leaving the deployed action policy unchanged.", "topics": ["grounding", "reinforcement-learning", "evaluation"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.36939"}
{"aliases": [], "authors": ["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": ["Yong Du", "Tongbo Chen", "Zhengxi Lu", "Yizhou Liu", "Bofan Chen", "Tao Jiang", "Wenhao Xu", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8f3745b6b497c8819dc990bdb5d74e7af463d23654cbc7cdd76ade0faf2b3e0c", "submitted_at": "2026-09-30T17:40:09Z", "updated_at": "2026-10-01T17:50:14Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.40253v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.40253", "identifiers": {"arxiv": "2609.40253"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.40253"}], "name": "ComputerSD: Online Self-Distillation from Real-Time Feedback for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces ComputerSD, which turns feedback from executed GUI transitions into online self-distillation guidance and step-level scores. A GUI analyzer supplies these signals, which are combined with trajectory-level reinforcement learning in asynchronous training.", "topics": ["reinforcement-learning", "evaluation"], "updated_at": "2026-10-01", "url": "https://arxiv.org/abs/2609.40253"}
{"aliases": [], "authors": ["Yifan Zhang", "Yutong Dai", "Viraj Prabhu", "Zhiyuan Hu", "Ran Xu", "Zeyuan Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2f3e9860574951cd6b749d47da748e4aec36d4f5afe226f4a3d0d5459fd07935", "submitted_at": "2026-10-05T17:57:51Z", "updated_at": "2026-10-05T17:57:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.06829v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.06829", "identifiers": {"arxiv": "2610.06829"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.06829"}], "name": "CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CLIFT, which certifies agent-generated verification signals against training-time judges and reuses them for per-step rewards and trajectory selection. At inference, a frozen question bank helps compare candidate web rollouts without querying an external judge.", "topics": ["reinforcement-learning", "evaluation", "web-navigation"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2610.06829"}
{"aliases": [], "authors": ["Chinmay Savadikar", "Zhaoyu Zhang", "Mingyu Zhao", "Shuang Xie", "Han Li", "Tianfu Wu", "Lingyun Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4890856d5961395a9ac8c83c60b988da4adeaab1d99edd511e4046e4f99fd25d", "submitted_at": "2026-10-05T17:11:07Z", "updated_at": "2026-10-05T17:11:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.07118v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.07118", "identifiers": {"arxiv": "2610.07118"}, "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/2610.07118"}], "name": "AMBER: Training Long-Horizon Web Agents through Append-Only Memory", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AMBER, an append-only memory framework in which a web agent learns to reason, act, and retain evidence using outcome-reward reinforcement learning. It studies whether preserving entries avoids the loss of facts and corrective feedback seen with rewritten memory.", "topics": ["memory", "long-horizon", "reinforcement-learning"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2610.07118"}
{"aliases": [], "authors": ["Yeji Park", "Jaeyun Shim", "Taesik Gong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b792cc0355a0947c61014c471c482aef173f761f7396a939c14816ac644abe66", "submitted_at": "2026-10-06T08:38:30Z", "updated_at": "2026-10-06T08:38:30Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.07972v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.07972", "identifiers": {"arxiv": "2610.07972"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.07972"}], "name": "Can Agents Work for Everyone? Cross-User Reliability for Mobile GUI Agents in Personalized User Interfaces", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces PAIR for controlled evaluation of mobile GUI agents across user-conditioned application states and RePAIR for training from cross-user subgoal outcomes. The study examines failures when personalized interfaces change which items an agent must identify.", "topics": ["evaluation", "robustness", "mobile-navigation", "reinforcement-learning"], "updated_at": "2026-10-06", "url": "https://arxiv.org/abs/2610.07972"}
{"aliases": [], "authors": ["Sarim Hashmi", "Mukul Ranjan", "Kshitij Mishra", "Mikhail Kuznetsov", "Praneeth Vepakomma", "Nils Lukas"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7113025e0dfd784d6994108c24681000625b13598e33ad0e4324e90fd2832c19", "submitted_at": "2026-10-06T17:56:43Z", "updated_at": "2026-10-06T17:56:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.08773v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.08773", "identifiers": {"arxiv": "2610.08773"}, "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/2610.08773"}], "name": "AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents AdvSim2Real, which jointly adapts web-task curricula, prompt-injection adversaries, and agent policies inside a frozen web world model. It evaluates robustness against an unseen adversary and transfer from simulated interaction to a real browser.", "topics": ["safety-security", "world-models", "reinforcement-learning"], "updated_at": "2026-10-06", "url": "https://arxiv.org/abs/2610.08773"}
{"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": ["Jiahao Lu", "Yize Shen", "Chuanyuan Tan", "Mianqiu Huang", "Shicheng Fang", "Xinchi Chen", "Yining Zheng", "Yuxin Wang", "Xipeng Qiu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73073d2d533307cd71b05afef2441a8a7b8f198f804f11f83536d4297db2491a", "submitted_at": "2026-10-12T12:52:21Z", "updated_at": "2026-10-12T12:52:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.16720v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.16720", "identifiers": {"arxiv": "2610.16720"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.16720"}], "name": "Learn from Move: the Next Step for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces LatentGUIWorld to test learning interface semantics and dynamics through interaction, and studies finer movement actions with intermediate visual feedback and exploration demonstrations.", "topics": ["evaluation", "grounding", "reinforcement-learning"], "updated_at": "2026-10-12", "url": "https://arxiv.org/abs/2610.16720"}
{"aliases": [], "authors": ["Zirui Shang", "Yang Liu", "Xin Shu", "Zhi Gao", "Xinxiao Wu", "Lifeng Fan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "46ea5a1a350ccfcc12d22ab3783e05e3c580719962831561b01aeb3fcc6f6c28", "submitted_at": "2026-10-13T04:37:18Z", "updated_at": "2026-10-13T04:37:18Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.17710v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.17710", "identifiers": {"arxiv": "2610.17710"}, "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/2610.17710"}], "name": "Adaptive-Group Process Rewards for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Groups GUI trajectories by subgoal and combines action-level and group-level assessments into intermediate rewards for training agents with sparse task outcomes.", "topics": ["reinforcement-learning", "long-horizon", "planning"], "updated_at": "2026-10-13", "url": "https://arxiv.org/abs/2610.17710"}
{"aliases": [], "authors": ["Hang He", "Linting Wang", "Guanming Liu", "Bo Mao", "Jinxiu Liu", "Yiran Suo", "Maosen Zhou", "Zhiheng Xi", "Guanting Dong", "Zhou Shao", "Rui Zheng", "Jie Zhou", "Chengcheng Wan", "Tao Gui", "Qi Zhang", "Xipeng Qiu", "Xuanjing Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6add363f25f87ea1814c7b79ebf6b51f7431b3745a7d2d910960b78e2d97d13f", "submitted_at": "2026-10-13T09:52:02Z", "updated_at": "2026-10-13T09:52:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.18067v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.18067v1"], "reason": "Document declaration only; code, data and weights permissions are not inferred."}}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:39Z", "role": "primary", "sha256": "c52e4e283e0dae88ef8b3ce67b7424b1d4b48e8f41015d872f4da69f98782eaa", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.39924%2C2610.00346%2C2610.00583%2C2610.01939%2C2610.05099%2C2610.09127%2C2610.09484%2C2610.09633%2C2610.11251%2C2610.11922%2C2610.13098%2C2610.13166%2C2610.13361%2C2610.13623%2C2610.16117%2C2610.17953%2C2610.18067%2C2610.19366%2C2610.20424&max_results=19"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T19:02:20Z", "role": "primary", "sha256": "f1fbdd2ecb2a54b7723372118e2b0ac818420f3a3dd280c2486c8e064cf486d6", "supports": ["name", "kind", "summary", "scope", "environments", "topics", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.18067v1"}], "id": "arxiv:2610.18067", "identifiers": {"arxiv": "2610.18067"}, "kind": "paper", "last_verified_at": "2026-10-20T19:02:20Z", "licenses": {"code": null, "data": null, "document": {"basis": "The exact-v1 arXiv HTML infobox explicitly declares CC BY 4.0 for this paper.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T19:02:20Z", "url": "https://arxiv.org/html/2610.18067v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.18067"}], "name": "WorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment Evolution", "notes": "App WorkEnvs expose UI observations and interaction interfaces alongside file and tool workflows; no specific operating system or pixel-control success is established. Verifier shaping depends on deterministic-check and dependency coverage. External reference scores use differing harnesses or subsets; GDPval Elo is not comparable to percentage metrics. A successful-trace analysis subset is separate from online RL outcomes. No implementation release or independent reproduction was verified.", "published_at": "2026-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Synthesizes resettable workspaces with files, application observations and skills; repairs environment incompatibilities and trains long-task agents using verifier-linked turn credit.", "topics": ["data-generation", "long-horizon", "evaluation", "reward-models", "reinforcement-learning", "workflow"], "updated_at": "2026-10-13", "url": "https://arxiv.org/abs/2610.18067"}
{"aliases": [], "authors": ["Qingsong Xie", "Jiahao Tian", "Jiacheng Tang", "Zhifeng Li", "Madalina Ciobanu", "Navan Preet Singh", "Qingqing Mao", "Ritankar Das"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "08d90bcec690b238be96207602ca694759c8672fecbed85e30372dd314e4006e", "submitted_at": "2026-10-13T19:37:49Z", "updated_at": "2026-10-13T19:37:49Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.19100v1"}}, "environments": ["desktop", "web", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.19100", "identifiers": {"arxiv": "2610.19100"}, "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/2610.19100"}], "name": "GroundGUI: GUI Grounding by High-Quality Data and Distillation-to-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": "2026-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces ScreenRef screenshot grounding data and a training sequence from supervised learning through teacher-guided distillation to reinforcement learning for GroundGUI-8B.", "topics": ["grounding", "data-generation", "reinforcement-learning"], "updated_at": "2026-10-13", "url": "https://arxiv.org/abs/2610.19100"}
{"aliases": [], "authors": ["Mingyu Ouyang", "Yiran Wang", "Siyuan Hu", "Xingyilang Yin", "Huiqiang Sun", "Guangzhi Wang", "Junfu Pu", "Cheng Cheng", "Honghao Chen", "Ying Shan", "Mike Zheng Shou", "Hwee Tou Ng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "67b539468bd2bbf0d3b49a5a7eed4791a8ddca4b56ad9515d23a29204d1f069f", "submitted_at": "2026-10-14T17:57:12Z", "updated_at": "2026-10-14T17:57:12Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.20584v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}], "id": "arxiv:2610.20584", "identifiers": {"arxiv": "2610.20584"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.20584"}], "name": "Player-1: Learning to Play from Synthetic Experience", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Extends computer-use policies to game control through synthetic training environments and richer keyboard and pointer actions. Its evaluation concerns games rather than ordinary productivity workflows.", "topics": ["data-generation", "reinforcement-learning", "evaluation"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.20584"}
{"aliases": [], "authors": ["Panrong Tong", "Quyu Kong", "Pengxiang Li", "Tianyu Xia", "Hanzhang Zhou", "Lei Wang", "Chenglin Cai", "Long Li", "Long Chen", "Jianan Zhang", "Zhengwen Zeng", "Xu Zhang", "Gongjie Zhang", "Yue Wang", "Steven Hoi"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a5f4eda178718aac211eddc7281d572dc1d189dfd0cd130d4eba9f30f5a05ca2", "submitted_at": "2026-10-14T21:06:01Z", "updated_at": "2026-10-14T21:06:01Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.20939v1"}}, "environments": ["cross-platform", "desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.20939", "identifiers": {"arxiv": "2610.20939"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.20939"}], "name": "MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "Unifies GUI evaluation, trajectory collection, environment synthesis, and online reinforcement learning across mobile, computer, and web tasks.", "topics": ["sandbox", "evaluation", "reinforcement-learning"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.20939"}
{"aliases": [], "authors": ["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": ["Tengda Guo", "Jiawang Hu", "Yifei Tao", "Yin Wang", "Haotian Hu", "Fanyi Wang", "Liujuan Tang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "025c7018c86ae4de79a659242070b2f91c5913333e9ecd4bd7fd357d13d27a34", "submitted_at": "2026-10-14T14:39:41Z", "updated_at": "2026-10-14T14:39:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22408v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T12:58:20.217517Z", "role": "primary", "sha256": "f275c47a8b6d349a70e39a9e55eea5a09d99d5f35cc896100f3bcfd5b0923e0b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22computer-use+agents%22+OR+abs%3A%22GUI+agents%22+OR+abs%3A%22browser+agents%22+OR+abs%3A%22mobile+GUI%22%29+AND+submittedDate%3A%5B202609200000+TO+202610201255%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:54.814500Z", "role": "primary", "sha256": "0389d96a8f6c4975129f5c526e8df9ca4690f31364a2f4e66aaea4a75f0eabbc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.22408v1"}], "id": "arxiv:2610.22408", "identifiers": {"arxiv": "2610.22408"}, "kind": "paper", "last_verified_at": "2026-10-20T13:12:54.814500Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.22408"}], "name": "ECHOES: Seeing Beyond the Current Image via Recursive History Compression and Privileged On-Policy Distillation", "notes": "Privileged screenshot history is used during training, while deployment retains self-generated textual state. GUI results use action/step metrics and model-specific history settings; online Sokoban transfer is separate. GRPO rewards action correctness, valid output format and compression length. Diagnostic memory replay is distinct from the validation gap used to schedule training. No independent reproduction.", "published_at": "2026-10-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "ECHOES learns recursive textual task memory through privileged screenshot-history distillation and GRPO, then predicts GUI actions from the current screenshot and memory.", "topics": ["memory", "interaction-history", "state-tracking", "long-horizon", "imitation-learning", "reinforcement-learning", "efficiency", "evaluation"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.22408"}
{"aliases": [], "authors": ["Chulhoon Yoon", "Kyungri Park", "Jinyeong Yun", "Daeyun Jeong", "Sukmin Yun", "Woohwan Jung"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b9585696d256bacc710dce57bfd7cda6a8037ce5725a30f959028352bc07e97c", "submitted_at": "2026-10-16T06:44:56Z", "updated_at": "2026-10-16T06:44:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.23012v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-19T04:42:05Z", "role": "primary", "sha256": "ffbb3f28a3a0fc8888f30ea8b2cdaf441239685aaacc2d80f311581f4b5a1d5c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610050000+TO+202610190000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T10:34:04.380045+00:00", "role": "primary", "sha256": "75c025a862a046b8bcbdd84b800a37132edcc1ec271d40c57c18ec984652a0b0", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.23012v1"}], "id": "arxiv:2610.23012", "identifiers": {"arxiv": "2610.23012"}, "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/2610.23012"}], "name": "Adaptation of Computer Use Agents to New Software with User Manuals", "notes": "Reviewed the complete retained abstract and selected exact-v1 methods, evaluation and validation sections. Execution plans stay fixed per episode; policy training is separate and application-specific. Screenshot-based training/Slack judgments can miss task conditions, and OSWorld tasks may overlap backbone pretraining. No independent reproduction, safe-deployment guarantee or artifact-release verification.", "published_at": "2026-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "MAGE retrieves software-manual sections into fixed execution plans and generates setup/main task-plan pairs whose GUI rollouts support application-specific policy adaptation with round-based GRPO.", "topics": ["planning", "data-generation", "reinforcement-learning"], "updated_at": "2026-10-16", "url": "https://arxiv.org/abs/2610.23012"}
{"aliases": [], "authors": ["Zhanhui Zhou", "Weichen Zhang", "Haoran Liu", "Lingjie Chen", "Tianneng Shi", "Kevin Lin", "Zhengyuan Yang", "Lijuan Wang", "Dawn Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0a297401f7ad4dcd2f485ade6bab295e365280bb803dcecc111aebf0693dd203", "submitted_at": "2026-10-19T09:30:54Z", "updated_at": "2026-10-19T09:30:54Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26071v1"}}, "environments": ["desktop", "linux", "windows", "web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:28:55.967592+00:00", "role": "primary", "sha256": "d4231fa9677e916ca8cc27493dfbbe9fb34c7430207150e6b9be10c2bad43bfb", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26071v1"}], "id": "arxiv:2610.26071", "identifiers": {"arxiv": "2610.26071"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26071"}], "name": "CUA-Lite: An Open Platform for Computer Use Agents", "notes": "Reviewed the complete official abstract and selected v1 methods, training and limitations sections. WebVoyager uses offline mirrors; VM-free sandboxes are Linux-only, and reported desktop/mobile gains do not establish unseen-app transfer. No results or released artifacts were independently reproduced.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "CUA-Lite unifies environment interfaces, supervised data formats and per-model harnesses for evaluation, reinforcement learning and supervised fine-tuning across desktop, browser and mobile tasks.", "topics": ["tool-integration", "data-generation", "evaluation", "imitation-learning", "reinforcement-learning", "sandbox"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26071"}
{"aliases": [], "authors": ["Zhizhou Wang", "Yaxi Lu", "Zhong Zhang", "Zikun Hu", "Xin Cong", "Yankai Lin", "Fangming Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "08ef7c9cdda652f60d3f2d4ab89e7218ad5a2b85f5c76f0277c289db06fd2796", "submitted_at": "2026-10-19T14:17:33Z", "updated_at": "2026-10-19T14:17:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26469v1"}}, "environments": ["desktop", "linux", "web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:28:59.073204+00:00", "role": "primary", "sha256": "f13f0225c721d5aa430bb8aea1d1e7a7da5973760c4c3b82f26ba648aade82ba", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26469v1"}], "id": "arxiv:2610.26469", "identifiers": {"arxiv": "2610.26469"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26469"}], "name": "VeriCUA: Learning Proactive Verification for Reliable GUI Agent Rewards", "notes": "Reviewed the complete abstract and selected v1 method/evaluation sections. The Judge is trained on OSWorld and frozen for MobileWorld/WebArena transfer; downstream Actor RL is separate. Evaluation uses screened terminal outcomes, and reported cost counts Probe tokens. No independent reproduction.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "VeriCUA trains a Judge to select verification checks and stopping decisions while a frozen Probe inspects the post-task GUI. Single-attempt reinforcement learning balances verdict correctness and Probe use.", "topics": ["reward-models", "reinforcement-learning", "evaluation", "efficiency"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26469"}
{"aliases": [], "authors": ["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": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "3e5f24b3175afeba8635dacc7b66eba1ce89fea49996a20281ff2df5a41f1e24", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "ae4f966004dd672fd5b994add5a5cba1fee9233d4d81b224e94c339728fafd87", "supports": ["identity"], "url": "https://api.github.com/repos/amazon-science/PAE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "b1e24a3f38fae67f71033a3f920164d282d3f8bb072f438349dbe9eb23730a7a", "supports": ["release_status"], "url": "https://api.github.com/repos/amazon-science/PAE/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:15Z", "role": "primary", "sha256": "db2e35513dbadcdc67f5819a3bfee2777786538dd3531620cd5fbd4b6ed6e538", "supports": ["licenses"], "url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/LICENSE.txt"}], "id": "github:amazon-science/pae", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:02:06Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/amazon-science/PAE"}], "name": "Proposer-Agent-Evaluator (PAE)", "notes": "Source code and browser-environment modules are present. The repository LICENSE is CC-BY-NC-4.0, while its README describes that license as applying to a dataset, leaving code scope ambiguous. The separately linked dataset is gated and was not accessed; model and dataset terms must be checked separately.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/README.md", "target_id": "arxiv:2412.13194", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A browser-agent training system that proposes tasks, collects visual interaction trajectories, and uses automated evaluation to drive policy improvement.", "topics": ["reinforcement-learning", "web-navigation", "data-generation"], "updated_at": null, "url": "https://github.com/amazon-science/PAE"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/LICENSE"], "reason": "The inspected root LICENSE is complete MIT text for repository code. The local data README says same as parent repository, but the linked Hub dataset/model cards and artifacts were not inspected; their licenses are not inferred."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:51:34.292633Z", "role": "primary", "sha256": "7cda5207d8882bc8dedaefb93d2ccf54b310354917283b8f4890fffdc5764958", "supports": ["name", "kind", "summary", "topics", "scope", "notes", "environments", "relations"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/README.md"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:04:53.470789+00:00", "role": "primary", "sha256": "3d76d934e98641746085772b98ee17059749f980a616435498a7ad483bd4cec1", "supports": ["id", "url"], "url": "https://api.github.com/search/repositories?q=topic%3Agui-grounding+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-20&per_page=100&page=1&sort=updated&order=asc"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-20T11:50:58.795155Z", "role": "primary", "sha256": "08b89c5e6143afa4c0cdde95f8ec016d9033ee004f7631fadd4457144c2b87c9", "supports": ["id", "url", "notes"], "url": "https://github.com/ChenghengLi/WinDOM"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:52:41.278660Z", "role": "primary", "sha256": "09488140d35408cc11bb9c12b1bb435b21814cf6525937a8309475c9326c4004", "supports": ["environments", "notes", "licenses.data"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/data/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:52:45.091369Z", "role": "primary", "sha256": "0ee33b1e416ea4c5d8834688e6f42fbd15ef493f1b82398d3bb985838cbdffb9", "supports": ["summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/src/windom/training/sfd_4b.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:52:48.511963Z", "role": "primary", "sha256": "c481b2f91f2d82bf801632da6ab6cce01bc42c780104d743156c272942d2cef3", "supports": ["summary", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/src/windom/evaluation/screenspot.py"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T11:51:44.940106Z", "role": "primary", "sha256": "d5a37ff86df9b231440ab1bcf3ecf77304a0a53e772182cdb622a3ab5d043447", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/LICENSE"}], "id": "github:chenghengli/windom", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-20T11:51:34.292633Z", "licenses": {"code": {"basis": "Complete MIT license text in the repository root at the inspected commit; repository code only.", "declared": "MIT", "retrieved_at": "2026-10-20T11:51:44.940106Z", "url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ChenghengLi/WinDOM"}], "name": "WinDOM code", "notes": "The Windows-style environment is win11React in Playwright, not native Windows control. The repository documents its predecessor trajectory pipeline but explicitly omits the final corpus harvester. The inspected ScreenSpot evaluator defaults to a privileged ground-truth-center demonstration; --no-demo and --official-prompt select other modes, so evaluation results require their prompt and coordinate settings. Linked datasets, weights and reported scores were not independently verified; MIT applies to inspected repository code.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ChenghengLi/WinDOM/ab52acffe2003fd0c98943a542b4114ef929de0d/README.md", "target_id": "arxiv:2606.25964", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "GUI-grounding training and evaluation code with same-family distillation, GRPO recipes and point-in-box scoring, using a Windows-style web environment.", "topics": ["grounding", "imitation-learning", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://github.com/ChenghengLi/WinDOM"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:09:28Z", "role": "primary", "sha256": "33c55f7d1df7f812c9c9b9c4d97a4de7408b1b1916e58eed266b1f88389b50cd", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/DigiRL-agent/digiq/HEAD/README.md"}], "id": "github:digirl-agent/digiq", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T11:09:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/DigiRL-agent/digiq"}], "name": "Digi-Q", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/DigiRL-agent/digiq/HEAD/README.md", "target_id": "arxiv:2502.15760", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Offline device-control training code that learns action values and uses them to select promising actions, with linked trajectories and checkpoints.", "topics": ["reinforcement-learning", "efficiency"], "updated_at": null, "url": "https://github.com/DigiRL-agent/digiq"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "464e26d88420cd2694abe4c014bfd088417622c3a980aadf19b2bf01988d5b65", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/DigiRL-agent/digirl/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/DigiRL-agent/digirl/HEAD/LICENSE"}], "id": "github:digirl-agent/digirl", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/DigiRL-agent/digirl/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/DigiRL-agent/digirl"}], "name": "DigiRL", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/DigiRL-agent/digirl/HEAD/README.md", "target_id": "arxiv:2406.11896", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Reinforcement learning for autonomous mobile device control.", "topics": ["reinforcement-learning"], "updated_at": null, "url": "https://github.com/DigiRL-agent/digirl"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "f501f8de6aac53758f60882a8c595a5ac27145f327c73b0cfa7921b45ccae75a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/DistRL-lab/distrl-open/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/DistRL-lab/distrl-open/HEAD/LICENSE"}], "id": "github:distrl-lab/distrl-open", "identifiers": {}, "kind": "framework", "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/DistRL-lab/distrl-open/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/DistRL-lab/distrl-open"}], "name": "DistRL", "notes": "The README identifies model weights and dataset release as pending; this entry covers the code framework.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/DistRL-lab/distrl-open/HEAD/README.md", "target_id": "arxiv:2410.14803", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Distributed online reinforcement-learning framework for Android agents, with environment setup and trajectory collection tools.", "topics": ["reinforcement-learning", "efficiency"], "updated_at": null, "url": "https://github.com/DistRL-lab/distrl-open"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["configurable coordinate actions", "element-reference actions", "keyboard input"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md", "population": "README collection size", "qualifier": "more-than", "unit": "web environments", "value": 100}], "evaluation": "", "evidence_urls": ["https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md", "https://miniwob.farama.org/content/observation_space/", "https://miniwob.farama.org/content/action_space/"], "focus": ["small web interaction tasks"], "notes": "The README gives a lower bound, not an exact active-environment inventory. Action presets expose different capabilities, so comparisons must identify the preset. Environment-specific reward details are not normalized here. The repository is in maintenance mode.", "observations": ["task instructions", "screenshots", "visible DOM-element attributes"], "setting": "interactive"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:54Z", "role": "primary", "sha256": "90499776fa1b23e08ac85190ad3be709caa9172ff4f2a0a106b0fa350568e86f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:40:54Z", "role": "primary", "sha256": "90499776fa1b23e08ac85190ad3be709caa9172ff4f2a0a106b0fa350568e86f", "supports": ["details.benchmark_protocol", "relations", "notes"], "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:23Z", "role": "primary", "sha256": "511039e0e3c26b117e7e0a001746c26ec6d94a8ab8c467f0be0bc1a0b01e870a", "supports": ["details.benchmark_protocol"], "url": "https://miniwob.farama.org/content/observation_space/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:23Z", "role": "primary", "sha256": "ae7bf6a3b98b198b985d7ca7b41cb2bdab3cf48e6445467ad9a8d4a2c8df240f", "supports": ["details.benchmark_protocol"], "url": "https://miniwob.farama.org/content/action_space/"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:07Z", "role": "primary", "sha256": "fd6529fb8f648d4130fd969e3d7fc740c9992a0bd94a1998f52555a873311511", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/LICENSE"}], "id": "github:farama-foundation/miniwob-plusplus", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:40:54Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:07Z", "url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Farama-Foundation/miniwob-plusplus"}], "name": "MiniWoB++", "notes": "The maintainer README identifies MiniWoB++ as the extension introduced in the 2018 workflow-guided exploration paper. It separately links World of Bits as the source for the original MiniWoB predecessor. The current Farama interface, environment inventory, and maintenance status belong to the retained repository snapshot.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Farama-Foundation/miniwob-plusplus/HEAD/README.md", "target_id": "arxiv:1802.08802", "type": "introduced_by"}], "review_status": "curated", "scope": "core", "summary": "Over 100 web-interaction environments extending OpenAI's MiniWoB, controlled from Python through a Gymnasium API that drives the browser with Selenium.", "topics": ["evaluation", "reinforcement-learning"], "updated_at": null, "url": "https://github.com/Farama-Foundation/miniwob-plusplus"}
{"aliases": ["github:deepmind/android_env", "deepmind/android_env", "https://github.com/deepmind/android_env"], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "b8dd2ea004647008265f4fb5c5a122569992311b6758f413080984266e53c25a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/google-deepmind/android_env/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/google-deepmind/android_env/HEAD/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T15:37:19Z", "role": "primary", "sha256": "de82c4101e81a6980789904be59c3733ab13f2c1f0e769e5fcb1ab35f847cb3e", "supports": ["aliases"], "url": "https://api.github.com/repos/deepmind/android_env"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T15:37:20Z", "role": "primary", "sha256": "de82c4101e81a6980789904be59c3733ab13f2c1f0e769e5fcb1ab35f847cb3e", "supports": ["aliases"], "url": "https://api.github.com/repos/google-deepmind/android_env"}], "id": "github:google-deepmind/android_env", "identifiers": {}, "kind": "infrastructure", "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/google-deepmind/android_env/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/google-deepmind/android_env"}], "name": "AndroidEnv", "notes": "Formerly published as deepmind/android_env. That address now redirects to this repository, and both paths serve the same reviewed README, so the project is counted once.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/google-deepmind/android_env/HEAD/README.md", "target_id": "arxiv:2105.13231", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Android device-control platform with screenshots, accessibility telemetry, input actions, and an optional reinforcement-learning task interface.", "topics": ["sandbox", "reinforcement-learning", "automation"], "updated_at": null, "url": "https://github.com/google-deepmind/android_env"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "3ef331612819a006d4279ac59203971a22a56fb6d9a0047f85a82668d09c327b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/hud-evals/hud-python/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c88cc185e5547fce850b56f3b4072bed161658dc7277707e509437b9d760d470", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/hud-evals/hud-python/HEAD/LICENSE"}], "id": "github:hud-evals/hud-python", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/hud-evals/hud-python/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/hud-evals/hud-python"}], "name": "HUD", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "SDK for RL environments and evals spanning coding, browser, computer-use and robotics agents, whose environments expose SSH, MCP, CDP or VNC capabilities behind a minimal start-and-grade protocol.", "topics": ["evaluation", "reinforcement-learning"], "updated_at": null, "url": "https://github.com/hud-evals/hud-python"}
{"aliases": ["github:dvlab-research/arpo", "dvlab-research/ARPO", "https://github.com/dvlab-research/ARPO"], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/JIA-Lab-research/ARPO/HEAD/LICENSE"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate."}}, "repository_tree_sha": "7489ae309e90d8b3b9e823f495b63e5116434060", "repository_url": "https://github.com/JIA-Lab-research/ARPO"}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:12Z", "role": "primary", "sha256": "839b2de0e0c614fe8d2df4782d0f6a6e7d33f5c60d8020c9d4a01c61d533f39e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/JIA-Lab-research/ARPO/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:40:39Z", "role": "primary", "sha256": "bb64c7fc1b5d88fc070e919516a51eefac4acb89e86773464cfa0b3de18f1898", "supports": ["identity"], "url": "https://api.github.com/repos/JIA-Lab-research/ARPO"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:25:23Z", "role": "primary", "sha256": "b3ff7f4e350223827cae0fffd986ee672e0b4466349c262ed2dc9e63a8480aeb", "supports": ["identity", "release_status", "details.repository_tree_sha"], "url": "https://api.github.com/repos/JIA-Lab-research/ARPO/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T15:37:00Z", "role": "primary", "sha256": "bb64c7fc1b5d88fc070e919516a51eefac4acb89e86773464cfa0b3de18f1898", "supports": ["identity"], "url": "https://api.github.com/repos/dvlab-research/ARPO"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:19Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/JIA-Lab-research/ARPO/HEAD/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-17T15:37:00Z", "role": "primary", "sha256": "bb64c7fc1b5d88fc070e919516a51eefac4acb89e86773464cfa0b3de18f1898", "supports": ["aliases"], "url": "https://api.github.com/repos/dvlab-research/ARPO"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-18T04:40:39Z", "role": "primary", "sha256": "bb64c7fc1b5d88fc070e919516a51eefac4acb89e86773464cfa0b3de18f1898", "supports": ["aliases"], "url": "https://api.github.com/repos/JIA-Lab-research/ARPO"}], "id": "github:jia-lab-research/arpo", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-18T04:27:12Z", "licenses": {"code": {"basis": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate.", "declared": "Apache-2.0", "retrieved_at": "2026-10-18T04:35:19Z", "url": "https://raw.githubusercontent.com/JIA-Lab-research/ARPO/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/JIA-Lab-research/ARPO"}], "name": "ARPO training framework", "notes": "Official GitHub API resolves the former dvlab-research/ARPO URL to JIA-Lab-research/ARPO with the same numeric repository ID. Training/evaluation scripts are released; OSWorld environments and the separate model checkpoint are not bundled into this code record. Reported benchmark scores were not reproduced.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/JIA-Lab-research/ARPO/HEAD/README.md", "target_id": "arxiv:2505.16282", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Replay-based policy-optimization framework for GUI agents, with distributed OSWorld rollouts and screenshot/action histories.", "topics": ["reinforcement-learning", "long-horizon", "automation"], "updated_at": null, "url": "https://github.com/JIA-Lab-research/ARPO"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:00Z", "role": "primary", "sha256": "eb2a2767741b53e5a6659ef4b0cc325ac18ffdc4222deef1d0e7671089fcaa39", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/KDEGroup/UI-AGILE/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:01Z", "role": "primary", "sha256": "a145c4ec8fd43f59ebe97708b6ee4cef8112959036957eaa4739b8da9a297e8a", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/KDEGroup/UI-AGILE/HEAD/LICENSE"}], "id": "github:kdegroup/ui-agile", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:36:00Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:42:01Z", "url": "https://raw.githubusercontent.com/KDEGroup/UI-AGILE/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/KDEGroup/UI-AGILE"}], "name": "UI-AGILE", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/KDEGroup/UI-AGILE/HEAD/README.md", "target_id": "arxiv:2507.22025", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "GUI agent trained by RL with a continuous reward, a simple-thinking reward and cropping-based resampling on about 9k samples, plus a plug-in decomposed grounding-and-selection inference method.", "topics": ["grounding", "reinforcement-learning"], "updated_at": null, "url": "https://github.com/KDEGroup/UI-AGILE"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/kwai/MobileForge/main/LICENSE", "https://raw.githubusercontent.com/kwai/MobileForge/main/README.md"], "reason": "The complete root LICENSE declares Apache-2.0 for repository code. Linked data and model weights, and third-party dependencies, retain separate licensing requirements."}}, "repository_id": 1278021888}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:04.230792+00:00", "role": "primary", "sha256": "f78f58a6e3b8dd2ae3e8c49a15473782ae258ecd70719f7eeddfa43a10441adf", "supports": ["name", "kind", "summary", "topics", "scope", "relations", "notes", "environments", "licenses.code"], "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:29.533431+00:00", "role": "primary", "sha256": "58187fe25d8398ef2ae31de02debd4bc3b932ff1925db86a87f5de65b016aed1", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/training/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:07.420029+00:00", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code", "notes"], "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-19T19:10:26.954787+00:00", "role": "primary", "sha256": "0ae4477b885b0915e43874399c2a6b91f7d07954697370204550df931f7db82c", "supports": ["details.repository_id"], "url": "https://api.github.com/search/repositories?q=topic%3Agui-agent+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-18&per_page=100&page=1&sort=updated&order=asc"}], "id": "github:kwai/mobileforge", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-20T03:55:04.230792+00:00", "licenses": {"code": {"basis": "Complete repository root Apache License 2.0 text; code declaration only, not linked datasets or weights.", "declared": "Apache-2.0", "retrieved_at": "2026-10-20T03:55:07.420029+00:00", "url": "https://raw.githubusercontent.com/kwai/MobileForge/main/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/kwai/MobileForge"}], "name": "MobileForge", "notes": "The root README describes MobileGym exploration/feedback and HiFPO optimization. The training README limits its public training path to the Qwen3-VL-8B experiment. Linked model and dataset collections, executable source completeness and benchmark results were not independently verified; the repository code license is not a model or dataset license.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/kwai/MobileForge/main/README.md", "target_id": "arxiv:2606.19930", "type": "introduced_by"}], "review_status": "curated", "scope": "core", "summary": "Android GUI-agent adaptation framework combining app exploration, generated curricula and rollout feedback with hint-conditioned step-level GRPO; the documented public training path targets Qwen3-VL-8B.", "topics": ["mobile-navigation", "data-generation", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://github.com/kwai/MobileForge"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:00Z", "role": "primary", "sha256": "f55ac6a0c50d298d46c5f7083a0879959fd7cd3c1d91496af295ebf2afcc13a3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/lll6gg/UI-R1/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:01Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/lll6gg/UI-R1/HEAD/LICENSE"}], "id": "github:lll6gg/ui-r1", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:36:00Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:01Z", "url": "https://raw.githubusercontent.com/lll6gg/UI-R1/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/lll6gg/UI-R1"}], "name": "UI-R1", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/lll6gg/UI-R1/HEAD/README.md", "target_id": "arxiv:2503.21620", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Rule-based reinforcement learning for GUI action prediction, releasing UI-R1-3B, trained from Qwen2.5-VL-3B, the efficiency-oriented UI-R1-E-3B, and training data.", "topics": ["reinforcement-learning", "grounding"], "updated_at": null, "url": "https://github.com/lll6gg/UI-R1"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "b00a072dfde61f774db6acf94fcde9bd32337b1b4f16a178998378d842bbbe4d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/meituan/EvoCUA/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:00Z", "role": "primary", "sha256": "a8d1d88a0f9cce3cdbff73b03a80dc5e5d69e3ba2443946c268815a550c9e0ec", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/meituan/EvoCUA/HEAD/LICENSE"}], "id": "github:meituan/evocua", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:35:59Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:00Z", "url": "https://raw.githubusercontent.com/meituan/EvoCUA/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/meituan/EvoCUA"}], "name": "EvoCUA", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/meituan/EvoCUA/HEAD/README.md", "target_id": "arxiv:2601.15876", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Computer-use model and training resources from the EvoCUA project.", "topics": ["reinforcement-learning", "planning"], "updated_at": null, "url": "https://github.com/meituan/EvoCUA"}
{"aliases": [], "authors": [], "details": {"availability": "Public training and evaluation source; trained checkpoints are generated by the documented workflow.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/microsoft/GUI-Agent-RL/HEAD/LICENSE"], "reason": "Repository LICENSE grants MIT for code; linked training imagery and pretrained model licenses remain separate."}}, "repository_revision": "6d2381ef549d980452a828471221a2ce8023bf41"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:14Z", "role": "primary", "sha256": "d21a1ae4359296c0f3034ac655888dce43448d10afe0b1909d6b6520678c2526", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/microsoft/GUI-Agent-RL/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:13Z", "role": "primary", "sha256": "791c6e56e62ce5e803a4ba313557bac0a39dfeba27499a3865ad444a860b426a", "supports": ["identity"], "url": "https://api.github.com/repos/microsoft/GUI-Agent-RL"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:06Z", "role": "primary", "sha256": "5b799a5bdae78ff50652c0b464b6cad12d9dda265d4752e5d5f9eab65e9d5c6f", "supports": ["release_status"], "url": "https://api.github.com/repos/microsoft/GUI-Agent-RL/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:06Z", "role": "primary", "sha256": "c2cfccb812fe482101a8f04597dfc5a9991a6b2748266c47ac91b6a5aae15383", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/microsoft/GUI-Agent-RL/HEAD/LICENSE"}], "id": "github:microsoft/gui-agent-rl", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T09:12:14Z", "licenses": {"code": {"basis": "Repository LICENSE grants MIT for code; linked training imagery and pretrained model licenses remain separate.", "declared": "MIT", "retrieved_at": "2026-10-17T09:14:06Z", "url": "https://raw.githubusercontent.com/microsoft/GUI-Agent-RL/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/microsoft/GUI-Agent-RL"}], "name": "VEM GUI-Agent-RL", "notes": "The public tree includes critic-training configurations, policy training and offline/online evaluation scripts. The README obtains base models and AITW imagery separately, then instructs users to train critic and policy checkpoints locally; this entry does not claim that VEM-trained weights are released. MIT applies to repository code only.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/microsoft/GUI-Agent-RL/HEAD/README.md", "target_id": "arxiv:2502.18906", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Training code that learns a value environment model from offline Android interaction data and uses its frozen value estimates to guide GUI policy optimization.", "topics": ["reinforcement-learning", "reward-models", "mobile-navigation"], "updated_at": null, "url": "https://github.com/microsoft/GUI-Agent-RL"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:20Z", "role": "primary", "sha256": "2fb7c0d70ccfa93605d5250a8f6d5182324f0c53d70e1c35ba18cb00f4ada548", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/night-chen/WorkForceAgent-R1/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:19Z", "role": "primary", "sha256": "0ad3ef1d900200e3ecab86cd44f79038000e8748fb285cdd5f32882455d2e36e", "supports": ["identity"], "url": "https://api.github.com/repos/night-chen/WorkForceAgent-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:39Z", "role": "primary", "sha256": "521d065188a6354416c1aafe88a0eb34486f2c2ec011c1cdf8ed008760dd9c55", "supports": ["release_status"], "url": "https://api.github.com/repos/night-chen/WorkForceAgent-R1/git/trees/main?recursive=1"}], "id": "github:night-chen/workforceagent-r1", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:03:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/night-chen/WorkForceAgent-R1"}], "name": "WorkForceAgent-R1", "notes": "The release contains WorkArena preprocessing and training source. The inspected root has no explicit project license; dependencies retain their own terms. Although the README discusses trained 7B/14B models, it does not provide a verified checkpoint release, so this is cataloged only as a framework.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/night-chen/WorkForceAgent-R1/HEAD/README.md", "target_id": "arxiv:2505.22942", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Web-task reward functions, WorkArena data processing and GRPO training configurations for reasoning-oriented language agents.", "topics": ["reinforcement-learning", "web-navigation", "planning"], "updated_at": null, "url": "https://github.com/night-chen/WorkForceAgent-R1"}
{"aliases": [], "authors": [], "details": {"availability": "Public archived client source; remote benchmark services and images are not verified usable.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/openai/universe/HEAD/LICENSE"], "reason": "Official Universe client-source LICENSE; no license or availability claim is transferred to external remote applications or container contents."}}, "maintenance_status": "archived and deprecated", "repository_revision": "cc9ce6ec241821bfb0f3b85dd455bd36e4ee7a8c"}, "environments": ["web", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:17Z", "role": "primary", "sha256": "b1ca938480d4d36c76fa939c5b10cef11c24465cc1a6a8ee857cec4b939f9d61", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/openai/universe/HEAD/README.rst"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:15Z", "role": "primary", "sha256": "94a7a4d383f754b96601bc0486b90e62760b8bfc35abc607077c8eb6e7934b2d", "supports": ["identity", "details.availability", "notes"], "url": "https://api.github.com/repos/openai/universe"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:19:47Z", "role": "primary", "sha256": "1808d7524506d1386dad507de8a8fa0f89080443a970ae032cea4c204699d113", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/openai/universe/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:05Z", "role": "primary", "sha256": "ecd4a4aacc4512a31bbdc3252ffc014d37ac7ef56ed34bc2a5d0f6337df6daa5", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/openai/universe/HEAD/LICENSE"}], "id": "github:openai/universe", "identifiers": {}, "kind": "infrastructure", "last_verified_at": "2026-10-17T11:20:17Z", "licenses": {"code": {"basis": "Official Universe client-source LICENSE; no license or availability claim is transferred to external remote applications or container contents.", "declared": "MIT", "retrieved_at": "2026-10-17T11:23:05Z", "url": "https://raw.githubusercontent.com/openai/universe/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/openai/universe"}], "name": "Universe legacy GUI interaction client", "notes": "GitHub marks this repository archived, and the README declares the library deprecated. It contains the client side of Universe; the README says remote-environment source repositories would be released separately. Original MiniWoB task-server or container availability was not established: alpha.openai.com no longer resolved in this audit and the official Quay tag metadata endpoint returned 401. This entry records the released client infrastructure, not a currently runnable original MiniWoB benchmark. Legacy host requirements are documented; no container or code was executed.", "published_at": null, "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "An archived Gym-compatible client that exposes screen observations and keyboard/mouse actions over VNC for remote application and browser environments, including the historical World of Bits integration.", "topics": ["automation", "reinforcement-learning", "history"], "updated_at": null, "url": "https://github.com/openai/universe"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "033d2de815e11c208862c5c5601b78d58668b79cadd37a750cad9edd4dacabed", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OpenGVLab/ZeroGUI/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OpenGVLab/ZeroGUI/HEAD/LICENSE"}], "id": "github:opengvlab/zerogui", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:04Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/OpenGVLab/ZeroGUI/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OpenGVLab/ZeroGUI"}], "name": "ZeroGUI", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OpenGVLab/ZeroGUI/HEAD/README.md", "target_id": "arxiv:2505.23762", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Online RL framework for GUI agents in which VLMs generate the training tasks and judge success from trajectory screenshots, with OSWorld and AndroidLab checkpoints trained from UI-TARS-7B-DPO.", "topics": ["reinforcement-learning", "data-generation"], "updated_at": null, "url": "https://github.com/OpenGVLab/ZeroGUI"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop", "web", "cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:46:05Z", "role": "primary", "sha256": "a495bb4c86cf54cd5e2276433cfecb401fd133287406b1a7e04c3eaaac6305ac", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/PrentisAI/GroundGUI/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T01:52:51Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/PrentisAI/GroundGUI/HEAD/LICENSE"}], "id": "github:prentisai/groundgui", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T01:46:05Z", "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/PrentisAI/GroundGUI/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/PrentisAI/GroundGUI"}], "name": "GroundGUI training and evaluation code", "notes": "The repository documents separate training-stage environments. Code, model weights, and ScreenRef data have distinct license declarations; no training or benchmark execution was reproduced here.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/PrentisAI/GroundGUI/HEAD/README.md", "target_id": "arxiv:2610.19100", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Code for screenshot grounding through supervised training, teacher-guided distillation, reinforcement learning, and benchmark evaluation.", "topics": ["grounding", "reinforcement-learning", "data-generation"], "updated_at": null, "url": "https://github.com/PrentisAI/GroundGUI"}
{"aliases": [], "authors": [], "details": {"availability": "Public simulator, task/judge and online-RL source, plus official release metadata for the companion data archive.", "benchmark_protocol": {"actions": ["tap", "type", "swipe", "back", "home", "drag", "complete"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "population": "publisher benchmark overview", "qualifier": "exact", "unit": "task templates", "value": 416}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "population": "publisher benchmark overview", "qualifier": "exact", "unit": "simulated apps", "value": 28}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/train.txt", "population": "bench_env/splits/train.txt", "qualifier": "exact", "unit": "task identifiers", "value": 160}, {"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/test.txt", "population": "bench_env/splits/test.txt", "qualifier": "exact", "unit": "task identifiers", "value": 256}], "evaluation": "Deterministic task judges inspect state changes; reported metrics distinguish success, progress, false completion and unexpected side effects.", "evidence_urls": ["https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/train.txt", "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/test.txt"], "focus": ["mobile GUI task completion in simulation", "state verification and parallel training"], "notes": "The split-ID inventory is not a count of sampled trajectories, live Android apps or real user accounts. Extra runtime/data dependencies and third-party content terms remain distinct.", "observations": ["simulated mobile screenshots", "structured simulator state for judges"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE"], "reason": "Core source LICENSE; mobilegym-rl has a separate matching Apache-2.0 license and retains vendored notices."}, "data": {"evidence_urls": ["https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE-DATA"], "reason": "Limited to maintainers’ dataset compilation/curation and authored synthetic content; third-party user content remains subject to platform terms and its authors’ rights."}}, "release_tag": "v0.1.0", "repository_revision": "5fbd3f34be0e9780ae4b1e9d5089963ddbd05d38"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:19Z", "role": "primary", "sha256": "815294dd2aa4bf8264978c3bfedee85a3a2c835c511860c732314a4c282a974a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:18Z", "role": "primary", "sha256": "48df04a2fbf479f956d5c7f22d7006c22f591b8857cab8a1d64c196349045495", "supports": ["identity"], "url": "https://api.github.com/repos/Purewhiter/mobilegym"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:56Z", "role": "primary", "sha256": "68f1fa16c0999079bccc3aa1b6c3c22b72b570a5b971837c5026ea5408853375", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/Purewhiter/mobilegym/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:08Z", "role": "primary", "sha256": "cb1057b0a0a2ef7ab781414c168bb41f0f3da88afc042ce266d6063493561bde", "supports": ["release_status", "details.release_tag"], "url": "https://api.github.com/repos/Purewhiter/mobilegym/releases/tags/v0.1.0"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:09Z", "role": "primary", "sha256": "ad0c86a6cdde2caba8bd1f98b121803817d93718319a863b0b833da2760c23aa", "supports": ["release_status"], "url": "https://api.github.com/repos/Purewhiter/mobilegym/releases/tags/data-v0.1.0"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:59Z", "role": "primary", "sha256": "7618cd7de039c264759ca8ca0b9130c6de63a4536dcf436baedd4cfbf639073e", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:27:50Z", "role": "primary", "sha256": "58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/mobilegym-rl/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:59Z", "role": "primary", "sha256": "e222fa518c5003d425ee9b371e987c95ba10e52b54fea078571254ae6a8eaf23", "supports": ["licenses.data", "notes"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE-DATA"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:27:49Z", "role": "primary", "sha256": "0ecef460fecfd8aa91dcbb2cd4cca17b033f620f85a50d7f0b598ed19a27c0eb", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/train.txt"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:27:48Z", "role": "primary", "sha256": "a3d485015d480f4f1ebbd12dbbf02c25c0aa667d1afcf5f6fa35f0263a82c567", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/bench_env/splits/test.txt"}], "id": "github:purewhiter/mobilegym", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T11:18:19Z", "licenses": {"code": {"basis": "Core source LICENSE; mobilegym-rl has a separate matching Apache-2.0 license and retains vendored notices.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T11:22:59Z", "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE"}, "data": {"basis": "Limited to maintainers’ dataset compilation/curation and authored synthetic content; third-party user content remains subject to platform terms and its authors’ rights.", "declared": "CC-BY-NC-4.0", "retrieved_at": "2026-10-17T11:22:59Z", "url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/LICENSE-DATA"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Purewhiter/mobilegym"}], "name": "MobileGym", "notes": "The inspected code release is v0.1.0, with a separately listed data-v0.1.0 archive. The interaction target is a browser-hosted Android-like simulation; real-device transfer results are separate publisher-reported experiments. Source split manifests contain 160 training IDs and 256 test IDs with no duplicates or overlap. Platform/training code declares Apache-2.0. LICENSE-DATA applies CC-BY-NC-4.0 only to maintainer-created compilation, curation and synthetic content, and explicitly leaves third-party UGC/API terms and author rights separate. No runtime, archive or model was executed/downloaded.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Purewhiter/mobilegym/HEAD/README.md", "target_id": "arxiv:2605.26114", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A browser-hosted mobile-app simulator with programmable state, parameterized tasks and deterministic state-checking judges for parallel GUI-agent evaluation and online reinforcement learning.", "topics": ["evaluation", "mobile-navigation", "reinforcement-learning", "sandbox"], "updated_at": null, "url": "https://github.com/Purewhiter/mobilegym"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:00Z", "role": "primary", "sha256": "5fd1399c88bdf406e7689765ccf0599b3b7e0ddbed082fa92bffdd7b120bd41a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/Reallm-Labs/InfiGUI-R1/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:01Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Reallm-Labs/InfiGUI-R1/HEAD/LICENSE"}], "id": "github:reallm-labs/infigui-r1", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:36:00Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:01Z", "url": "https://raw.githubusercontent.com/Reallm-Labs/InfiGUI-R1/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Reallm-Labs/InfiGUI-R1"}], "name": "InfiGUI-R1", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Reallm-Labs/InfiGUI-R1/HEAD/README.md", "target_id": "arxiv:2504.14239", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "3B GUI agent trained in two stages: distilling spatial reasoning from teacher models, then reinforcement learning that rewards accurate sub-goals and trains error recovery.", "topics": ["reinforcement-learning", "planning"], "updated_at": null, "url": "https://github.com/Reallm-Labs/InfiGUI-R1"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "94c7c78bede11dde727940f56592a04594c1c59438812dff47540bf53c1bed7d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ritzz-ai/GUI-R1/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:42Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ritzz-ai/GUI-R1/HEAD/LICENSE"}], "id": "github:ritzz-ai/gui-r1", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T10:43:11Z", "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/ritzz-ai/GUI-R1/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ritzz-ai/GUI-R1"}], "name": "GUI-R1", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ritzz-ai/GUI-R1/HEAD/README.md", "target_id": "arxiv:2504.10458", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "GUI action-model training and evaluation code with verifiable rewards, plus linked reinforcement-learning data and checkpoints.", "topics": ["reinforcement-learning", "grounding"], "updated_at": null, "url": "https://github.com/ritzz-ai/GUI-R1"}
{"aliases": [], "authors": [], "details": {"availability": "Public historical training and environment-wrapper source; legacy dependencies documented.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/LICENSE"], "reason": "Stanford WGE source license; not a blanket license for all third-party tasks, embeddings or demonstration resources."}}, "repository_revision": "bed526657a62a64a163f80deb4a8627cb63871a3"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:16Z", "role": "primary", "sha256": "50ee5234bf62320d096fbe4c11c7be21f50c9e09dae97754a16a5bbde2f1ab19", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:15Z", "role": "primary", "sha256": "66fcd6577d1867abf84e62fe9edf2e3f223292724cbe6b9d0ef02b582db5c8b2", "supports": ["identity"], "url": "https://api.github.com/repos/stanfordnlp/wge"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:19:50Z", "role": "primary", "sha256": "926aff11e14c529d05d9bc51fe12aadb007818119da55f99c2fa47d0387f466d", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/stanfordnlp/wge/git/trees/master?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:17Z", "role": "primary", "sha256": "249b23ff62e21bd446b3e411b1d7ea7f173c7e19e42ed75e97800ed32201bb3a", "supports": ["details", "notes"], "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/third-party/miniwob-sandbox/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:05Z", "role": "primary", "sha256": "3df3dd1e3149cd6cc1a7b052a237a6653449a7c5f18edf19a3a79294c7d541c1", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/LICENSE"}], "id": "github:stanfordnlp/wge", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T11:18:16Z", "licenses": {"code": {"basis": "Stanford WGE source license; not a blanket license for all third-party tasks, embeddings or demonstration resources.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T11:23:05Z", "url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/stanfordnlp/wge"}], "name": "Workflow-Guided Exploration research code", "notes": "The released repository includes training policies, experiment configuration and a MiniWoB sandbox extended for the 2018 paper. This is the WGE implementation, not another copy of the already cataloged Farama MiniWoB++ benchmark and not the original OpenAI MiniWoB release. Its documented dependency is PyTorch 0.1.12, and current runtime compatibility was not tested. Stanford code declares Apache-2.0; bundled third-party browser tasks and externally downloaded demonstration/embedding data retain their own terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/stanfordnlp/wge/HEAD/README.md", "target_id": "arxiv:1802.08802", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Research code that learns browser-control policies by constraining exploration with workflows induced from demonstrations, with a Selenium-based MiniWoB environment and experiment harness.", "topics": ["reinforcement-learning", "web-navigation", "workflow", "history"], "updated_at": null, "url": "https://github.com/stanfordnlp/wge"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "8b327dcd58c18e48e1e7f5b19ebf5673da4b5ff9320aa960e391844adf3a8a55", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/THUDM/ComputerRL/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "a8d1d88a0f9cce3cdbff73b03a80dc5e5d69e3ba2443946c268815a550c9e0ec", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/THUDM/ComputerRL/HEAD/LICENSE"}], "id": "github:thudm/computerrl", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/THUDM/ComputerRL/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/ComputerRL"}], "name": "ComputerRL", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/ComputerRL/HEAD/README.md", "target_id": "arxiv:2508.14040", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Reinforcement-learning infrastructure for computer-use agents.", "topics": ["reinforcement-learning"], "updated_at": null, "url": "https://github.com/THUDM/ComputerRL"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "752fad8ebd8b8eda03648329a6fc2585e5bcb2b96875b8aa2ce7d9f3fff78153", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/THUDM/MobileRL/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "1bbbee919161337e2456c587c6be809214406482cd9f1292a9eb62428f6961f3", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/THUDM/MobileRL/HEAD/LICENSE"}], "id": "github:thudm/mobilerl", "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/THUDM/MobileRL/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/MobileRL"}], "name": "MobileRL", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/MobileRL/HEAD/README.md", "target_id": "arxiv:2509.18119", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Online agentic RL framework for mobile GUI agents built around the difficulty-adaptive ADAGRPO algorithm, with Dockerized AndroidWorld and AndroidLab evaluation and a MobileRL-9B checkpoint.", "topics": ["reinforcement-learning"], "updated_at": null, "url": "https://github.com/THUDM/MobileRL"}
{"aliases": [], "authors": [], "details": {}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:09:28Z", "role": "primary", "sha256": "d4a68b00acc4becf2a25688520595babea9b694b56541e709f066413b5b80aea", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/THUDM/SCALE-CUA/HEAD/README.md"}], "id": "github:thudm/scale-cua", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T11:09:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/SCALE-CUA"}], "name": "SCALE-CUA (THUDM)", "notes": "A distinct project from OpenGVLab/ScaleCUA, which documents arXiv:2509.15221. Similar names do not establish a shared project identity.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/SCALE-CUA/HEAD/README.md", "target_id": "arxiv:2607.11185", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Computer-use training framework with VeriGen task synthesis, adaptive rollout sampling, and visual-context segmentation.", "topics": ["reinforcement-learning", "data-generation", "efficiency"], "updated_at": null, "url": "https://github.com/THUDM/SCALE-CUA"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "eb217b96c54c3c2ccfd3a27255c6fd740df40c2d6743ae811858339bf7f740b0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "42496cf0cff06a7f9e794f4313a10a670bc48df9909ec6d57763cb902bb229ea", "supports": ["identity"], "url": "https://api.github.com/repos/THUDM/WebRL"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:05Z", "role": "primary", "sha256": "bb1683c6fb5ff658c89f0ebf0f8baf8026bb703840e174c9e46b223c85a16856", "supports": ["release_status"], "url": "https://api.github.com/repos/THUDM/WebRL/git/trees/main?recursive=1"}], "id": "github:thudm/webrl", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:02:06Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/THUDM/WebRL"}], "name": "WebRL", "notes": "Training code and checkpoint links are released. No root code-license declaration was found in the inspected repository metadata; bundled dependency licenses do not establish a license for the whole project. Weights have separate Hub terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md", "target_id": "arxiv:2411.02337", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Training pipeline that evolves web tasks from prior attempts and uses an outcome reward model to refine browser policies.", "topics": ["reinforcement-learning", "web-navigation", "data-generation"], "updated_at": null, "url": "https://github.com/THUDM/WebRL"}
{"aliases": [], "authors": [], "details": {}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:42:12Z", "role": "primary", "sha256": "5f460f9b2a18315d07b8c48d78dcc764892d09d9250e95f5e591b5966fe59328", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/vivo-ai-lab/BlueLM-GUI/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:41Z", "role": "primary", "sha256": "6036aa17e52be8051fc3c8ede37fc4d0b2bc2589518be7623566c75868612178", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/vivo-ai-lab/BlueLM-GUI/HEAD/LICENSE"}], "id": "github:vivo-ai-lab/bluelm-gui", "identifiers": {}, "kind": "model", "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/vivo-ai-lab/BlueLM-GUI/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/vivo-ai-lab/BlueLM-GUI"}], "name": "BlueLM-GUI", "notes": "The repository lists the agent as in-house. Public model weights were not established by this review; linked benchmark assets have their own access terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/vivo-ai-lab/BlueLM-GUI/HEAD/README.md", "target_id": "arxiv:2609.12394", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Mobile GUI agent research release with technical report, benchmark data, and evaluation tooling from vivo AI Lab.", "topics": ["mobile-navigation", "reinforcement-learning"], "updated_at": null, "url": "https://github.com/vivo-ai-lab/BlueLM-GUI"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "813b7580cdae140917db573522b83a7986448025b31064442d34757292a6935c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/WebAgent-R1/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:19Z", "role": "primary", "sha256": "db1f71491c1d9482f403db30ea5599bd76e93d1db02c4ba8134d1a5cae5f0487", "supports": ["identity"], "url": "https://api.github.com/repos/weizhepei/WebAgent-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:39Z", "role": "primary", "sha256": "cc1a74037c4a2c4c094b11f6874ef67f88b24d14293ec09568ac6f378035c0b0", "supports": ["release_status"], "url": "https://api.github.com/repos/weizhepei/WebAgent-R1/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/LICENSE"}, {"method": "license_header", "retrieved_at": "2026-10-17T02:05:22Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/LICENSE"}], "id": "github:weizhepei/webagent-r1", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T02:05:22Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T02:05:22Z", "url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/weizhepei/WebAgent-R1"}], "name": "WebAgent-R1", "notes": "The implementation is inside the WebAgent-R1 subdirectory and includes training, evaluation and WebArena setup components. Root source licensing is Apache-2.0. No downloadable author checkpoint was established by the reviewed README, so this entry represents code.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/weizhepei/WebAgent-R1/HEAD/WebAgent-R1/README.md", "target_id": "arxiv:2505.16421", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Environment setup, multi-turn reinforcement-learning and evaluation tooling for browser agents interacting with WebArena.", "topics": ["reinforcement-learning", "web-navigation", "evaluation"], "updated_at": null, "url": "https://github.com/weizhepei/WebAgent-R1"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "2d64349b514c43fb9d0260247300412d33cadc2a029d8279a00cb2a12f26a25a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/Yuqi-Zhou/GUI-G1/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T10:48:43Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Yuqi-Zhou/GUI-G1/HEAD/LICENSE"}], "id": "github:yuqi-zhou/gui-g1", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T10:43:11Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T10:48:43Z", "url": "https://raw.githubusercontent.com/Yuqi-Zhou/GUI-G1/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Yuqi-Zhou/GUI-G1"}], "name": "GUI-G1", "notes": "The README retains an older release-pending notice alongside a later checkpoint-release announcement. Check the linked model card for current weight availability and terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Yuqi-Zhou/GUI-G1/HEAD/README.md", "target_id": "arxiv:2505.15810", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "GUI-grounding training code that modifies reasoning templates, reward functions, and the GRPO objective, with a linked model release.", "topics": ["grounding", "reinforcement-learning"], "updated_at": null, "url": "https://github.com/Yuqi-Zhou/GUI-G1"}
{"aliases": [], "authors": [], "details": {}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:03Z", "role": "primary", "sha256": "3f0aaefce32210dfba428fcf9e609e04e75d4d1520d24a3a6377611a9b83fa6d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:42:05Z", "role": "primary", "sha256": "af69329b4667040c498cc87dfcc165db1f5f9452fe2ad3024ad7eef63d467666", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/LICENSE"}], "id": "github:zju-real/clawgui", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T06:36:03Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "Apache-2.0", "retrieved_at": "2026-10-16T06:42:05Z", "url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ZJU-REAL/ClawGUI"}], "name": "ClawGUI", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/ZJU-REAL/ClawGUI/HEAD/README.md", "target_id": "arxiv:2604.11784", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Framework for GUI agent training, evaluation, and deployment.", "topics": ["reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://github.com/ZJU-REAL/ClawGUI"}
{"aliases": [], "authors": [], "details": {"availability": "Public region-voting evaluation and policy-optimization source; model/data dependencies external.", "license_components_note": "Apache-2.0 notices in VLM-R1 subdirectories do not establish a root-project license.", "license_review": {"code": {"evidence_urls": ["https://api.github.com/repos/ZJU-REAL/GUI-RCPO/git/trees/main?recursive=1", "https://raw.githubusercontent.com/ZJU-REAL/GUI-RCPO/HEAD/VLM-R1/LICENSE", "https://raw.githubusercontent.com/ZJU-REAL/GUI-RCPO/HEAD/VLM-R1/src/open-r1-multimodal/LICENSE"], "reason": "Only subdirectory Apache-2.0 notices were established; no blanket project-source declaration is inferred."}}, "repository_revision": "af15ed5fe8b89b0fe5043a3e94f2984c7b126a4b"}, "environments": ["cross-platform", "desktop", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:55Z", "role": "primary", "sha256": "41c235305d28f6ba77ba1c5d1d9298fd6b1bd68ae3c9ea97eb34a8d22d0ba1dc", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/zju-real/gui-rcpo/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:30:55Z", "role": "primary", "sha256": "f5823de05f4828a8ab367ef13ec3a8338e020f0876e153ac2e944e4eb0e242c0", "supports": ["identity"], "url": "https://api.github.com/repos/zju-real/gui-rcpo"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:32:06Z", "role": "primary", "sha256": "98ed8f0fa50699b8a3d0e8a71cd008113009a3072b971b8fadf2b92f68877e1e", "supports": ["release_status", "details.repository_revision", "licenses.code"], "url": "https://api.github.com/repos/ZJU-REAL/GUI-RCPO/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["notes", "licenses.code"], "url": "https://raw.githubusercontent.com/ZJU-REAL/GUI-RCPO/HEAD/VLM-R1/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:34:10Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["notes", "licenses.code"], "url": "https://raw.githubusercontent.com/ZJU-REAL/GUI-RCPO/HEAD/VLM-R1/src/open-r1-multimodal/LICENSE"}], "id": "github:zju-real/gui-rcpo", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T12:30:55Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/ZJU-REAL/GUI-RCPO"}], "name": "GUI-RC / GUI-RCPO", "notes": "The tree contains GUI-RC evaluation code and VLM-R1-based training scripts, with example JSONL-format paths. It requires separately supplied models and benchmark images; no released trained checkpoint is established here. Apache-2.0 files exist inside VLM-R1 subdirectories, but no repository-wide license is present for all newly added code, so the code slot remains unknown and the component notices are retained.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/zju-real/gui-rcpo/HEAD/README.md", "target_id": "arxiv:2508.05615", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "GUI-grounding research code for spatial voting across sampled predictions and test-time policy optimization using region-consistency rewards.", "topics": ["grounding", "reinforcement-learning", "efficiency"], "updated_at": null, "url": "https://github.com/ZJU-REAL/GUI-RCPO"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/UI-S1/README.md", "population": "public example subset", "qualifier": "exact", "unit": "training examples", "value": 1000}], "evidence_urls": ["https://huggingface.co/datasets/mPLUG/UI_S1_dataset/raw/main/README.md", "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/UI-S1/README.md"], "formats": ["JSONL", "Parquet"], "modalities": ["images", "text", "actions"], "notes": "The count is explicitly reported by the publisher, not inferred from the filename. Full training data remains outside this release.", "splits": ["training example subset"], "supervision": ["GUI actions"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/mPLUG/UI_S1_dataset/tree/main", "gated": false, "hub_id": "mPLUG/UI_S1_dataset"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "d826463e2aa159f083dc74869bfabb8abb5b8ea4f2cdf4aeb7676172b18aaaa5", "supports": ["name", "kind", "relations", "licenses.data"], "url": "https://huggingface.co/datasets/mPLUG/UI_S1_dataset/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "cdfbcbf23a3b1b4a609e1943e80dcf144ca1224a74ebf2308fde4fcb4326ab3c", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/mPLUG/UI_S1_dataset"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:06:57Z", "role": "primary", "sha256": "f668c890a5a87953a243715d0d47367c23f0c99b524d80f48bbdde8163f3e942", "supports": ["summary", "environments", "details.dataset_profile", "topics", "scope"], "url": "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/UI-S1/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "d826463e2aa159f083dc74869bfabb8abb5b8ea4f2cdf4aeb7676172b18aaaa5", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/mPLUG/UI_S1_dataset/raw/main/README.md"}], "id": "url:16184830c0f4bdb146fa", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T04:07:19Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T04:07:19Z", "url": "https://huggingface.co/datasets/mPLUG/UI_S1_dataset/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/mPLUG/UI_S1_dataset"}], "name": "UI-S1 released training examples", "notes": "The release contains train_1000.jsonl and images.parquet. The official README explicitly describes 1,000 released examples and calls the release only part of the training dataset. Apache-2.0 is declared by the dataset card; this is not the complete training corpus.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/mPLUG/UI_S1_dataset/raw/main/README.md", "target_id": "arxiv:2509.11543", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A public example subset of the UI-S1 mobile training data, packaged as JSONL annotations and Parquet images.", "topics": ["reinforcement-learning", "data-generation", "mobile-navigation"], "updated_at": null, "url": "https://huggingface.co/datasets/mPLUG/UI_S1_dataset"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/mPLUG/UI-S1-7B/tree/main", "gated": false, "hub_id": "mPLUG/UI-S1-7B"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:41Z", "role": "primary", "sha256": "341abd3833131080682b27fe0b79e310d7cf9465602b77b8e399e45200552d61", "supports": ["name", "kind", "relations", "licenses.weights"], "url": "https://huggingface.co/mPLUG/UI-S1-7B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:40Z", "role": "primary", "sha256": "d0c25fa2fad8372ffc8b51b63701c7e5548cca08a612c0081c4c317f9532894d", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/mPLUG/UI-S1-7B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:06:57Z", "role": "primary", "sha256": "f668c890a5a87953a243715d0d47367c23f0c99b524d80f48bbdde8163f3e942", "supports": ["summary", "environments", "topics", "scope"], "url": "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/UI-S1/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T04:03:41Z", "role": "primary", "sha256": "341abd3833131080682b27fe0b79e310d7cf9465602b77b8e399e45200552d61", "supports": ["licenses.weights"], "url": "https://huggingface.co/mPLUG/UI-S1-7B/raw/main/README.md"}], "id": "url:1c4ee99a9a55ae71a7ac", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T04:03:41Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T04:03:41Z", "url": "https://huggingface.co/mPLUG/UI-S1-7B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/mPLUG/UI-S1-7B"}], "name": "UI-S1 7B", "notes": "Four public safetensors shards are present. The card declares Apache-2.0 for the weights. The project publishes only a portion of its training data; no complete training corpus or independent reproduction is inferred.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/mPLUG/UI-S1-7B/raw/main/README.md", "target_id": "arxiv:2509.11543", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A mobile GUI checkpoint trained with semi-online reinforcement learning that uses offline trajectories to improve multi-turn action prediction.", "topics": ["reinforcement-learning", "mobile-navigation", "action-prediction"], "updated_at": null, "url": "https://huggingface.co/mPLUG/UI-S1-7B"}
{"aliases": [], "authors": [], "details": {"artifact_type": "model-checkpoint", "disabled": false, "download_url": "https://huggingface.co/meituan/EvoCUA-8B-20260105/tree/main", "gated": false, "hub_id": "meituan/EvoCUA-8B-20260105", "languages": ["en", "zh"], "modalities": [], "size_category_reported": null}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "3fd1e50729e97746fec57c7d5da6f093d76fb9174e74c3f2c630fed5486fe995", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/api/models/meituan/EvoCUA-8B-20260105"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "3fd1e50729e97746fec57c7d5da6f093d76fb9174e74c3f2c630fed5486fe995", "supports": ["details", "licenses.weights"], "url": "https://huggingface.co/api/models/meituan/EvoCUA-8B-20260105"}], "id": "url:2a2682f86dcb8ebc9a80", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:50:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card", "declared": "apache-2.0", "retrieved_at": "2026-10-16T06:50:49Z", "url": "https://huggingface.co/meituan/EvoCUA-8B-20260105"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/meituan/EvoCUA-8B-20260105"}], "name": "EvoCUA 8B 20260105", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/api/models/meituan/EvoCUA-8B-20260105", "target_id": "github:meituan/evocua", "type": "variant_of"}], "review_status": "curated", "scope": "core", "summary": "EvoCUA computer-use agent checkpoint (20260105 release) with about 8.8B BF16 parameters in the Qwen3-VL architecture, tagged for OSWorld and for English and Chinese.", "topics": ["reinforcement-learning", "planning"], "updated_at": null, "url": "https://huggingface.co/meituan/EvoCUA-8B-20260105"}
{"aliases": ["url:cab7bd78abb1faa469e7", "Reallm-Labs/InfiGUI-R1-3B", "https://huggingface.co/Reallm-Labs/InfiGUI-R1-3B"], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/InfiX-ai/InfiGUI-R1-3B/tree/main", "gated": false, "hub_id": "InfiX-ai/InfiGUI-R1-3B"}, "environments": ["web", "desktop", "android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:41Z", "role": "primary", "sha256": "73496fb30838ba9ea291bb6d1e14004ac3318ac03a1ae7b5d6044e0dc14fd46f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/InfiX-ai/InfiGUI-R1-3B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:41Z", "role": "primary", "sha256": "213a5181a3516cf15d5f7b746986bc1e36cd6579acabd8aa1c2fb16a23a51ee0", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/InfiX-ai/InfiGUI-R1-3B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:33Z", "role": "primary", "sha256": "213a5181a3516cf15d5f7b746986bc1e36cd6579acabd8aa1c2fb16a23a51ee0", "supports": ["identity"], "url": "https://huggingface.co/api/models/Reallm-Labs/InfiGUI-R1-3B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:00Z", "role": "primary", "sha256": "5fd1399c88bdf406e7689765ccf0599b3b7e0ddbed082fa92bffdd7b120bd41a", "supports": ["environments"], "url": "https://raw.githubusercontent.com/Reallm-Labs/InfiGUI-R1/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:05:41Z", "role": "primary", "sha256": "73496fb30838ba9ea291bb6d1e14004ac3318ac03a1ae7b5d6044e0dc14fd46f", "supports": ["licenses.weights"], "url": "https://huggingface.co/InfiX-ai/InfiGUI-R1-3B/raw/main/README.md"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:15Z", "role": "primary", "sha256": "213a5181a3516cf15d5f7b746986bc1e36cd6579acabd8aa1c2fb16a23a51ee0", "supports": ["aliases"], "url": "https://huggingface.co/api/models/Reallm-Labs/InfiGUI-R1-3B"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:16Z", "role": "primary", "sha256": "213a5181a3516cf15d5f7b746986bc1e36cd6579acabd8aa1c2fb16a23a51ee0", "supports": ["aliases"], "url": "https://huggingface.co/api/models/InfiX-ai/InfiGUI-R1-3B"}], "id": "url:35b9879a351cc3b3d7d5", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:05:41Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T02:05:41Z", "url": "https://huggingface.co/InfiX-ai/InfiGUI-R1-3B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/InfiX-ai/InfiGUI-R1-3B"}], "name": "InfiGUI-R1 3B", "notes": "The official Hub API resolves the former Reallm-Labs model namespace to InfiX-ai. Two safetensors shards and model/tokenizer files are public and ungated. The card declares Apache-2.0 for this checkpoint; base-model terms remain separate. Its older loading example still uses the prior namespace.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/InfiX-ai/InfiGUI-R1-3B/raw/main/README.md", "target_id": "arxiv:2504.14239", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A Qwen2.5-VL checkpoint trained with the Actor2Reasoner procedure to combine interface grounding with planning and recovery reasoning.", "topics": ["grounding", "reinforcement-learning", "self-reflection"], "updated_at": null, "url": "https://huggingface.co/InfiX-ai/InfiGUI-R1-3B"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data/tree/main", "gated": false, "hub_id": "KDEGroup/UI-AGILE-Data"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:43:19Z", "role": "primary", "sha256": "ce35413bc1ef28338bc6fd9f132fa271a4f5e0bf23cecd7346f54adf0a964fd7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:43:19Z", "role": "primary", "sha256": "7782422de88830a497d9eb530bd2f0fb6283904cbb25dafcf2ff4d03358a377a", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/KDEGroup/UI-AGILE-Data"}], "id": "url:41abb1b7206689cf5057", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:43:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data"}], "name": "UI-AGILE-Data", "notes": "Public JSON, image archives and Parquet evaluation files are present. The card has no data-license declaration; the model’s MIT declaration is not inherited. The publisher describes fixes to AndroidControl action parameters, but those corrections were not independently reproduced or applied to upstream resources.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data/raw/main/README.md", "target_id": "arxiv:2507.22025", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "UI-AGILE training annotations and screenshots together with converted GUI-grounding and AndroidControl evaluation files.", "topics": ["grounding", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md", "population": "RL training corpus", "qualifier": "exact", "unit": "prompts", "value": 12600}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md", "population": "RL training corpus", "qualifier": "exact", "unit": "screenshots", "value": 11205}, {"basis": "reported", "evidence_url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md", "population": "RL training corpus", "qualifier": "exact", "unit": "desktop applications", "value": 87}], "evidence_urls": ["https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md"], "formats": ["Parquet"], "modalities": ["screenshots", "text", "coordinates"], "notes": "Publisher-reported counts cover a selection and reformatting of GroundCUA, not a new disjoint image collection. No independent image-overlap audit was performed.", "splits": ["train"], "supervision": ["instruction-to-click grounding", "target bounding boxes"]}}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T01:48:10Z", "role": "primary", "sha256": "34f0025f686d0e48e2baa5244056a4362018aaad77ac54abea7f1f6eb25c4276", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T01:48:10Z", "role": "primary", "sha256": "34f0025f686d0e48e2baa5244056a4362018aaad77ac54abea7f1f6eb25c4276", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md"}], "id": "url:4e71dc456c4112eed727", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T01:48:10Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T01:48:10Z", "url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL"}], "name": "GroundGUI RL Dataset", "notes": "The card declares MIT and attributes the source examples to GroundCUA. Its separate claim about UI-Vision overlap is not independently validated by this catalog.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL/raw/main/README.md", "target_id": "arxiv:2610.19100", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Desktop grounding prompts selected from GroundCUA and formatted for click-based reinforcement learning with target-box rewards.", "topics": ["grounding", "reinforcement-learning", "data-generation"], "updated_at": null, "url": "https://huggingface.co/datasets/PrentisAI/GroundGUI_RL"}
{"aliases": [], "authors": ["Tianlin Shi", "Andrej Karpathy", "Linxi Fan", "Jonathan Hernandez", "Percy Liang"], "details": {"publication": {"authors": ["Tianlin Shi", "Andrej Karpathy", "Linxi Fan", "Jonathan Hernandez", "Percy Liang"], "date_basis": "Publisher citation_publication_date metadata explicitly states 2017/07/17; this is separate from the conference interval.", "date_parts": [2017, 7, 17], "doi": null, "event_dates": "06--11 Aug (conference interval reported in the publisher BibTeX month field)", "evidence_urls": ["https://proceedings.mlr.press/v70/shi17a.html", "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"], "identity_basis": "No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.", "issue": null, "notes": "Publisher citation uses Tianlin Shi and Linxi Fan; the PDF also prints the nicknames Tim and Jim. The ordered publisher author names are retained. DOI not supplied in the retained publisher citation. Publisher metadata supplies the publication date separately from the conference interval.", "pages": "3135-3144", "pdf_url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf", "publisher": "PMLR", "series": "Proceedings of Machine Learning Research", "title": "World of Bits: An Open-Domain Platform for Web-Based Agents", "type": "conference-paper", "url": "https://proceedings.mlr.press/v70/shi17a.html", "venue": "Proceedings of the 34th International Conference on Machine Learning", "volume": "70"}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:21Z", "role": "primary", "sha256": "749d2a150e4c2bcde1b219a0ecb86c95546134a057892544e8495b5e45323c44", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.event_dates", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.publisher", "details.publication.series", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.mlr.press/v70/shi17a.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:54:39Z", "role": "primary", "sha256": "087008efb9d056df5fd9fbc724413d6a3f20cb14c7f1597d00bacac53d8342d6", "supports": ["details.publication.notes", "details.publication.pdf_url"], "url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:21Z", "role": "primary", "sha256": "749d2a150e4c2bcde1b219a0ecb86c95546134a057892544e8495b5e45323c44", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.series", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume", "kind", "name", "scope", "summary", "topics"], "url": "https://proceedings.mlr.press/v70/shi17a.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:54:39Z", "role": "primary", "sha256": "087008efb9d056df5fd9fbc724413d6a3f20cb14c7f1597d00bacac53d8342d6", "supports": ["details.publication.notes", "details.publication.pdf_url", "environments"], "url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"}], "id": "url:52a3261eb6a19b402e88", "identifiers": {}, "kind": "paper", "last_verified_at": "2026-10-17T09:53:21Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://proceedings.mlr.press/v70/shi17a.html"}, {"rel": "publication_pdf", "url": "https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"}], "name": "World of Bits: An Open-Domain Platform for Web-Based Agents", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. Publisher citation uses Tianlin Shi and Linxi Fan; the PDF also prints the nicknames Tim and Jim. The ordered publisher author names are retained.", "published_at": "2017-07-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces World of Bits, a web-task platform combining human demonstrations, keyboard and mouse actions, and cached HTTP traffic for repeatable task environments. It evaluates agents trained through behavioral cloning and reinforcement learning.", "topics": ["web-navigation", "reinforcement-learning", "evaluation", "data-generation"], "updated_at": null, "url": "https://proceedings.mlr.press/v70/shi17a.html"}
{"aliases": ["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/yifeizhou/pae-llava-7b/tree/main", "gated": false, "hub_id": "yifeizhou/pae-llava-7b"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "c0994582f81415c9b0d37e4e5018752e48e511178c4b3898c207a042992db845", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "1b76d2aebbe15954eff2b8859b884ec6504f8b67ed0b1045d8b3a83c0d132d3a", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/yifeizhou/pae-llava-7b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "3e5f24b3175afeba8635dacc7b66eba1ce89fea49996a20281ff2df5a41f1e24", "supports": ["environments", "summary"], "url": "https://raw.githubusercontent.com/amazon-science/PAE/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "c0994582f81415c9b0d37e4e5018752e48e511178c4b3898c207a042992db845", "supports": ["licenses.weights"], "url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md"}], "id": "url:5d61c1b5fc8dd650518d", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:02:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "cc-by-nc-4.0", "retrieved_at": "2026-10-17T02:02:28Z", "url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/yifeizhou/pae-llava-7b"}], "name": "PAE LLaVA 7B", "notes": "The model is publicly ungated and has seven safetensors shards plus tokenizer and configuration files. Its card declares CC-BY-NC-4.0 for this model. This weight declaration does not settle the separate PAE dataset or code licensing questions.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/yifeizhou/pae-llava-7b/raw/main/README.md", "target_id": "arxiv:2412.13194", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A released LLaVA checkpoint for visual browser interaction from the PAE autonomous task-proposal and policy-learning system.", "topics": ["web-navigation", "reinforcement-learning"], "updated_at": null, "url": "https://huggingface.co/yifeizhou/pae-llava-7b"}
{"aliases": [], "authors": [], "details": {"artifact_type": "model-checkpoint", "disabled": false, "download_url": "https://huggingface.co/meituan/EvoCUA-32B-20260105/tree/main", "gated": false, "hub_id": "meituan/EvoCUA-32B-20260105", "languages": ["en", "zh"], "modalities": [], "size_category_reported": null}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "f42d9537a05f2562b57e5d508f95c3d29a9107d06574318c5715c7da5fb2a86d", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/api/models/meituan/EvoCUA-32B-20260105"}, {"method": "huggingface_api", "retrieved_at": "2026-10-16T06:50:49Z", "role": "primary", "sha256": "f42d9537a05f2562b57e5d508f95c3d29a9107d06574318c5715c7da5fb2a86d", "supports": ["details", "licenses.weights"], "url": "https://huggingface.co/api/models/meituan/EvoCUA-32B-20260105"}], "id": "url:60a95f34072aeb2771cc", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T06:50:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card", "declared": "apache-2.0", "retrieved_at": "2026-10-16T06:50:49Z", "url": "https://huggingface.co/meituan/EvoCUA-32B-20260105"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/meituan/EvoCUA-32B-20260105"}], "name": "EvoCUA 32B 20260105", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/api/models/meituan/EvoCUA-32B-20260105", "target_id": "github:meituan/evocua", "type": "variant_of"}], "review_status": "curated", "scope": "core", "summary": "Larger EvoCUA computer-use agent checkpoint (20260105 release) with about 33B BF16 parameters in the Qwen3-VL architecture, tagged for OSWorld and for English and Chinese.", "topics": ["reinforcement-learning", "planning"], "updated_at": null, "url": "https://huggingface.co/meituan/EvoCUA-32B-20260105"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/ritzzai/GUI-R1/tree/main", "gated": false, "hub_id": "ritzzai/GUI-R1"}, "environments": ["desktop", "windows", "linux", "macos", "android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:30Z", "role": "primary", "sha256": "7b14996d1a99675990cb2f3aeada2b2f9898f9daee46d3680988e16f6365abe3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/ritzzai/GUI-R1/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "3eff86951de7c36dbd4af006ab3f1350213bdfea384494e59c8abc007793c979", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/ritzzai/GUI-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "94c7c78bede11dde727940f56592a04594c1c59438812dff47540bf53c1bed7d", "supports": ["environments"], "url": "https://raw.githubusercontent.com/ritzz-ai/GUI-R1/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:44:30Z", "role": "primary", "sha256": "7b14996d1a99675990cb2f3aeada2b2f9898f9daee46d3680988e16f6365abe3", "supports": ["licenses.weights"], "url": "https://huggingface.co/ritzzai/GUI-R1/raw/main/README.md"}], "id": "url:61a5321db10fc8a312b9", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:44:30Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:44:30Z", "url": "https://huggingface.co/ritzzai/GUI-R1/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/ritzzai/GUI-R1"}], "name": "GUI-R1 checkpoint bundle", "notes": "The manifest contains two safetensors shards in GUI-R1-3B and four in GUI-R1-7B. Both are counted as one Hub resource rather than duplicate root URLs. The model card declares MIT; this does not automatically apply to source training data or upstream base models.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/ritzzai/GUI-R1/raw/main/README.md", "target_id": "arxiv:2504.10458", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A single Hub release containing GUI-R1 3B and 7B action-model checkpoints trained with reinforcement learning across interface platforms.", "topics": ["grounding", "reinforcement-learning", "action-prediction"], "updated_at": null, "url": "https://huggingface.co/ritzzai/GUI-R1"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/openbmb/AgentCPM-GUI/tree/main", "gated": false, "hub_id": "openbmb/AgentCPM-GUI"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:33Z", "role": "primary", "sha256": "9016a68e833a7717c922d9363a6d6c56cc7cf6a7087543f5694507668fe9ab75", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/openbmb/AgentCPM-GUI/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:33Z", "role": "primary", "sha256": "4d76c8cc19f1ff4c850ba6cc566d50bbd2355e10921a2214e38c921cedfeed4c", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/openbmb/AgentCPM-GUI"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:03:33Z", "role": "primary", "sha256": "9016a68e833a7717c922d9363a6d6c56cc7cf6a7087543f5694507668fe9ab75", "supports": ["licenses.weights"], "url": "https://huggingface.co/openbmb/AgentCPM-GUI/raw/main/README.md"}], "id": "url:8ef3985dce697e5feea9", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:03:33Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T02:03:33Z", "url": "https://huggingface.co/openbmb/AgentCPM-GUI/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/openbmb/AgentCPM-GUI"}], "name": "AgentCPM-GUI 8B checkpoint", "notes": "The Hub provides four safetensors shards and associated model files without gating. Weight metadata declares Apache-2.0, while the prose license section refers specifically to code; these scopes are kept distinct. The card describes Android-app operation, and the CAGUI evaluation data remains a separate existing catalog asset.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/openbmb/AgentCPM-GUI/raw/main/README.md", "target_id": "arxiv:2506.01391", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "An eight-billion-parameter MiniCPM-V-based model that consumes smartphone screenshots and predicts compact actions for Chinese and English Android app tasks.", "topics": ["mobile-navigation", "grounding", "reinforcement-learning"], "updated_at": null, "url": "https://huggingface.co/openbmb/AgentCPM-GUI"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/inclusionAI/GUI-G2-7B/tree/main", "gated": false, "hub_id": "inclusionAI/GUI-G2-7B"}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "cf53a868a010a20b165386edaaafed8918071c35a3babf1766fbb51f020cb4a1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/inclusionAI/GUI-G2-7B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "c1f11fea30d3a64445e983a011db4efb9515a4221ad770e3a5223699e665eebe", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/inclusionAI/GUI-G2-7B"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "cf53a868a010a20b165386edaaafed8918071c35a3babf1766fbb51f020cb4a1", "supports": ["licenses.weights"], "url": "https://huggingface.co/inclusionAI/GUI-G2-7B/raw/main/README.md"}], "id": "url:90eea6f6d099a3297850", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:44:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T02:44:29Z", "url": "https://huggingface.co/inclusionAI/GUI-G2-7B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/inclusionAI/GUI-G2-7B"}], "name": "GUI-G2 7B", "notes": "The model card and four public safetensors shards establish this specific release. Apache-2.0 is declared for the model. The release targets element grounding; its grounding results do not establish long-horizon task-completion success.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/inclusionAI/GUI-G2-7B/raw/main/README.md", "target_id": "arxiv:2507.15846", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A Qwen2.5-VL grounding checkpoint trained with Gaussian point and coverage rewards whose granularity adapts to interface-element scale.", "topics": ["grounding", "reinforcement-learning", "reward-models"], "updated_at": null, "url": "https://huggingface.co/inclusionAI/GUI-G2-7B"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/Yuqi-Zhou/GUI-G1-3B-v1/tree/main", "gated": false, "hub_id": "Yuqi-Zhou/GUI-G1-3B-v1"}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:30Z", "role": "primary", "sha256": "b7a4628a4d65ae226b50086075f3977b8f69ad12d15238a194675875bb9113e8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/Yuqi-Zhou/GUI-G1-3B-v1/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:30Z", "role": "primary", "sha256": "2f4b2084ca9cfd67f0952ea8c578c87f3a5d721064b5de3a191d7489373125aa", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/Yuqi-Zhou/GUI-G1-3B-v1"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:44:30Z", "role": "primary", "sha256": "b7a4628a4d65ae226b50086075f3977b8f69ad12d15238a194675875bb9113e8", "supports": ["licenses.weights"], "url": "https://huggingface.co/Yuqi-Zhou/GUI-G1-3B-v1/raw/main/README.md"}], "id": "url:9dc9d8d3b9a4b32d6f4b", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:44:30Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:44:30Z", "url": "https://huggingface.co/Yuqi-Zhou/GUI-G1-3B-v1/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/Yuqi-Zhou/GUI-G1-3B-v1"}], "name": "GUI-G1 3B v1", "notes": "Two public safetensors shards and model/tokenizer configuration are present. The card declares MIT. Its inference example loads the v1 model but names a different 0.1K processor checkpoint; that example was not executed, and compatibility is not asserted.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/Yuqi-Zhou/GUI-G1-3B-v1/raw/main/README.md", "target_id": "arxiv:2505.15810", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A Qwen2.5-VL-3B checkpoint released for R1-style visual grounding experiments, predicting interface target locations from screenshots and instructions.", "topics": ["grounding", "reinforcement-learning"], "updated_at": null, "url": "https://huggingface.co/Yuqi-Zhou/GUI-G1-3B-v1"}
{"aliases": ["url:2f67f5ad7954632dfc4e", "THUDM/webrl-llama-3.1-8b", "https://huggingface.co/THUDM/webrl-llama-3.1-8b"], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/tree/main", "gated": false, "hub_id": "zai-org/webrl-llama-3.1-8b"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "4712446382f106a115a317d0ec6273a975545386b9cedf56946a96a56479dc9f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "6523b7f198d8debec2f3074190812955c6251590b61f8a6ea1bbb681b8ea128f", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/zai-org/webrl-llama-3.1-8b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "6523b7f198d8debec2f3074190812955c6251590b61f8a6ea1bbb681b8ea128f", "supports": ["identity"], "url": "https://huggingface.co/api/models/THUDM/webrl-llama-3.1-8b"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "4712446382f106a115a317d0ec6273a975545386b9cedf56946a96a56479dc9f", "supports": ["licenses.weights"], "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:06Z", "role": "primary", "sha256": "96621f060dddec91453d8c2b451b3793a34b34810d82c7cd13ce4cb05ca90a76", "supports": ["aliases"], "url": "https://huggingface.co/api/models/THUDM/webrl-llama-3.1-8b"}, {"method": "huggingface_repository_identity", "retrieved_at": "2026-10-17T19:15:06Z", "role": "primary", "sha256": "96621f060dddec91453d8c2b451b3793a34b34810d82c7cd13ce4cb05ca90a76", "supports": ["aliases"], "url": "https://huggingface.co/api/models/zai-org/webrl-llama-3.1-8b"}], "id": "url:a6ee20fea67438d91e06", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:05:40Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "other", "retrieved_at": "2026-10-17T02:05:40Z", "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b"}], "name": "WebRL Llama 3.1 8B", "notes": "The official Hub API resolves the former THUDM namespace to zai-org. Seven safetensors shards and configuration/tokenizer files are public and ungated. The card marks the weight license as other without naming complete terms, so no permissive weight license is inferred. The card’s evaluation prose also names GLM-4 despite the Llama checkpoint identity; no score is imported.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b/raw/main/README.md", "target_id": "arxiv:2411.02337", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A WebRL browser-policy checkpoint for performing actions in the five website environments of WebArena.", "topics": ["web-navigation", "reinforcement-learning"], "updated_at": null, "url": "https://huggingface.co/zai-org/webrl-llama-3.1-8b"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/KDEGroup/UI-AGILE/tree/main", "gated": false, "hub_id": "KDEGroup/UI-AGILE"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:30Z", "role": "primary", "sha256": "75aed1ebff7dd54ab1dc0112fe524a118d9e8845c2e079080d2bb7954f7ab36e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/KDEGroup/UI-AGILE/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "a52032bb185885d47466f0bea26c15ccefeb8d6ff89f3bacc5eb4192c35d45f7", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/KDEGroup/UI-AGILE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:43:19Z", "role": "primary", "sha256": "ce35413bc1ef28338bc6fd9f132fa271a4f5e0bf23cecd7346f54adf0a964fd7", "supports": ["environments"], "url": "https://huggingface.co/datasets/KDEGroup/UI-AGILE-Data/raw/main/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:44:30Z", "role": "primary", "sha256": "75aed1ebff7dd54ab1dc0112fe524a118d9e8845c2e079080d2bb7954f7ab36e", "supports": ["licenses.weights"], "url": "https://huggingface.co/KDEGroup/UI-AGILE/raw/main/README.md"}], "id": "url:b94e285c3f757e2d0f3a", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:44:30Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:44:30Z", "url": "https://huggingface.co/KDEGroup/UI-AGILE/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/KDEGroup/UI-AGILE"}], "name": "UI-AGILE checkpoint", "notes": "Four safetensors shards and model/tokenizer files are public. The card declares MIT for these weights; the companion dataset has no declared license and is separate. The short card does not state a parameter count, so this entry does not infer one from shard sizes.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/KDEGroup/UI-AGILE/raw/main/README.md", "target_id": "arxiv:2507.22025", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "The released UI-AGILE model for GUI grounding with reward shaping, crop-based resampling and decomposed selection during inference.", "topics": ["grounding", "reinforcement-learning", "efficiency"], "updated_at": null, "url": "https://huggingface.co/KDEGroup/UI-AGILE"}
{"aliases": [], "authors": ["Fei Tang", "Zhangxuan Gu", "Zhengxi Lu", "Shangzhan Zhang", "Zhengwen Zeng", "Shuheng Shen", "Changhua Meng", "Yuchen Yan", "Wenqi Zhang", "Yongliang Shen", "Weiming Lu", "Yueting Zhuang"], "details": {"publication": {"authors": ["Fei Tang", "Zhangxuan Gu", "Zhengxi Lu", "Shangzhan Zhang", "Zhengwen Zeng", "Shuheng Shen", "Changhua Meng", "Yuchen Yan", "Wenqi Zhang", "Yongliang Shen", "Weiming Lu", "Yueting Zhuang"], "date_basis": "Explicit publisher metadata: month = {June}, year = {2026}; precision=month.", "date_parts": [2026, 6], "doi": null, "event_dates": null, "evidence_urls": ["https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html", "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"], "identity_basis": "No title/DOI/canonical URL match in the current paper catalog or raw archive; closest substantive same-team works were reread and have different contributions.", "issue": null, "notes": "CVF open-access accepted version; PDF states that the final proceedings version is on IEEE Xplore.", "pages": "13007-13016", "pdf_url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf", "publisher": null, "series": null, "title": "GUI-SAGE: Enhancing GUI Automation with Self-Explanatory Learning", "type": "conference-paper", "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html", "venue": "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)", "volume": null}}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:48+00:00", "role": "primary", "sha256": "1d5ed4e24cf9763cc43f0a8ce14f0e98c01504c73578e5632cea573dfb7da647", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:56:04+00:00", "role": "primary", "sha256": "244f40d77a40dfd184e78cf8e1e2dd9e469d3e594a8ce214279c2ce17b8f1012", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.title"], "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:53:48+00:00", "role": "primary", "sha256": "1d5ed4e24cf9763cc43f0a8ce14f0e98c01504c73578e5632cea573dfb7da647", "supports": ["details.publication.date_basis", "details.publication.date_parts", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.type", "details.publication.url", "details.publication.venue", "environments", "kind", "scope", "summary", "topics"], "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:56:04+00:00", "role": "primary", "sha256": "244f40d77a40dfd184e78cf8e1e2dd9e469d3e594a8ce214279c2ce17b8f1012", "supports": ["details.publication.authors", "details.publication.identity_basis", "details.publication.notes", "details.publication.title", "environments", "name", "scope", "summary", "topics"], "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"}], "id": "url:beb8f1fd6b945ffc9d93", "identifiers": {}, "kind": "paper", "last_verified_at": "2026-10-17T09:53:48+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}, {"rel": "publication_pdf", "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.pdf"}], "name": "GUI-SAGE: Enhancing GUI Automation with Self-Explanatory Learning", "notes": "The annotation was reviewed against retained primary publication sources. Experiments have not been independently reproduced. New work proposal from the official CVF accepted open-access version. No invented day, DOI, or arXiv identifier; original annotations describe the reviewed sources and not independent reproduction.", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-SAGE generates reasoning conditioned on known actions and weights policy updates using confidence and reward signals. Its experiments use AndroidControl and GUIOdyssey data and separately evaluate interactive AndroidWorld tasks.", "topics": ["reinforcement-learning", "data-generation", "evaluation", "robustness"], "updated_at": null, "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/zai-org/webrl-orm-llama-3.1-8b/tree/main", "gated": false, "hub_id": "zai-org/webrl-orm-llama-3.1-8b", "languages": [], "modalities": [], "size_category_reported": null}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "8bafcdf63fe484b6005ca7082da1310459f8c87d4c011e84293a3cc8fca6ae9e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/api/models/zai-org/webrl-orm-llama-3.1-8b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:29Z", "role": "primary", "sha256": "8bafcdf63fe484b6005ca7082da1310459f8c87d4c011e84293a3cc8fca6ae9e", "supports": ["identity"], "url": "https://huggingface.co/api/models/THUDM/webrl-orm-llama-3.1-8b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "eb217b96c54c3c2ccfd3a27255c6fd740df40c2d6743ae811858339bf7f740b0", "supports": ["summary", "environments", "relations"], "url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T02:05:40Z", "role": "primary", "sha256": "8bafcdf63fe484b6005ca7082da1310459f8c87d4c011e84293a3cc8fca6ae9e", "supports": ["details", "licenses.weights"], "url": "https://huggingface.co/api/models/zai-org/webrl-orm-llama-3.1-8b"}], "id": "url:d1e85aeb6d96859c109d", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:05:40Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/zai-org/webrl-orm-llama-3.1-8b"}], "name": "WebRL outcome reward model 8B", "notes": "The official API resolves the former THUDM namespace to zai-org and lists four safetensors shards. No model card or weight-license declaration is present in the inspected manifest. The official WebRL README identifies this as its outcome-supervised reward model; it is a training component rather than an action policy.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/WebRL/HEAD/README.md", "target_id": "arxiv:2411.02337", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "The outcome-supervision checkpoint used by WebRL to score browser trajectories during policy training.", "topics": ["reward-models", "reinforcement-learning", "web-navigation"], "updated_at": null, "url": "https://huggingface.co/zai-org/webrl-orm-llama-3.1-8b"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/xuyifan/MobileRL-9B/tree/main", "gated": false, "hub_id": "xuyifan/MobileRL-9B", "languages": ["en"], "modalities": [], "size_category_reported": null}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:40Z", "role": "primary", "sha256": "a00f13f5a86f2de7eb9c866cc4d08b958d21b2ac30dc3b9e1662e4bb336b7cb3", "supports": ["name", "kind", "relations", "licenses.weights", "identity", "release_status", "details"], "url": "https://huggingface.co/api/models/xuyifan/MobileRL-9B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:03:41Z", "role": "primary", "sha256": "5c2d5eb247958f1af0646cbac831ecfbc7f839baa236a0d733075592da81f58e", "supports": ["licenses.weights"], "url": "https://huggingface.co/xuyifan/MobileRL-9B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:04Z", "role": "primary", "sha256": "752fad8ebd8b8eda03648329a6fc2585e5bcb2b96875b8aa2ce7d9f3fff78153", "supports": ["summary", "environments", "relations", "topics", "scope"], "url": "https://raw.githubusercontent.com/THUDM/MobileRL/HEAD/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T04:03:40Z", "role": "primary", "sha256": "a00f13f5a86f2de7eb9c866cc4d08b958d21b2ac30dc3b9e1662e4bb336b7cb3", "supports": ["details", "licenses.weights"], "url": "https://huggingface.co/api/models/xuyifan/MobileRL-9B"}], "id": "url:d5667c854845c77b9f5b", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T04:03:40Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card", "declared": "apache-2.0", "retrieved_at": "2026-10-17T04:03:40Z", "url": "https://huggingface.co/xuyifan/MobileRL-9B"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/xuyifan/MobileRL-9B"}], "name": "MobileRL 9B checkpoint", "notes": "Five public safetensors shards are present. The short card declares Apache-2.0 and a GLM-4.1V-9B base. The official project confirms the checkpoint and Android targets, and distinguishes MobileRL bounding-box actions from coordinate-format alternatives.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/THUDM/MobileRL/HEAD/README.md", "target_id": "arxiv:2509.18119", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A GLM-4.1V-based mobile action checkpoint released by the MobileRL project for Android interaction.", "topics": ["reinforcement-learning", "mobile-navigation"], "updated_at": null, "url": "https://huggingface.co/xuyifan/MobileRL-9B"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/InfiX-ai/InfiGUI-G1-7B/tree/main", "gated": false, "hub_id": "InfiX-ai/InfiGUI-G1-7B"}, "environments": ["cross-platform"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "253055f0f77266a64d2db5de8fd04c4202f4cda3be733dd1b00d8565116dd240", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/InfiX-ai/InfiGUI-G1-7B/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "10278eb0f1600557e432a0f8dde2c40cfffe12330ddb949b6a0fc02bfbc6c9ee", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/InfiX-ai/InfiGUI-G1-7B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:00Z", "role": "primary", "sha256": "4b68c82a2dd0063a450bde34c8bffbd931565a91a2982a5b47be7e711815b284", "supports": ["environments"], "url": "https://raw.githubusercontent.com/InfiXAI/InfiGUI-G1/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:44:29Z", "role": "primary", "sha256": "253055f0f77266a64d2db5de8fd04c4202f4cda3be733dd1b00d8565116dd240", "supports": ["licenses.weights"], "url": "https://huggingface.co/InfiX-ai/InfiGUI-G1-7B/raw/main/README.md"}], "id": "url:d70a755e3b2102f25f34", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:44:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-17T02:44:29Z", "url": "https://huggingface.co/InfiX-ai/InfiGUI-G1-7B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/InfiX-ai/InfiGUI-G1-7B"}], "name": "InfiGUI-G1 7B", "notes": "Four safetensors shards are public. The model card declares Apache-2.0 and identifies Qwen2.5-VL-7B-Instruct as its base. It is the G1 grounding checkpoint, distinct from the previously cataloged InfiGUI-R1 reasoning checkpoint; benchmark scores remain publisher claims.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/InfiX-ai/InfiGUI-G1-7B/raw/main/README.md", "target_id": "arxiv:2508.05731", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A GUI-grounding checkpoint trained with adaptive exploration rewards and multiple candidate answers to improve instruction-to-element alignment.", "topics": ["grounding", "reinforcement-learning"], "updated_at": null, "url": "https://huggingface.co/InfiX-ai/InfiGUI-G1-7B"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/Fanbin/ARPO_UITARS1.5_7B/tree/2c5604d86c529843ef346698edcf33b170c1e817", "gated": false, "hub_id": "Fanbin/ARPO_UITARS1.5_7B", "hub_revision": "2c5604d86c529843ef346698edcf33b170c1e817", "license_review": {"weights": {"evidence_urls": ["https://huggingface.co/Fanbin/ARPO_UITARS1.5_7B/raw/2c5604d86c529843ef346698edcf33b170c1e817/README.md"], "reason": "The inspected model card contains no checkpoint-license declaration. The code repository Apache license is not transferred to weights."}}, "repository_url": "https://github.com/JIA-Lab-research/ARPO"}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:12Z", "role": "primary", "sha256": "839b2de0e0c614fe8d2df4782d0f6a6e7d33f5c60d8020c9d4a01c61d533f39e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/JIA-Lab-research/ARPO/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:21Z", "role": "primary", "sha256": "6e19a5339c4e52ca3753cce26ac2034883b96f5cb26383a4b41bc87a5fbf5bf6", "supports": ["identity", "release_status", "details.hub_id", "details.hub_revision", "details.gated", "details.disabled", "details.download_url"], "url": "https://huggingface.co/api/models/Fanbin/ARPO_UITARS1.5_7B"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:16Z", "role": "primary", "sha256": "b6957fad37a7995e9dde4c7d4a4e0f6e2153b85354b2d234587c2a94ff95d0bd", "supports": ["summary", "notes", "licenses.weights"], "url": "https://huggingface.co/Fanbin/ARPO_UITARS1.5_7B/raw/2c5604d86c529843ef346698edcf33b170c1e817/README.md"}], "id": "url:db6fc5f29f186d13991c", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-18T04:27:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/Fanbin/ARPO_UITARS1.5_7B"}], "name": "ARPO UI-TARS-1.5-7B checkpoint", "notes": "The public Hub manifest lists four Safetensors shards, an index and configuration/tokenizer files. The model card reports a 15-step evaluation setting; no performance result is independently reproduced. No weight license is declared, despite Apache-2.0 being established for the separate code repository.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/JIA-Lab-research/ARPO/HEAD/README.md", "target_id": "arxiv:2505.16282", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "UI-TARS-1.5-7B-derived checkpoint trained with ARPO on OSWorld desktop tasks.", "topics": ["reinforcement-learning", "action-prediction"], "updated_at": null, "url": "https://huggingface.co/Fanbin/ARPO_UITARS1.5_7B"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/datasets/ritzzai/GUI-R1/tree/main", "gated": false, "hub_id": "ritzzai/GUI-R1"}, "environments": ["desktop", "windows", "linux", "macos", "android", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "7b14996d1a99675990cb2f3aeada2b2f9898f9daee46d3680988e16f6365abe3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "f31ae9547d94967dd91f5cde5e5c66430d7d59ee3a525c0914bc600a79999af7", "supports": ["identity", "release_status", "details", "licenses.data"], "url": "https://huggingface.co/api/datasets/ritzzai/GUI-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "94c7c78bede11dde727940f56592a04594c1c59438812dff47540bf53c1bed7d", "supports": ["summary", "environments", "release_status"], "url": "https://raw.githubusercontent.com/ritzz-ai/GUI-R1/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:42:32Z", "role": "primary", "sha256": "7b14996d1a99675990cb2f3aeada2b2f9898f9daee46d3680988e16f6365abe3", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md"}], "id": "url:df680385f18b01a825be", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T02:42:32Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:42:32Z", "url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/ritzzai/GUI-R1"}], "name": "GUI-R1 data bundle", "notes": "The Hub is a dataset repository with Parquet payloads, although the short card mistakenly calls it a model. The official project identifies it as training/evaluation data and describes subsequent expanded mixtures. MIT is declared in this dataset card; upstream dataset terms remain separate. No 3K/800K aggregate count is inferred for the mutable bundle.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/ritzzai/GUI-R1/raw/main/README.md", "target_id": "arxiv:2504.10458", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Training and evaluation Parquet bundles accompanying GUI-R1, including GUI grounding and mobile/web action-prediction data.", "topics": ["grounding", "reinforcement-learning", "evaluation"], "updated_at": null, "url": "https://huggingface.co/datasets/ritzzai/GUI-R1"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/LZXzju/Qwen2.5-VL-3B-UI-R1/tree/main", "gated": false, "hub_id": "LZXzju/Qwen2.5-VL-3B-UI-R1"}, "environments": ["web", "desktop", "android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:33Z", "role": "primary", "sha256": "620fbee20a6002bbc13562030ce4988d41ef17b0c033ec6872d83741f597adb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/LZXzju/Qwen2.5-VL-3B-UI-R1/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:03:33Z", "role": "primary", "sha256": "6a206702a1a4faef428460e07f52aaf8fd1f1a144d711d807ca42479b248f697", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/LZXzju/Qwen2.5-VL-3B-UI-R1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:00Z", "role": "primary", "sha256": "f55ac6a0c50d298d46c5f7083a0879959fd7cd3c1d91496af295ebf2afcc13a3", "supports": ["environments"], "url": "https://raw.githubusercontent.com/lll6gg/UI-R1/HEAD/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-17T02:03:33Z", "role": "primary", "sha256": "620fbee20a6002bbc13562030ce4988d41ef17b0c033ec6872d83741f597adb8", "supports": ["licenses.weights"], "url": "https://huggingface.co/LZXzju/Qwen2.5-VL-3B-UI-R1/raw/main/README.md"}], "id": "url:f13474d7aa734a30624d", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:03:33Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:03:33Z", "url": "https://huggingface.co/LZXzju/Qwen2.5-VL-3B-UI-R1/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/LZXzju/Qwen2.5-VL-3B-UI-R1"}], "name": "UI-R1 3B", "notes": "Two safetensors shards and tokenizer/configuration files are publicly ungated. The model card declares MIT; upstream base-model terms are separate. The mutable card compares v1, v2 and UI-R1-E results without pinning every result to a commit, so this record identifies the Hub release and does not import an unversioned score.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/LZXzju/Qwen2.5-VL-3B-UI-R1/raw/main/README.md", "target_id": "arxiv:2503.21620", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A Qwen2.5-VL checkpoint trained with reinforcement learning for GUI action prediction and evaluated on mobile, desktop and web grounding tasks.", "topics": ["grounding", "reinforcement-learning", "action-prediction"], "updated_at": null, "url": "https://huggingface.co/LZXzju/Qwen2.5-VL-3B-UI-R1"}