{"aliases": [], "authors": ["Shuyan Zhou", "Frank F. Xu", "Hao Zhu", "Xuhui Zhou", "Robert Lo", "Abishek Sridhar", "Xianyi Cheng", "Tianyue Ou", "Yonatan Bisk", "Daniel Fried", "Uri Alon", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0e9eca0165573ba53b663acbb1c7962fd2d64bc6633f10397f9cad229772a363", "submitted_at": "2023-07-25T22:59:32Z", "updated_at": "2024-04-16T15:13:18Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2307.13854v4"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:20Z", "role": "primary", "sha256": "cbe8bfcbce4170a94a6cf42de01fbe0a448e7b1d29fa75a0de877105a98ec08e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=1802.08802%2C1808.09132%2C2005.03776%2C2101.09465%2C2103.16057%2C2104.08560%2C2105.13231%2C2107.03374%2C2107.13731%2C2108.03353%2C2202.02312%2C2202.08137%2C2204.01691%2C2205.11029%2C2207.01206%2C2209.08199%2C2209.08655%2C2209.14927%2C2209.15099%2C2210.02663%2C2210.03629%2C2210.03945%2C2211.07615%2C2301.10165%2C2301.13280%2C2302.04761%2C2303.07280%2C2303.08774%2C2303.17491%2C2303.17580%2C2304.04370%2C2304.07061%2C2305.04091%2C2305.05662%2C2305.08144%2C2305.10601%2C2305.11854%2C2305.12487%2C2305.14257%2C2305.14325%2C2305.16504%2C2305.16653%2C2305.19308%2C2306.00245%2C2306.06070%2C2306.07863%2C2307.10088%2C2307.12856%2C2307.13854%2C2308.01990&max_results=50"}], "id": "arxiv:2307.13854", "identifiers": {"arxiv": "2307.13854"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2307.13854"}], "name": "WebArena: A Realistic Web Environment for Building Autonomous Agents", "notes": "Editorial summary based on the primary paper abstract; experiments and released artifacts have not been independently reproduced.", "published_at": "2023-07-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces reproducible web environments with task-completion checks for evaluating autonomous web agents.", "topics": ["evaluation", "web-navigation", "long-horizon"], "updated_at": "2024-04-16", "url": "https://arxiv.org/abs/2307.13854"}
{"aliases": [], "authors": ["Hiroki Furuta", "Yutaka Matsuo", "Aleksandra Faust", "Izzeddin Gur"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4b00e0a9c74b7ac8819bdda540a1069c33cbac45f0c7fb853dee995f2feb9a59", "submitted_at": "2023-11-30T17:50:47Z", "updated_at": "2024-12-31T04:24:16Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2311.18751v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:24Z", "role": "primary", "sha256": "3aac6e9748769efff3c475ede0da192776daa03431d231b5f67e2aa50f8a3be9", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2308.03688%2C2308.15272%2C2309.08172%2C2309.11436%2C2310.04716%2C2310.04869%2C2310.05126%2C2310.06770%2C2310.08740%2C2310.10634%2C2310.11441%2C2310.16042%2C2310.18331%2C2311.01767%2C2311.07562%2C2311.09278%2C2311.12983%2C2311.18751%2C2312.03003%2C2312.06677%2C2312.08914%2C2312.11671%2C2312.13108%2C2312.13771%2C2312.15820%2C2312.16062%2C2401.01614%2C2401.04124%2C2401.07781%2C2401.10935%2C2401.13178%2C2401.13649%2C2401.13919%2C2401.14268%2C2401.16158%2C2402.01030%2C2402.04476%2C2402.04615%2C2402.05930%2C2402.06596%2C2402.07456%2C2402.07939%2C2402.07945%2C2402.08567%2C2402.11941%2C2402.14073%2C2402.15057%2C2402.16965%2C2402.17553%2C2403.02713&max_results=50"}], "id": "arxiv:2311.18751", "identifiers": {"arxiv": "2311.18751"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2311.18751"}], "name": "Exposing Limitations of Language Model Agents in Sequential-Task Compositions on the Web", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2023-11-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "CompWoB adds 50 composed web-automation tasks to study generalization from simpler interactions. The paper compares prompted and fine-tuned agents, balances task training data, and probes sensitivity to changes in instruction order.", "topics": ["evaluation", "long-horizon", "robustness", "web-navigation"], "updated_at": "2024-12-31", "url": "https://arxiv.org/abs/2311.18751"}
{"aliases": [], "authors": ["Shulin Tian", "Ziniu Zhang", "Liangyu Chen", "Ziwei Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "667b1f5053f3ae6ea712999932d5e9ee2f42e402cc72e713048d0790f5931976", "submitted_at": "2024-04-15T17:59:50Z", "updated_at": "2025-06-29T05:24:54Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.09992v2"}, "publication": {"authors": ["Shulin Tian", "Ziniu Zhang", "Liang-Yu Chen", "Ziwei Liu"], "date_basis": "Explicit publisher citation_publication_date: 2025/7. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2025, 7], "doi": "10.18653/v1/2025.findings-acl.703", "event_dates": null, "evidence_urls": ["https://aclanthology.org/2025.findings-acl.703/"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "13682-13697", "pdf_url": "https://aclanthology.org/2025.findings-acl.703.pdf", "publisher": "Association for Computational Linguistics", "series": null, "title": "MMInA: Benchmarking Multihop Multimodal Internet Agents", "type": "conference-paper", "url": "https://aclanthology.org/2025.findings-acl.703/", "venue": "Findings of the Association for Computational Linguistics: ACL 2025", "volume": null}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:09:51Z", "role": "primary", "sha256": "5e725b77b94b484284adaab9ec70674c9fc13a7fce34b764e59561f9180b915f", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.publisher", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue"], "url": "https://aclanthology.org/2025.findings-acl.703/"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2404.09992", "identifiers": {"arxiv": "2404.09992"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2404.09992"}, {"rel": "publication", "url": "https://aclanthology.org/2025.findings-acl.703/"}, {"rel": "publication_pdf", "url": "https://aclanthology.org/2025.findings-acl.703.pdf"}], "name": "MMInA: Benchmarking Multihop Multimodal Internet Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates multimodal agents on compositional tasks across changing websites and studies replaying prior trajectories as memory support.", "topics": ["evaluation", "long-horizon", "memory"], "updated_at": "2025-06-29", "url": "https://arxiv.org/abs/2404.09992"}
{"aliases": [], "authors": ["Zilong Wang", "Yuedong Cui", "Li Zhong", "Zimin Zhang", "Da Yin", "Bill Yuchen Lin", "Jingbo Shang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c443e6a508c438332ac5990cf254a4441f5bdcc24354fd8aeca36bcbf4604a29", "submitted_at": "2024-07-26T19:27:17Z", "updated_at": "2024-07-26T19:27:17Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.19056v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.19056", "identifiers": {"arxiv": "2407.19056"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.19056"}], "name": "OfficeBench: Benchmarking Language Agents across Multiple Applications for Office Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates language agents on office workflows requiring information integration, long-term planning, and switching between applications.", "topics": ["evaluation", "cross-app", "long-horizon"], "updated_at": "2024-07-26", "url": "https://arxiv.org/abs/2407.19056"}
{"aliases": [], "authors": ["Yao Zhang", "Zijian Ma", "Yunpu Ma", "Zhen Han", "Yu Wu", "Volker Tresp"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9e3a34f16101df7a96ffdabbdf86b060e195494f16499f20bdf54110a0394cae", "submitted_at": "2024-08-28T17:49:29Z", "updated_at": "2024-08-28T17:49:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2408.15978v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2408.15978", "identifiers": {"arxiv": "2408.15978"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2408.15978"}], "name": "WebPilot: A Versatile and Autonomous Multi-Agent System for Web Task Execution with Strategic Exploration", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-08-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebPilot separates global subtask planning from local Monte Carlo tree search during web execution. It continually refines the high-level plan and adapts local search to uncertain transitions and incomplete observations, evaluated on WebArena and MiniWoB++.", "topics": ["planning", "web-navigation", "long-horizon"], "updated_at": "2024-08-28", "url": "https://arxiv.org/abs/2408.15978"}
{"aliases": [], "authors": ["Frank F. Xu", "Yufan Song", "Boxuan Li", "Yuxuan Tang", "Kritanjali Jain", "Mengxue Bao", "Zora Z. Wang", "Xuhui Zhou", "Zhitong Guo", "Murong Cao", "Mingyang Yang", "Hao Yang Lu", "Amaad Martin", "Zhe Su", "Leander Maben", "Raj Mehta", "Wayne Chi", "Lawrence Jang", "Yiqing Xie", "Shuyan Zhou", "Graham Neubig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "77f91dd9ec9f50fb5526a78cfe270a7a1705d5bd83a2aef26644453740c005d5", "submitted_at": "2024-12-18T18:55:40Z", "updated_at": "2025-09-10T08:35:19Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2412.14161v3"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2412.14161", "identifiers": {"arxiv": "2412.14161"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2412.14161"}], "name": "TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-12-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "Simulates a software-company workplace to evaluate agents on professional tasks involving websites, programs, files, and coworker communication.", "topics": ["evaluation", "long-horizon", "workflow"], "updated_at": "2025-09-10", "url": "https://arxiv.org/abs/2412.14161"}
{"aliases": [], "authors": ["Zhenhailong Wang", "Haiyang Xu", "Junyang Wang", "Xi Zhang", "Ming Yan", "Ji Zhang", "Fei Huang", "Heng Ji"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5cf28f084cac7fc38e5c2dc4654115356d2d57f16009aad8925e071b8005be9c", "submitted_at": "2025-01-20T20:35:46Z", "updated_at": "2025-01-28T16:58:02Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2501.11733v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "arxiv:2501.11733", "identifiers": {"arxiv": "2501.11733"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:38Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2501.11733"}], "name": "Mobile-Agent-E: Self-Evolving Mobile Assistant for Complex Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-01-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses hierarchical mobile agents and persistent tips and reusable action shortcuts to improve complex tasks from accumulated experience.", "topics": ["mobile-navigation", "memory", "long-horizon"], "updated_at": "2025-01-28", "url": "https://arxiv.org/abs/2501.11733"}
{"aliases": [], "authors": ["Lutfi Eren Erdogan", "Nicholas Lee", "Sehoon Kim", "Suhong Moon", "Hiroki Furuta", "Gopala Anumanchipalli", "Kurt Keutzer", "Amir Gholami"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6b03d84f29bb524d7950039624e0d460d8381f2ba63d70c69f5f958013730db9", "submitted_at": "2025-03-12T17:40:52Z", "updated_at": "2025-04-22T17:56:22Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2503.09572v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:45:04Z", "role": "primary", "sha256": "bab88dfed88ca440c231aa768530374a0f19498777a70921481a904fddd46bb8", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2303.17760%2C2310.03720%2C2312.11190%2C2403.02502%2C2406.12639%2C2409.18060%2C2410.00201%2C2501.01702%2C2501.11425%2C2502.02533%2C2502.02955%2C2502.12110%2C2502.15760%2C2502.17110%2C2502.19328%2C2503.02197%2C2503.02403%2C2503.02682%2C2503.06580%2C2503.07365%2C2503.09572%2C2503.14021%2C2503.24180%2C2504.02357%2C2504.11675%2C2504.13936&max_results=26"}], "id": "arxiv:2503.09572", "identifiers": {"arxiv": "2503.09572"}, "kind": "paper", "last_verified_at": "2026-10-16T06:45:04Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2503.09572"}], "name": "Plan-and-Act: Improving Planning of Agents for Long-Horizon 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": "2025-03-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Separates planning from execution and creates synthetic plan annotations for training agents on extended web-navigation tasks.", "topics": ["planning", "long-horizon", "data-generation"], "updated_at": "2025-04-22", "url": "https://arxiv.org/abs/2503.09572"}
{"aliases": [], "authors": ["Suyu Ye", "Haojun Shi", "Darren Shih", "Hyokun Yun", "Tanya Roosta", "Tianmin Shu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "268adf347e1f224f434e6dba838873c99a4050686cae1495643d9638d840d809", "submitted_at": "2025-04-14T17:36:46Z", "updated_at": "2025-12-01T15:22:40Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.10445v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2504.10445", "identifiers": {"arxiv": "2504.10445"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.10445"}], "name": "RealWebAssist: A Benchmark for Long-Horizon Web Assistance with Real-World Users", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "RealWebAssist evaluates web assistance through sequences of instructions collected from real users across multiple websites. It tests intent interpretation, user-specific routines, long-term context, and visual grounding when instructions are ambiguous or evolve during interaction.", "topics": ["evaluation", "long-horizon", "human-in-the-loop", "grounding"], "updated_at": "2025-12-01", "url": "https://arxiv.org/abs/2504.10445"}
{"aliases": [], "authors": ["Mariya Davydova", "Daniel Jeffries", "Patrick Barker", "Arturo Márquez Flores", "Sinéad Ryan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e8bb4c6f8080ed1ee91e1089233cae6f231ab3256b38ac7d6403ebea3ae56fe0", "submitted_at": "2025-05-06T14:29:47Z", "updated_at": "2025-05-06T14:29:47Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.03570v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:47Z", "role": "primary", "sha256": "27144160e8981bb9f36b60c391689774913ff3f026eeeffc2923154c404a6529", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2504.03206%2C2504.03601%2C2504.06821%2C2504.09723%2C2504.11301%2C2504.11543%2C2504.12609%2C2504.14233%2C2504.14348%2C2504.14603%2C2504.16073%2C2504.16419%2C2504.17892%2C2504.18580%2C2504.20965%2C2505.00019%2C2505.00693%2C2505.01578%2C2505.02094%2C2505.03570%2C2505.03574%2C2505.05026%2C2505.05849%2C2505.06507%2C2505.07003%2C2505.07473%2C2505.10468%2C2505.10887%2C2505.12053%2C2505.12490%2C2505.12728%2C2505.13291%2C2505.14030%2C2505.14454%2C2505.14646%2C2505.15801%2C2505.15816%2C2505.17020%2C2505.18746%2C2505.19095%2C2505.19713%2C2505.20023%2C2505.20732%2C2505.20795%2C2505.21055%2C2505.22304%2C2505.22429%2C2505.23450%2C2505.23559%2C2506.00641&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:17Z", "role": "primary", "sha256": "39de8f4a9aee74cc0eae64e24efc08ed070714eaaea4a30dbc20cb08397eb9bd", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2505.03570v1"}], "id": "arxiv:2505.03570", "identifiers": {"arxiv": "2505.03570"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:17Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.03570"}], "name": "OSUniverse: Benchmark for Multimodal GUI-navigation AI Agents", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OSUniverse, a desktop GUI benchmark spanning precise clicks through multistep and multiapplication tasks. It provides both manual scoring and an automated validation mechanism.", "topics": ["evaluation", "cross-app", "long-horizon"], "updated_at": "2025-05-06", "url": "https://arxiv.org/abs/2505.03570"}
{"aliases": [], "authors": ["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": ["Brandon Man", "Ghadi Nehme", "Md Ferdous Alam", "Faez Ahmed"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f913b87f07fdad0e4c45264c58c8d9b9d13ca26695306d9e0644eb856f18ced7", "submitted_at": "2025-05-30T17:39:52Z", "updated_at": "2025-11-08T18:27:03Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2505.24838v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T12:02:34.311949+00:00", "role": "primary", "sha256": "a303d95a811331cd6ee23b0340f1b9ef2dd996835bfbc66656a871ff94eead50", "supports": ["environments", "notes"], "url": "https://arxiv.org/html/2505.24838v2"}], "id": "arxiv:2505.24838", "identifiers": {"arxiv": "2505.24838"}, "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/2505.24838"}], "name": "VideoCAD: A Dataset and Model for Learning Long-Horizon 3D CAD UI Interactions from Video", "notes": "The contribution was reviewed from the retained abstract. Sections 1 and 3 of the exact-version v2 paper identify Onshape as the browser-based interaction environment; the use of pixel-level input does not make the operated surface a desktop application. No experimental results were independently reproduced.", "published_at": "2025-05-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces synthetic CAD interaction videos, a model for learning interface actions from video, and evaluation tasks for spatial and long-horizon reasoning.", "topics": ["video", "data-generation", "science", "long-horizon"], "updated_at": "2025-11-08", "url": "https://arxiv.org/abs/2505.24838"}
{"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": ["Yuan Guo", "Tingjia Miao", "Zheng Wu", "Pengzhou Cheng", "Ming Zhou", "Zhuosheng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8dbc84aa582b94775d490cf84e17eff05be84af7dbcc2de7de61b886fa3c9178", "submitted_at": "2025-06-10T16:45:29Z", "updated_at": "2025-06-10T16:45:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.08972v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2506.08972", "identifiers": {"arxiv": "2506.08972"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.08972"}], "name": "Atomic-to-Compositional Generalization for Mobile Agents with A New Benchmark and Scheduling System", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-NEXUS evaluates compositional mobile tasks in local utility apps and online Chinese and English service apps. AGENT-NEXUS decomposes those long tasks into self-contained atomic subtasks to study generalization beyond isolated screens or short action chains.", "topics": ["evaluation", "long-horizon", "planning", "mobile-navigation"], "updated_at": "2025-06-10", "url": "https://arxiv.org/abs/2506.08972"}
{"aliases": [], "authors": ["Yuquan Xie", "Zaijing Li", "Rui Shao", "Gongwei Chen", "Kaiwen Zhou", "Yinchuan Li", "Dongmei Jiang", "Liqiang Nie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9b353606ce29a5bc60d263e86827a825af46883e25da5bd290b48affa9dd3838", "submitted_at": "2025-06-12T06:21:19Z", "updated_at": "2025-06-12T06:21:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.10387v1"}}, "environments": ["android", "web", "cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2506.10387", "identifiers": {"arxiv": "2506.10387"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.10387"}], "name": "Mirage-1: Augmenting and Updating GUI Agent with Hierarchical Multimodal Skills", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Mirage-1 organizes interaction trajectories into execution, core, and meta-skills, then uses them to guide Monte Carlo tree search during online GUI tasks. It introduces AndroidLH for long-horizon evaluation and tests the approach on Android and web-navigation benchmarks.", "topics": ["planning", "memory", "long-horizon", "mobile-navigation", "web-navigation"], "updated_at": "2025-06-12", "url": "https://arxiv.org/abs/2506.10387"}
{"aliases": [], "authors": ["Jingxu Xie", "Dylan Xu", "Xuandong Zhao", "Dawn Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ae36b271f9b4799093831cabee772845edfaeca97107414e424e170ccf89904d", "submitted_at": "2025-06-17T05:46:52Z", "updated_at": "2026-03-02T00:51:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2506.14205v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:13+00:00", "role": "primary", "sha256": "42e8b8dead7caea3f8b21e93a075e8c895dfc30a25e1470bf97a90940bbf0106", "supports": ["scope", "environments"], "url": "https://raw.githubusercontent.com/sunblaze-ucb/AgentSynth/HEAD/README.md"}], "id": "arxiv:2506.14205", "identifiers": {"arxiv": "2506.14205"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2506.14205"}], "name": "AgentSynth: Scalable Task Generation for Generalist Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-06-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "AgentSynth composes easily generated subtasks into longer computer-use tasks and trajectories with adjustable difficulty. Its evaluation uses the number of subtasks to probe how agent success changes as task complexity grows.", "topics": ["data-generation", "evaluation", "long-horizon"], "updated_at": "2026-03-02", "url": "https://arxiv.org/abs/2506.14205"}
{"aliases": [], "authors": ["Ning Li", "Xiangmou Qu", "Jiamu Zhou", "Jun Wang", "Muning Wen", "Kounianhua Du", "Xingyu Lou", "Qiuying Peng", "Jun Wang", "Weinan Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6d58a188742fef5862b299c1dcf17bb6be139e81aafd3d2e1bdbe26f88001384", "submitted_at": "2025-07-21T09:37:05Z", "updated_at": "2025-07-21T09:37:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2507.16853v1"}}, "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:2507.16853", "identifiers": {"arxiv": "2507.16853"}, "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/2507.16853"}], "name": "MobileUse: A GUI Agent with Hierarchical Reflection for Autonomous Mobile Operation", "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": "MobileUse reflects on errors at action and task timescales, invoking reflection when needed and using proactive exploration to learn unfamiliar apps. It evaluates this mobile execution framework on AndroidWorld and AndroidLab and describes a toolkit for physical-device operation.", "topics": ["mobile-navigation", "self-reflection", "robustness", "long-horizon"], "updated_at": "2025-07-21", "url": "https://arxiv.org/abs/2507.16853"}
{"aliases": [], "authors": ["Zihan Zheng", "Tianle Cui", "Taoran Wang", "Fengtao Wang", "Jiahui Pan", "Lewei He", "Qianglong Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "57d89909534af90cbc10b97a992238bde3ae976ac6cc97778b4345635cef552c", "submitted_at": "2025-08-02T11:53:41Z", "updated_at": "2026-04-20T02:44:32Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2508.01330v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2508.01330", "identifiers": {"arxiv": "2508.01330"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.01330"}], "name": "NaturalGAIA: A Verifiable Benchmark and Hierarchical Framework for Long-Horizon GUI Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "NaturalGAIA builds verifiable long-horizon GUI tasks from human interaction intents while separating causal task structure from its verbal description. Its LightManus-Jarvis framework divides dynamic planning and context management from execution based on visual and structural interface information.", "topics": ["evaluation", "long-horizon", "planning"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2508.01330"}
{"aliases": [], "authors": ["Jaewoo Ahn", "Junseo Kim", "Heeseung Yun", "Jaehyeon Son", "Dongmin Park", "Jaewoong Cho", "Gunhee Kim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3bf86652aaeb252972f3010be38b75b78dfaa9a8063c480c80d0cea47397c2e7", "submitted_at": "2025-09-01T01:33:16Z", "updated_at": "2025-10-15T10:33:27Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.01052v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2509.01052", "identifiers": {"arxiv": "2509.01052"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.01052"}], "name": "FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-09-01", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "FlashAdventure evaluates GUI agents on completing story arcs in 34 adventure games. It pairs an automated gameplay judge with COAST, a framework that keeps long-term clue memory to plan and complete sequential objectives.", "topics": ["evaluation", "long-horizon", "memory", "planning"], "updated_at": "2025-10-15", "url": "https://arxiv.org/abs/2509.01052"}
{"aliases": [], "authors": ["Jingru Fan", "Yufan Dang", "Jingyao Wu", "Huatao Li", "Runde Yang", "Xiyuan Yang", "Yuheng Wang", "Chen Qian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "49a887b12b3395821aa100a66e61aac808309ef947ae752475df3d2db9a4fcd8", "submitted_at": "2025-09-02T15:48:21Z", "updated_at": "2025-10-17T00:57:58Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2509.02444v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:01:10.946679+00:00", "role": "primary", "sha256": "594b5261dd2039777d36c46cdb8525d973ad57a154b902ff1a171bcbed2c64cc", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+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=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:01:10.946679+00:00", "role": "primary", "sha256": "594b5261dd2039777d36c46cdb8525d973ad57a154b902ff1a171bcbed2c64cc", "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=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:14:33.619660+00:00", "role": "primary", "sha256": "6c2591dad8893f87cffdea211584a3733dc2137dafa9416c1885d1b0126fd540", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/pdf/2509.02444v2"}], "id": "arxiv:2509.02444", "identifiers": {"arxiv": "2509.02444"}, "kind": "paper", "last_verified_at": "2026-10-18T10:01:10.946679+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.02444"}], "name": "AppCopilot: Toward General, Accurate, Long-Horizon, and Efficient Mobile Agent", "notes": "Trace reuse assumes stable layouts and fixed tasks. The described function-call routes are user-selected, and coordinate calibration depends on valid detected widgets. No particular operating-system coverage, latency guarantee or reproduced result is asserted. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (4.2 OCR/OR Region Localization; 4.2.2 Coordinate Calibration; 4.3 Multi-Agent Voting; 5.2 Task Decomposition; 6.2 Experience Reuse; 6.3 Function Calling); not an entire-paper reading or reproduction claim.", "published_at": "2025-09-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a mobile-agent stack combining region-based coordinate correction, hierarchical task planning, multi-agent voting, and reuse of successful interaction traces.", "topics": ["planning", "grounding", "workflow", "mobile-navigation", "long-horizon"], "updated_at": "2025-10-17", "url": "https://arxiv.org/abs/2509.02444"}
{"aliases": [], "authors": ["Gowen Loo", "Chang Liu", "Qinghong Yin", "Xiang Chen", "Jiawei Chen", "Jingyuan Zhang", "Yu Tian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a8b2db00f62afe350e55e2266b3627de9b0cf089178b0cf80eb13b06275eee15", "submitted_at": "2025-09-04T05:22:42Z", "updated_at": "2025-09-04T05:22:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.03891v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:49Z", "role": "primary", "sha256": "c208e36ae75cef4b32cfcb2f8182dd4aaaa218e3119bb3d4a2c20a5832f8b1ee", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2506.10387%2C2506.14866%2C2506.20332%2C2507.00008%2C2507.02592%2C2507.03730%2C2507.04227%2C2507.05445%2C2507.05673%2C2507.05720%2C2507.05791%2C2507.14293%2C2507.14799%2C2507.15846%2C2507.16853%2C2507.19132%2C2507.19478%2C2507.21071%2C2507.22025%2C2507.22358%2C2507.23779%2C2508.00935%2C2508.01330%2C2508.03700%2C2508.03923%2C2508.04010%2C2508.04026%2C2508.04482%2C2508.04700%2C2508.05615%2C2508.05731%2C2508.09057%2C2508.09123%2C2508.13167%2C2508.14040%2C2508.15144%2C2508.18689%2C2508.19679%2C2508.20096%2C2509.02544%2C2509.03891%2C2509.06477%2C2509.10723%2C2509.11250%2C2509.11543%2C2509.15221%2C2509.15566%2C2509.15738%2C2509.18119%2C2510.01354&max_results=50"}], "id": "arxiv:2509.03891", "identifiers": {"arxiv": "2509.03891"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:49Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2509.03891"}], "name": "MobileRAG: Enhancing Mobile Agent with Retrieval-Augmented Generation", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-09-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileRAG combines retrieval and memory components to support mobile tasks that need external knowledge or long action sequences. MobileRAG-Eval is introduced to study these requirements on complex real-world app tasks.", "topics": ["memory", "mobile-navigation", "evaluation", "long-horizon"], "updated_at": "2025-09-04", "url": "https://arxiv.org/abs/2509.03891"}
{"aliases": [], "authors": ["Musen Lin", "Minghao Liu", "Taoran Lu", "Lichen Yuan", "Yiwei Liu", "Haonan Xu", "Yu Miao", "Yuhao Chao", "Zhaojian Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "035048459d2b303721ef0738212829d1fa3314ccd0736b6649cc786c6e975ad4", "submitted_at": "2025-09-19T08:09:18Z", "updated_at": "2025-09-19T08:09:18Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2509.15738v1"}}, "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:2509.15738", "identifiers": {"arxiv": "2509.15738"}, "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/2509.15738"}], "name": "GUI-ReWalk: Massive Data Generation for GUI Agent via Stochastic Exploration and Intent-Aware 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": "2025-09-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-ReWalk begins with stochastic interface exploration and then infers goals to generate coherent action trajectories. It supports long-horizon workflows across applications and studies training on the resulting data using grounding and navigation benchmarks.", "topics": ["data-generation", "cross-app", "long-horizon", "planning"], "updated_at": "2025-09-19", "url": "https://arxiv.org/abs/2509.15738"}
{"aliases": [], "authors": ["Gonzalo Gonzalez-Pumariega", "Vincent Tu", "Chih-Lun Lee", "Jiachen Yang", "Ang Li", "Xin Eric Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "29c00ba4713da26f912fec8046d498b4fd10cf71a3109f66e92fa23f7cebcf44", "submitted_at": "2025-10-02T17:37:08Z", "updated_at": "2026-02-03T23:48:45Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.02250v2"}}, "environments": ["desktop", "windows", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.02250", "identifiers": {"arxiv": "2510.02250"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.02250"}], "name": "Scaling Agents for Computer Use", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Behavior Judge, which converts execution trajectories into behavior narratives and compares multiple candidate rollouts. Experiments examine this selection approach on OSWorld, WindowsAgentArena and AndroidWorld.", "topics": ["evaluation", "reward-models", "long-horizon"], "updated_at": "2026-02-03", "url": "https://arxiv.org/abs/2510.02250"}
{"aliases": [], "authors": ["Shang Ma", "Xusheng Xiao", "Yanfang Ye"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4bf0a561460c080b14da6fd13f6d92781e01e52b08dc691fe019e70b88e44c86", "submitted_at": "2025-10-07T15:36:04Z", "updated_at": "2026-09-27T17:25:57Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2510.06042v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T10:01:10.946679+00:00", "role": "primary", "sha256": "594b5261dd2039777d36c46cdb8525d973ad57a154b902ff1a171bcbed2c64cc", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+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=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:01:10.946679+00:00", "role": "primary", "sha256": "594b5261dd2039777d36c46cdb8525d973ad57a154b902ff1a171bcbed2c64cc", "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=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T10:14:29.861666+00:00", "role": "primary", "sha256": "5e2cb2f053983140904ec15c097d060ff18315c6c3c956bbcacbec26921709c5", "supports": ["summary", "scope", "environments", "notes"], "url": "https://arxiv.org/pdf/2510.06042v3"}], "id": "arxiv:2510.06042", "identifiers": {"arxiv": "2510.06042"}, "kind": "paper", "last_verified_at": "2026-10-18T10:01:10.946679+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.06042"}], "name": "Agent+P: Guiding UI Agents via Symbolic Planning", "notes": "Planning guarantees are relative to the modeled graph and selected target. They do not establish complete graphs, correct grounding or faithful goal translation. The described design depends on static source analysis; no experiments were reproduced. Basis: complete retained abstract and source-reviewed sections of the exact-version paper (3.3 UI Automation to Classical Planning; 4.1 UTG Builder; 4.2 Node Selector; 4.3 Plan Generator; 4.4 UI Explorer; 5.1 Datasets); not an entire-paper reading or reproduction claim.", "published_at": "2025-10-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Guides UI agents with symbolic planning over a UI transition graph initialized from app-code analysis and updated through online exploration, with evaluation on AndroidWorld.", "topics": ["planning", "long-horizon", "mobile-navigation"], "updated_at": "2026-09-27", "url": "https://arxiv.org/abs/2510.06042"}
{"aliases": [], "authors": ["Yuanyi Song", "Heyuan Huang", "Qiqiang Lin", "Yin Zhao", "Xiangmou Qu", "Jun Wang", "Xingyu Lou", "Weiwen Liu", "Zhuosheng Zhang", "Jun Wang", "Yong Yu", "Weinan Zhang", "Zhaoxiang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fc45dc62068c971cd831fc3db64bd3d8995d4e5fd60e9c83953476d4fb24f7f3", "submitted_at": "2025-10-16T12:30:05Z", "updated_at": "2025-10-16T12:30:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.14621v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2510.14621", "identifiers": {"arxiv": "2510.14621"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2510.14621"}], "name": "ColorBench: Benchmarking Mobile Agents with Graph-Structured Framework for Complex Long-Horizon Tasks", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "ColorBench represents mobile tasks as graphs of states collected from real devices, providing a quasi-dynamic environment with multiple valid paths and error paths. It evaluates single-app and cross-app tasks using completion, subtask progress, and atomic diagnostics.", "topics": ["evaluation", "mobile-navigation", "long-horizon", "state-tracking"], "updated_at": "2025-10-16", "url": "https://arxiv.org/abs/2510.14621"}
{"aliases": [], "authors": ["Mathieu Andreux", "Märt Bakler", "Yanael Barbier", "Hamza Benchekroun", "Emilien Biré", "Antoine Bonnet", "Riaz Bordie", "Nathan Bout", "Matthias Brunel", "Aleix Cambray", "Pierre-Louis Cedoz", "Antoine Chassang", "Gautier Cloix", "Ethan Connelly", "Alexandra Constantinou", "Ramzi De Coster", "Hubert de la Jonquiere", "Aurélien Delfosse", "Maxime Delpit", "Alexis Deprez", "Augustin Derupti", "Mathieu Diaz", "Shannon D'Souza", "Julie Dujardin", "Abai Edmund", "Michael Eickenberg", "Armand Fatalot", "Wissem Felissi", "Isaac Herring", "Xavier Koegler", "Erwan Le Jumeau de Kergaradec", "Aurélien Lac", "Maxime Langevin", "Corentin Lauverjat", "Antonio Loison", "Avshalom Manevich", "Axel Moyal", "Axel Nguyen Kerbel", "Marinela Parovic", "Julien Revelle", "Guillaume Richard", "Mats Richter", "Ronan Riochet", "María Santos", "Romain Savidan", "Laurent Sifre", "Maxime Theillard", "Marc Thibault", "Ivan Valentini", "Tony Wu", "Laura Yie", "Kai Yuan", "Jevgenij Zubovskij"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "de25494af1bc2ceaa760a21da51803aaf89d9d2d1961887cfd0023fb87f28f63", "submitted_at": "2025-10-22T18:21:52Z", "updated_at": "2025-10-24T11:52:29Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.19949v2"}}, "environments": ["web", "desktop", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.19949", "identifiers": {"arxiv": "2510.19949"}, "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/2510.19949"}], "name": "Surfer 2: The Next Generation of Cross-Platform Computer Use Agents", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-10-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Surfer2 combines hierarchical context management with separate planning and execution, self-verification, and recovery for visual computer use. It evaluates the framework across web, desktop, and Android tasks without task-specific fine-tuning.", "topics": ["planning", "memory", "long-horizon", "robustness"], "updated_at": "2025-10-24", "url": "https://arxiv.org/abs/2510.19949"}
{"aliases": [], "authors": ["Weihua Cheng", "Junming Liu", "Yifei Sun", "Botian Shi", "Yirong Chen", "Ding Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d80312fbd8f57cdc53cc10e03f2c68e2152d06945e5089967dceda50c02b9d4b", "submitted_at": "2025-10-28T08:19:58Z", "updated_at": "2026-04-14T04:34:20Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2510.24168v3"}}, "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", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.24168", "identifiers": {"arxiv": "2510.24168"}, "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/2510.24168"}], "name": "MGA: Memory-Driven GUI Agent for Observation-Centric Interaction", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2025-10-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MGA, which separates task-agnostic screen-state reading from decision making and connects steps through structured state memory. Validated state deltas replace full trajectory replay, evaluated on OSWorld and application tasks.", "topics": ["memory", "state-tracking", "long-horizon"], "updated_at": "2026-04-14", "url": "https://arxiv.org/abs/2510.24168"}
{"aliases": [], "authors": ["Yuxiang Zhou", "Jichang Li", "Yanhao Zhang", "Haonan Lu", "Guanbin Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ede08df1c10d8eeede43af75d26608b4917dd83becde23f28cd3adf335f79736", "submitted_at": "2025-11-15T15:22:42Z", "updated_at": "2026-03-12T08:45:07Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2511.12254v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2511.12254", "identifiers": {"arxiv": "2511.12254"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2511.12254"}], "name": "Mobile-Agent-RAG: Driving Smart Multi-Agent Coordination with Contextual Knowledge Empowerment for Long-Horizon Mobile Automation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-11-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses separate retrieval systems for high-level mobile-task planning and precise interface operations, with a benchmark for extended cross-app tasks.", "topics": ["mobile-navigation", "memory", "long-horizon"], "updated_at": "2026-03-12", "url": "https://arxiv.org/abs/2511.12254"}
{"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": ["Shizuo Tian", "Hao Wen", "Yuxuan Chen", "Jiacheng Liu", "Shanhui Zhao", "Guohong Liu", "Ju Ren", "Yunxin Liu", "Yuanchun Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "72b2db237a52ae31eaba58183a2777e79379aa8e849d18de5d8633daf27ccbf9", "submitted_at": "2025-12-11T07:37:38Z", "updated_at": "2026-05-08T09:40:02Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2512.10371v2"}}, "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", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.10371", "identifiers": {"arxiv": "2512.10371"}, "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/2512.10371"}], "name": "AgentProg: Empowering Long-Horizon GUI Agents with Program-Guided Context Management", "notes": "Reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-12-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AgentProg, organizing mobile-agent interaction history as a program with variables and control flow to decide what context to retain. A global belief state handles partial observability, with evaluation on AndroidWorld and extended long-horizon tasks.", "topics": ["interaction-history", "memory", "long-horizon"], "updated_at": "2026-05-08", "url": "https://arxiv.org/abs/2512.10371"}
{"aliases": [], "authors": ["Quyu Kong", "Xu Zhang", "Zhenyu Yang", "Nolan Gao", "Chen Liu", "Panrong Tong", "Chenglin Cai", "Hanzhang Zhou", "Jianan Zhang", "Liangyu Chen", "Zhidan Liu", "Steven Hoi", "Yue Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cf3258fd2286b0ee6babb69b9bfc459ecf40536eec498a3af3403d2971791f3d", "submitted_at": "2025-12-22T14:31:28Z", "updated_at": "2025-12-30T07:15:15Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2512.19432v3"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}], "id": "arxiv:2512.19432", "identifiers": {"arxiv": "2512.19432"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.19432"}], "name": "MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive and MCP-Augmented Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces long mobile workflows that combine agent-user interaction, cross-application tasks, and MCP-based tool access.", "topics": ["evaluation", "tool-integration", "long-horizon"], "updated_at": "2025-12-30", "url": "https://arxiv.org/abs/2512.19432"}
{"aliases": [], "authors": ["Yue Cao", "Yingyao Wang", "Pi Bu", "Jingxuan Xing", "Wei Jiang", "Zekun Zhu", "Junpeng Ma", "Sashuai Zhou", "Tong Lu", "Jun Song", "Yu Cheng", "Yuning Jiang", "Bo Zheng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "597935923ce9e59720771f3d92e921f2842f908ad81b3ab31a87ed9d9f5c5774", "submitted_at": "2025-12-24T17:40:42Z", "updated_at": "2025-12-24T17:40:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.21302v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2512.21302", "identifiers": {"arxiv": "2512.21302"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2512.21302"}], "name": "AndroidLens: Long-latency Evaluation with Nested Sub-targets for Android GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "AndroidLens provides 571 extended mobile tasks in Chinese and English, with multiple valid paths and preserved environmental anomalies. Its dynamic evaluation uses nested milestones and Average Task Progress to measure partial completion of long workflows.", "topics": ["evaluation", "long-horizon", "mobile-navigation", "robustness"], "updated_at": "2025-12-24", "url": "https://arxiv.org/abs/2512.21302"}
{"aliases": [], "authors": ["Jihong Wang", "Jiamu Zhou", "Weiming Zhang", "Teng Wang", "Weiwen Liu", "Zhuosheng Zhang", "Xingyu Lou", "Weinan Zhang", "Huarong Deng", "Jun Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a2afe8d0d8d83f91003b7885635e098e19f69024306c43c3edb23c110e553f63", "submitted_at": "2026-01-12T07:08:42Z", "updated_at": "2026-04-23T09:34:51Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.07262v3"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.07262", "identifiers": {"arxiv": "2601.07262"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.07262"}], "name": "ColorBrowserAgent: Complex Long-Horizon Browser Agent with Adaptive Knowledge Evolution", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "ColorBrowserAgent converts human feedback into reusable website knowledge and uses that knowledge to guide progressive context summaries. It evaluates the resulting memory and adaptation mechanisms on long browser workflows.", "topics": ["memory", "human-in-the-loop", "long-horizon", "web-navigation"], "updated_at": "2026-04-23", "url": "https://arxiv.org/abs/2601.07262"}
{"aliases": [], "authors": ["Taofeng Xue", "Chong Peng", "Mianqiu Huang", "Linsen Guo", "Tiancheng Han", "Haozhe Wang", "Jianing Wang", "Xiaocheng Zhang", "Xin Yang", "Dengchang Zhao", "Jinrui Ding", "Xiandi Ma", "Yuchen Xie", "Peng Pei", "Xunliang Cai", "Xipeng Qiu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "38b2798ce9ebd69a46202f0fa8ceca859394acda4f5173b5e30248914c8c35ac", "submitted_at": "2026-01-22T11:36:43Z", "updated_at": "2026-01-23T08:52:00Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.15876v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.15876", "identifiers": {"arxiv": "2601.15876"}, "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/2601.15876"}], "name": "EvoCUA: Evolving Computer Use Agents via Learning from Scalable Synthetic Experience", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "EvoCUA combines task synthesis with executable validators, asynchronous sandbox rollouts and iterative policy updates. It turns successful routines and analyzed failures into supervision and evaluates the evolving agent on OSWorld.", "topics": ["data-generation", "sandbox", "long-horizon", "self-reflection"], "updated_at": "2026-01-23", "url": "https://arxiv.org/abs/2601.15876"}
{"aliases": [], "authors": ["Jing Wu", "Wenjie Ai", "Daphne Barretto", "Yiye Chen", "Qingyu Chen", "Yuhang He", "Pranit Chawla", "Nicholas Gydé", "Yanan Jian", "Vibhav Vineet"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0139f7a1402d8dbfdecf65cc53c032495b2084f24cc38ff43841642c9102f29c", "submitted_at": "2026-01-28T14:35:23Z", "updated_at": "2026-08-26T22:31:44Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2601.20650v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.20650", "identifiers": {"arxiv": "2601.20650"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.20650"}], "name": "OS-Marathon: Benchmarking Computer-Use Agents on Vast-Horizon, Repetitive Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-01-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces OS-Marathon to evaluate long repetitive computer-use workflows whose execution grows with the amount of input data. It studies failure accumulation under task decomposition and GraphDemo, which adapts agents using a demonstration of the recurring workflow.", "topics": ["evaluation", "long-horizon", "human-in-the-loop"], "updated_at": "2026-08-26", "url": "https://arxiv.org/abs/2601.20650"}
{"aliases": [], "authors": ["Ziyu Lu", "Tengjin Weng", "Yiying Yang", "Yuhang Zhao", "Xinxin Huang", "Wenhao Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fa36a9fbd0b189717b2a5311f06b6127d54f2eca09961cedd2a0b9d2a78b1b8b", "submitted_at": "2026-01-29T07:22:50Z", "updated_at": "2026-01-29T07:22:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.21352v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.21352", "identifiers": {"arxiv": "2601.21352"}, "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/2601.21352"}], "name": "BEAP-Agent: Backtrackable Execution and Adaptive Planning 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": "2026-01-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "BEAP-Agent treats GUI execution as depth-first search with coordinated planning, execution and task-state tracking. It supports backtracking across multiple levels of earlier states and evaluates recovery on OSWorld.", "topics": ["planning", "state-tracking", "robustness", "long-horizon"], "updated_at": "2026-01-29", "url": "https://arxiv.org/abs/2601.21352"}
{"aliases": [], "authors": ["Hongze Mi", "Yibo Feng", "WenJie Lu", "Song Cao", "Jinyuan Li", "Yanming Li", "Xuelin Zhang", "Haotian Luo", "Songyang Peng", "He Cui", "Tengfei Tian", "Jun Fang", "Hua Chai", "Naiqiang Tan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c0b3dbaacdf640a1314a40688133f7386c0eb8b901c43393ad21ec887814a606", "submitted_at": "2026-01-30T04:01:21Z", "updated_at": "2026-01-30T04:01:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.22528v1"}}, "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:2601.22528", "identifiers": {"arxiv": "2601.22528"}, "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/2601.22528"}], "name": "Darwinian Memory: A Training-Free Self-Regulating Memory System for GUI Agent Evolution", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Darwinian Memory splits GUI trajectories into reusable units and updates the memory by retaining useful paths and pruning poor ones. It evaluates this training-free memory mechanism on long multi-application tasks.", "topics": ["memory", "training-free", "cross-app", "long-horizon"], "updated_at": "2026-01-30", "url": "https://arxiv.org/abs/2601.22528"}
{"aliases": [], "authors": ["Zexue He", "Yu Wang", "Churan Zhi", "Yuanzhe Hu", "Tzu-Ping Chen", "Lang Yin", "Ze Chen", "Tong Arthur Wu", "Siru Ouyang", "Zihan Wang", "Jiaxin Pei", "Julian McAuley", "Yejin Choi", "Alex Pentland"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2317a94d54b4458a1b7c5e92754aeaefcbf270aabfa0bd8ea333c673cd4596b3", "submitted_at": "2026-02-18T09:49:14Z", "updated_at": "2026-09-17T22:55:03Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.16313v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}], "id": "arxiv:2602.16313", "identifiers": {"arxiv": "2602.16313"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:12Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.16313"}], "name": "MemoryArena: Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-02-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MemoryArena links tasks across sessions so that agents must distill earlier interaction experience and use it in later decisions. Its evaluation includes web navigation alongside planning, search and formal reasoning, testing memory through subsequent actions rather than recall alone.", "topics": ["memory", "interaction-history", "planning", "evaluation", "long-horizon"], "updated_at": "2026-09-17", "url": "https://arxiv.org/abs/2602.16313"}
{"aliases": [], "authors": ["Yibo Shi", "Jungang Li", "Linghao Zhang", "Zihao Dongfang", "Biao Wu", "Sicheng Tao", "Yibo Yan", "Chenxi Qin", "Weiting Liu", "Zhixin Lin", "Hanqian Li", "Yu Huang", "Song Dai", "Yonghua Hei", "Yue Ding", "Xiang Li", "Shikang Wang", "Chengdong Xu", "Jingqi Liu", "Xueying Ma", "Zhiwen Zheng", "Xiaofei Zhang", "Bincheng Wang", "Nichen Yang", "Jie Wu", "Lihua Tian", "Chen Li", "Xuming Hu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0ce6da91e59ac57fa24ff991ca023b84448c54d477cacb2d6fded9bf527dd29d", "submitted_at": "2026-03-19T02:45:21Z", "updated_at": "2026-03-19T02:45:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.18429v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2603.18429", "identifiers": {"arxiv": "2603.18429"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.18429"}], "name": "AndroTMem: From Interaction Trajectories to Anchored Memory in Long-Horizon GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-03-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AndroTMem-Bench for Android tasks whose later actions depend on earlier UI states. Anchored State Memory retains causally linked intermediate states for subgoal-specific retrieval and evaluates them against full-history and summary-based memory.", "topics": ["memory", "long-horizon", "evaluation"], "updated_at": "2026-03-19", "url": "https://arxiv.org/abs/2603.18429"}
{"aliases": [], "authors": ["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": ["Akash Ghosh", "Tajamul Ashraf", "Rishu Kumar Singh", "Numan Saeed", "Sriparna Saha", "Xiuying Chen", "Salman Khan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8fb6a294616e1be46c5da05f60cda7d43c05d8a1ce71116c5e30cb94507cd56f", "submitted_at": "2026-03-25T10:25:48Z", "updated_at": "2026-03-25T10:25:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.24157v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:16Z", "role": "primary", "sha256": "9289bc047b59e93f75703b72f5e92d9277e115238427519c343df550220e351b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.19870%2C2602.22769%2C2602.23166%2C2602.23320%2C2603.00198%2C2603.00842%2C2603.00991%2C2603.01050%2C2603.01145%2C2603.02236%2C2603.02795%2C2603.03278%2C2603.04549%2C2603.06713%2C2603.07106%2C2603.07670%2C2603.07853%2C2603.08706%2C2603.09716%2C2603.10002%2C2603.10664%2C2603.11139%2C2603.11320%2C2603.12483%2C2603.13026%2C2603.13247%2C2603.15714%2C2603.16065%2C2603.16289%2C2603.16839%2C2603.17216%2C2603.17378%2C2603.19313%2C2603.19685%2C2603.21489%2C2603.22527%2C2603.23638%2C2603.24157%2C2603.24639%2C2603.27646%2C2603.28069%2C2603.28407%2C2603.29418%2C2603.29847%2C2604.00528%2C2604.00886%2C2604.01194%2C2604.01438%2C2604.01560%2C2604.02985&max_results=50"}], "id": "arxiv:2603.24157", "identifiers": {"arxiv": "2603.24157"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:16Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2603.24157"}], "name": "CarePilot: A Multi-Agent Framework for Long-Horizon Computer Task Automation in Healthcare", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-03-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "CareFlow benchmarks long healthcare-software workflows, while CarePilot coordinates action proposals, execution checks and corrective feedback using short- and long-term memory. Its actor consumes visual interfaces and system state to select semantic actions.", "topics": ["automation", "planning", "memory", "evaluation", "long-horizon", "workflow"], "updated_at": "2026-03-25", "url": "https://arxiv.org/abs/2603.24157"}
{"aliases": [], "authors": ["Yuwen Zhai", "Runze Li", "Liang Wang", "Nian Shi", "Liwu Xu", "Wei Zhang", "Ran Lin", "Bo Xu", "Benlei Cui"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "856e9920aea7d821061b3258b31aa32fd37b5579d628f5507f45ba83f237b77b", "submitted_at": "2026-04-06T03:58:43Z", "updated_at": "2026-04-06T03:58:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.04399v1"}}, "environments": ["web", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.04399", "identifiers": {"arxiv": "2604.04399"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.04399"}], "name": "GUIDE: Interpretable GUI Agent Evaluation via Hierarchical Diagnosis", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUIDE evaluates GUI trajectories by segmenting them into subtasks, diagnosing each segment, and aggregating structured findings into a task verdict. It tests diagnostic judgments on e-commerce, web-agent and Android-control traces rather than relying on one holistic trajectory score.", "topics": ["evaluation", "long-horizon", "state-tracking"], "updated_at": "2026-04-06", "url": "https://arxiv.org/abs/2604.04399"}
{"aliases": [], "authors": ["Pranjal Aggarwal", "Graham Neubig", "Sean Welleck"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fdd9cb6daad8832261ebe861fc1cb74429b819305e14ecfeb9698616013e6ac8", "submitted_at": "2026-04-07T17:38:15Z", "updated_at": "2026-04-07T17:38:15Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.06126v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2604.06126", "identifiers": {"arxiv": "2604.06126"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.06126"}], "name": "Gym-Anything: Turn any Software into an Agent Environment", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Gym-Anything, where a setup agent configures software environments and an independent auditor checks the resulting evidence. CUA-World supplies long-horizon tasks across professional applications, with experiments on trajectory distillation and test-time auditing.", "topics": ["data-generation", "sandbox", "long-horizon", "evaluation"], "updated_at": "2026-04-07", "url": "https://arxiv.org/abs/2604.06126"}
{"aliases": [], "authors": ["Corby Rosset", "Pratyusha Sharma", "Andrew Zhao", "Miguel Gonzalez-Fernandez", "Ahmed Awadallah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b452684b305c4f20ded686289b448fbe4f08bb3b34f515a43e2ce5d6dc4ef506", "submitted_at": "2026-04-05T06:26:05Z", "updated_at": "2026-04-05T06:26:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.06240v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.06240", "identifiers": {"arxiv": "2604.06240"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.06240"}], "name": "The Art of Building Verifiers for Computer Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Describes a web-trajectory verifier built around structured rubrics, separate process and outcome judgments, and explicit treatment of uncontrollable failures. CUAVerifierBench supplies human labels for evaluating these choices and screenshot-based context management.", "topics": ["reward-models", "evaluation", "long-horizon"], "updated_at": "2026-04-05", "url": "https://arxiv.org/abs/2604.06240"}
{"aliases": [], "authors": ["Zhengxi Lu", "Fei Tang", "Guangyi Liu", "Kaitao Song", "Xu Tan", "Jin Ma", "Wenqi Zhang", "Weiming Lu", "Jun Xiao", "Yueting Zhuang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d6e81b448c9c22e2fd25dce519db3fdc0585b0265b94d448751d7a37ed910dfa", "submitted_at": "2026-04-15T12:55:46Z", "updated_at": "2026-04-15T12:55:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.13822v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:00Z", "role": "primary", "sha256": "4d09ebfb093e43a4c44adef86c8fa8374f90d907fc33bd8e12a1315e62d5a583", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2603.12823%2C2603.14707%2C2603.15039%2C2603.17357%2C2603.18429%2C2603.19191%2C2603.24440%2C2603.25226%2C2603.25864%2C2603.26211%2C2603.26648%2C2603.29318%2C2604.01535%2C2604.02947%2C2604.04399%2C2604.05477%2C2604.06126%2C2604.06182%2C2604.06367%2C2604.07429%2C2604.08455%2C2604.08516%2C2604.08523%2C2604.09155%2C2604.09574%2C2604.09587%2C2604.09815%2C2604.09937%2C2604.11259%2C2604.11784%2C2604.12284%2C2604.13019%2C2604.13822%2C2604.14113%2C2604.14262%2C2604.14956%2C2604.15093%2C2604.15309%2C2604.17817%2C2604.17821%2C2604.18860%2C2604.24964%2C2604.25380%2C2604.25562%2C2604.27151%2C2604.27253%2C2604.27776%2C2605.00642%2C2605.06393%2C2605.06664&max_results=50"}], "id": "arxiv:2604.13822", "identifiers": {"arxiv": "2604.13822"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:00Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.13822"}], "name": "UI-Copilot: Advancing Long-Horizon GUI Automation via Tool-Integrated Policy Optimization", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-Copilot separates GUI execution from on-demand memory retrieval and numerical assistance. Its training procedure treats tool selection and multi-turn task execution as distinct optimization targets while keeping persistent observations outside transient context.", "topics": ["tool-integration", "memory", "planning", "long-horizon", "mobile-navigation"], "updated_at": "2026-04-15", "url": "https://arxiv.org/abs/2604.13822"}
{"aliases": [], "authors": ["Lingfeng Zhang", "Yongan Sun", "Jinpeng Hu", "Hui Ma", "Yang Ying", "Kuien Liu", "Zenglin Shi", "Meng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "123577fbb4431f22bc18d2d32187744268f9d63b3149e9e1775fb44673771feb", "submitted_at": "2026-04-20T05:19:49Z", "updated_at": "2026-04-21T03:15:01Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.17821v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.17821", "identifiers": {"arxiv": "2604.17821"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.17821"}], "name": "WebUncertainty: Dual-Level Uncertainty Driven Planning and Reasoning For Autonomous Web Agent", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-04-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebUncertainty adapts planning mode to task uncertainty and guides action-level tree search with confidence-derived uncertainty estimates. Its evaluations examine autonomous execution of dynamic, long-horizon web tasks on WebArena and WebVoyager.", "topics": ["planning", "web-navigation", "long-horizon", "robustness"], "updated_at": "2026-04-21", "url": "https://arxiv.org/abs/2604.17821"}
{"aliases": [], "authors": ["Lawrence Keunho Jang", "Jing Yu Koh", "Daniel Fried", "Ruslan Salakhutdinov"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "db70a5736d04d26f10afe3d9a0cfb08955f65f69480a0345ff13b30b2e83df23", "submitted_at": "2026-04-27T20:05:41Z", "updated_at": "2026-04-27T20:05:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.24964v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.24964", "identifiers": {"arxiv": "2604.24964"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:39Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.24964"}], "name": "Odysseys: Benchmarking Web Agents on Realistic Long Horizon Tasks", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-04-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Odysseys, a live-web benchmark of long, multi-site tasks derived from browsing sessions. Graded task rubrics and a rubric-score-per-step measure evaluate both partial progress and trajectory efficiency.", "topics": ["evaluation", "long-horizon", "efficiency"], "updated_at": "2026-04-27", "url": "https://arxiv.org/abs/2604.24964"}
{"aliases": [], "authors": ["Jinbiao Wei", "Kangqi Ni", "Yilun Zhao", "Guo Gan", "Arman Cohan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "68183be37dbd78314aca35f7f1420e5f2f8da8d8c0042a258b45ae03c90606cf", "submitted_at": "2026-04-29T19:59:36Z", "updated_at": "2026-04-29T19:59:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.27151v1"}}, "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:2604.27151", "identifiers": {"arxiv": "2604.27151"}, "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/2604.27151"}], "name": "Step-level Optimization for Efficient 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": "2026-04-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Proposes a step-level model cascade that runs a smaller computer-use policy until learned monitors flag stalled progress or semantic drift. Separate history-based stuck detection and milestone checks determine when to escalate for recovery or verification without retraining the larger model.", "topics": ["efficiency", "robustness", "state-tracking", "interaction-history", "long-horizon"], "updated_at": "2026-04-29", "url": "https://arxiv.org/abs/2604.27151"}
{"aliases": [], "authors": ["Guibin Zhang", "Yaohui Ling", "Fanci Meng", "Kun Wang", "Shuicheng Yan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "729ff6b37da62bb8a9a911865554520e3b29219d9735e087c3186705428d7b05", "submitted_at": "2026-05-10T04:31:23Z", "updated_at": "2026-05-10T04:31:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.09317v1"}}, "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:2605.09317", "identifiers": {"arxiv": "2605.09317"}, "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/2605.09317"}], "name": "Mem-W: Latent Memory-Native GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2026-05-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Mem-W, which compresses previous trajectories and current-session segments into latent memory tokens used directly by the GUI policy. Self-distillation and outcome supervision train this memory representation for web and mobile navigation.", "topics": ["memory", "long-horizon", "imitation-learning"], "updated_at": "2026-05-10", "url": "https://arxiv.org/abs/2605.09317"}
{"aliases": [], "authors": ["Zerui Qin", "Sheng Yue", "Xingyuan Hua", "Yongjian Fu", "Ju Ren"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "388eac2721929fc8211f40cc4edde61be0b6e6145fa60fbaa444ec80822fcc36", "submitted_at": "2026-05-12T15:48:44Z", "updated_at": "2026-05-12T15:48:44Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.12294v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:39Z", "role": "primary", "sha256": "27c33576ed43c7bc63d4b290f6558553bd1d3aa112a7358138f4045e1bc14918", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=300&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.12294", "identifiers": {"arxiv": "2605.12294"}, "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/2605.12294"}], "name": "Executable Agentic Memory for GUI Agent", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Executable Agentic Memory stores GUI routines in a knowledge graph built through state-aware exploration and action-group mining. A learned value model guides tree search over this executable memory for retrieval-based planning, with AndroidWorld and efficiency evaluation.", "topics": ["memory", "planning", "long-horizon", "efficiency"], "updated_at": "2026-05-12", "url": "https://arxiv.org/abs/2605.12294"}
{"aliases": [], "authors": ["Kean Shi", "Zihang Li", "Tianyi Ma", "Zengji Tu", "Jialong Wu", "Xinbo Xu", "Qingyao Yang", "Ruoyu Wu", "Weichu Xie", "Ming Wu", "Jason Zeng", "Michael Heinrich", "Elvis Zhang", "Liang Chen", "Kuan Li", "Baobao Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c4ee2439dd31a375b6f64cbc7f1d0ff75c87e1346b1627a53db145ae42682f8a", "submitted_at": "2026-05-15T09:35:15Z", "updated_at": "2026-05-24T16:35:56Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.15777v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "topics"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:59Z", "role": "primary", "sha256": "8cfab6fca80090adb47c4ef634f89eedd49a6204dbdc2b5b8e0d6f04c32db289", "supports": ["summary", "environments", "scope", "notes"], "url": "https://raw.githubusercontent.com/UniPat-AI/SaaS-Bench/HEAD/README.md"}], "id": "arxiv:2605.15777", "identifiers": {"arxiv": "2605.15777"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.15777"}], "name": "SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows?", "notes": "The complete retained official abstract and cached official README were reviewed. The README establishes browser execution against running SaaS applications; no repository code was executed and benchmark results were not independently reproduced.", "published_at": "2026-05-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "SaaS-Bench evaluates browser agents on professional workflows across self-hosted applications. Weighted checkpoints measure task completion and partial progress, with application-state checks and configured content judges documented in the official harness.", "topics": ["web-navigation", "evaluation", "long-horizon", "workflow", "cross-app", "state-tracking"], "updated_at": "2026-05-24", "url": "https://arxiv.org/abs/2605.15777"}
{"aliases": [], "authors": ["Ziyun Zeng", "Hang Hua", "Bocheng Zou", "Mu Cai", "Rogerio Feris", "Jiebo Luo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d96c8844274f79adea167169a0853e3a7cd82fee4a8a666913f97898a0836e7c", "submitted_at": "2026-05-18T16:57:36Z", "updated_at": "2026-05-18T16:57:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.18652v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.18652", "identifiers": {"arxiv": "2605.18652"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.18652"}], "name": "MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-18", "relations": [], "review_status": "curated", "scope": "core", "summary": "MementoGUI learns a controller for selecting, compressing and retrieving textual and region-level visual memories while leaving the GUI-agent backbone fixed. It also converts computer-use traces into controller training data and evaluates semantic actions, progress and memory consistency on long-horizon tasks.", "topics": ["memory", "interaction-history", "long-horizon", "data-generation", "evaluation"], "updated_at": "2026-05-18", "url": "https://arxiv.org/abs/2605.18652"}
{"aliases": [], "authors": ["Yusong Lin", "Xinyuan Liang", "Haiyang Wang", "Qipeng Gu", "Siqi Cheng", "Jiangui Chen", "Shuzhe Wu", "Feiyang Pan", "Lue Fan", "Sanyuan Zhao", "Dandan Tu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "27999553928305cfa4223011feb3f9c2decda8a861e5800865537a81c7c08832", "submitted_at": "2026-05-25T17:50:04Z", "updated_at": "2026-05-25T17:50:04Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.26086v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}], "id": "arxiv:2605.26086", "identifiers": {"arxiv": "2605.26086"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:23Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.26086"}], "name": "Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-05-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Claw-Anything benchmarks persistent personal assistance across activity histories, backend services and integrated GUI/CLI interaction. Its simulated event histories introduce irrelevant or conflicting context, and a separate generation pipeline supplies training environments.", "topics": ["evaluation", "interaction-history", "long-horizon", "robustness", "data-generation", "workflow"], "updated_at": "2026-05-25", "url": "https://arxiv.org/abs/2605.26086"}
{"aliases": [], "authors": ["Yifan Sui", "Xin Huang", "Hongbing Li", "Fang Xu", "Jiahe Lv", "Haolong Yan", "Yeqing Shen", "Litao Liu", "Zhimin Fan", "Ziyang Meng", "Jia Wang", "Junbo Qi", "Kaijun Tan", "Zheng Ge", "Xiangyu Zhang", "Daxin Jiang", "Osamu Yoshie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0c89b8b5ef5aa17fe747201285104544648449804cc2229a77577d217763f497", "submitted_at": "2026-05-26T23:19:42Z", "updated_at": "2026-05-26T23:19:42Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.27761v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2605.27761", "identifiers": {"arxiv": "2605.27761"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2605.27761"}], "name": "AndroidDaily: A Verifiable Benchmark for Mobile GUI Agents on Real-World Closed-Source Applications", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-05-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "AndroidDaily evaluates mobile GUI agents on daily workflows in closed-source Android apps. Its GRADE evaluator follows visual trajectories against operational obligations, output-quality requirements and negative constraints, producing process-level diagnostics without hidden app state.", "topics": ["mobile-navigation", "evaluation", "long-horizon", "robustness"], "updated_at": "2026-05-26", "url": "https://arxiv.org/abs/2605.27761"}
{"aliases": [], "authors": ["Jing Yu Koh", "Ruslan Salakhutdinov", "Daniel Fried"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "acc6e8af676b7de5393469d379746e674ccd662f570399622d0d5b1b0e882939", "submitted_at": "2026-06-01T01:29:36Z", "updated_at": "2026-06-01T01:29:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.01533v1"}}, "environments": ["desktop", "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:2606.01533", "identifiers": {"arxiv": "2606.01533"}, "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/2606.01533"}], "name": "Multi-Agent Computer Use", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "MACU represents computer-use objectives as a dependency graph whose manager dispatches parallel subagents and revises tasks as new information arrives. Desktop and web experiments examine coordination, information handoff and task-completion time.", "topics": ["planning", "long-horizon", "workflow", "state-tracking", "efficiency"], "updated_at": "2026-06-01", "url": "https://arxiv.org/abs/2606.01533"}
{"aliases": [], "authors": ["Wenkai Wang", "Tao Xiong", "Jingchen Ni", "Yunpeng Bao", "Xiyun Li", "Tianqi Liu", "Hongcan Guo", "Zilong Huang", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf0fcafeb2d40e489ec80784a3541114b02c6a9077a330265d5c49775eca4582", "submitted_at": "2026-06-02T03:42:34Z", "updated_at": "2026-06-02T03:42:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.03103v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:23Z", "role": "primary", "sha256": "9a71e95803ce58c462c84e2b45a891555e31d1fb409e643bb3ca4675a44c96e6", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2605.12233%2C2605.12493%2C2605.12978%2C2605.13119%2C2605.14133%2C2605.16909%2C2605.17260%2C2605.17324%2C2605.17448%2C2605.17656%2C2605.17721%2C2605.19447%2C2605.19587%2C2605.20894%2C2605.22642%2C2605.23204%2C2605.25233%2C2605.25901%2C2605.26086%2C2605.26144%2C2605.26252%2C2605.26646%2C2605.27209%2C2605.27705%2C2605.27898%2C2605.28158%2C2605.28255%2C2605.29341%2C2605.29463%2C2605.29801%2C2605.30010%2C2605.30052%2C2605.30126%2C2605.30434%2C2605.30907%2C2605.30914%2C2605.31264%2C2605.31557%2C2606.00530%2C2606.00543%2C2606.00856%2C2606.01057%2C2606.01961%2C2606.02540%2C2606.03103%2C2606.03889%2C2606.03895%2C2606.04315%2C2606.04455%2C2606.05525&max_results=50"}], "id": "arxiv:2606.03103", "identifiers": {"arxiv": "2606.03103"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:23Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.03103"}], "name": "DeskCraft: Benchmarking Desktop Agents on Professional Workflows and Human-in-the-Loop Collaboration", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-06-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "DeskCraft evaluates professional desktop workflows and human-agent collaboration through mid-execution clarification, interruption and post-completion feedback. Its tasks cover creative applications and expose difficulties in long-horizon execution and proactive clarification.", "topics": ["evaluation", "human-in-the-loop", "long-horizon", "workflow"], "updated_at": "2026-06-02", "url": "https://arxiv.org/abs/2606.03103"}
{"aliases": [], "authors": ["Shubham Gaur", "Ian Lane"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a449bb2706d5d18cbd176dcfd2f14d814cf1c1a22a08e9210c99e0722e10ea93", "submitted_at": "2026-06-04T20:48:37Z", "updated_at": "2026-08-04T18:57:23Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.06708v2"}}, "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:2606.06708", "identifiers": {"arxiv": "2606.06708"}, "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/2606.06708"}], "name": "Signal-Driven Observation for Long-Horizon 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. The abstract presents an architecture proposal and open problems, without reporting an empirical evaluation.", "published_at": "2026-06-04", "relations": [], "review_status": "curated", "scope": "core", "summary": "Proposes Signal-Driven Observation: a separate call selects relevant DOM elements and selectors, and a detector requests new observations only after meaningful browser events or failures. The paper frames observation timing and compression as an architectural research problem for long-horizon web agents.", "topics": ["efficiency", "screen-understanding", "long-horizon"], "updated_at": "2026-08-04", "url": "https://arxiv.org/abs/2606.06708"}
{"aliases": [], "authors": ["Wanli Li", "Bowen Zhou", "Yunyao Yu", "Zhou Xu", "Yifan Yang", "Dongsheng Li", "Caihua Shan"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73b751efd63d98fc81166a1a80cb0c162d68bf1ec4ecca0a1a1a7717ad62e680", "submitted_at": "2026-06-08T12:39:23Z", "updated_at": "2026-07-06T14:25:02Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2606.09426v3"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.09426", "identifiers": {"arxiv": "2606.09426"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.09426"}], "name": "WeaveBench: A Long-Horizon, Real-World Benchmark for Computer-Use Agents with Hybrid Interfaces", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "WeaveBench requires agents to combine GUI, command-line and code operations within long tasks on an Ubuntu desktop. A trajectory-aware judge checks deliverables and interaction evidence, including shortcut behavior that outcome-only grading can miss.", "topics": ["evaluation", "long-horizon", "workflow", "tool-integration"], "updated_at": "2026-07-06", "url": "https://arxiv.org/abs/2606.09426"}
{"aliases": [], "authors": ["Liya Zhu", "Jingzhe Ding", "Jian Zhang", "Jianbo Xue", "Shihao Liang", "Ge Zhang", "Yi Zhu", "Duju Zeng", "Xiang Gao", "Qingshui Gu", "Mailun Gao", "Huimin Che", "Yan Zhao", "Peiheng Zhou", "Haojun Wang", "Chaobo Xian", "Lili Le", "Chi Wu", "Yiwei Liu", "Shengda Long", "Jiale Yang", "Fangzhi Xu", "Sijin Wu", "Haodong Duan", "Chao He", "Zhaojian Li", "Minchao Wang", "Huan Zhou", "Jiani Hou", "Chuqian Yu", "Weiran Shi", "Hongwan Gao", "Jiamin Chen", "Guanhong Chen", "Tingqin Luo", "Kaiyuan Zhang", "Zhixin Yao", "Qing Hua", "Yuhao Jiang", "Jin Chen", "Pu Chen", "Zhenyu Hu", "Xingyu Li", "Zhengxuan Jiang", "Meng Cao", "Tianfeng Long", "Haozhe Wang", "Mingzhang Wang", "Yichen Zhang", "Yiming Dai", "Chenchen Zhang", "Jiaying Wang", "Xinying Liu", "Xingzu Liu", "Lingling Zhang", "Xinjie Chen", "Yujia Qin", "Wangchunshu Zhou", "Zhiyong Wu", "Yang Liu", "Jiaheng Liu", "Lei Zhang", "Shen Yan", "Wenhao Huang", "Zaiyuan Wang", "Xiaolong Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "037a3197137d27318320a3fd23fb5e606a1aa748133e57213cd02fd10aad4c4d", "submitted_at": "2026-06-09T16:10:16Z", "updated_at": "2026-07-17T12:22:12Z", "version": 4, "versioned_url": "https://arxiv.org/abs/2606.11042v4"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.11042", "identifiers": {"arxiv": "2606.11042"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.11042"}], "name": "Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional Fields", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Workflow-GYM evaluates long GUI workflows in professional software rather than isolated interface actions. Its analysis examines failures such as omitted stages, accumulated errors and goal drift during specialized end-to-end work.", "topics": ["evaluation", "long-horizon", "workflow", "robustness"], "updated_at": "2026-07-17", "url": "https://arxiv.org/abs/2606.11042"}
{"aliases": [], "authors": ["Jaewoo Lee", "Zaid Khan", "Archiki Prasad", "Justin Chih-Yao Chen", "Supriyo Chakraborty", "Kartik Balasubramaniam", "Sambit Sahu", "Elias Stengel-Eskin", "Hyunji Lee", "Mohit Bansal"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a8ed399c8b489a80c5313a00232d2ecca0fbc8228b1f56745c02d4fedb6f6908", "submitted_at": "2026-06-09T16:39:10Z", "updated_at": "2026-06-09T16:39:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.11078v1"}}, "environments": ["web", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.11078", "identifiers": {"arxiv": "2606.11078"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.11078"}], "name": "A History-Aware Visually Grounded Critic for Computer Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "HiViG combines a compact history of completed actions with a visually grounded critic that checks proposed coordinates before execution. Its evaluations separate history-related planning failures from errors caused by targeting the wrong screen element.", "topics": ["interaction-history", "memory", "grounding", "reward-models", "long-horizon", "evaluation"], "updated_at": "2026-06-09", "url": "https://arxiv.org/abs/2606.11078"}
{"aliases": [], "authors": ["Anqi Zou", "Han Deng", "Chengyu Zhang", "Junquan Hu", "Yu Wang", "Yuxiang Xing", "Aokai Zhang", "Hanling Zhang", "Zhaoyang Liu", "Ben Fei", "Zhihui Wang", "Wanli Ouyang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7487f31097bc8de90dfbe836051d47f407c26e31a42bd751bedfd38cf4f1742c", "submitted_at": "2026-06-15T14:42:33Z", "updated_at": "2026-06-15T14:42:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.16802v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.16802", "identifiers": {"arxiv": "2606.16802"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.16802"}], "name": "LabOSBench: Benchmarking Computer Use Agents for Scientific Instrument Control", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "LabOSBench evaluates GUI agents on browser-based scientific instrument simulators with configurable tasks and execution-based checks. Its subtask and end-to-end tests examine feedback-driven adjustment and long workflows without controlling physical instruments.", "topics": ["evaluation", "science", "long-horizon"], "updated_at": "2026-06-15", "url": "https://arxiv.org/abs/2606.16802"}
{"aliases": [], "authors": ["Ziyan Jiang", "Li An", "Yujian Liu", "Jiabao Ji", "Qiucheng Wu", "Jacob Andreas", "Yang Zhang", "Shiyu Chang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2605cedb4359aa0cf1e694455903871ed334dd37b7b8e15edf31bac9d5e51b13", "submitted_at": "2026-06-16T19:57:07Z", "updated_at": "2026-06-16T19:57:07Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.18448v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.18448", "identifiers": {"arxiv": "2606.18448"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.18448"}], "name": "VISUALSKILL: Multimodal Skills for Computer-Use Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-06-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "VISUALSKILL organizes application-specific text and figures into hierarchical skills retrieved on demand through an MCP tool. Skills combine authored documentation with live UI exploration, and matched multimodal/text-only comparisons test their utility for interface identification and workflow-state verification.", "topics": ["memory", "tool-integration", "screen-understanding", "long-horizon"], "updated_at": "2026-06-16", "url": "https://arxiv.org/abs/2606.18448"}
{"aliases": [], "authors": ["Guangyi Liu", "Gao Wu", "Congxiao Liu", "Pengxiang Zhao", "Liang Liu", "Mading Li", "Qi Zhang", "Mengyan Wang", "Liang Guo", "Yong Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cbbb1bd57057e465ce9d991d3f6e2bb52112fc133a5846fc8cb6ce51eddc2121", "submitted_at": "2026-06-18T08:26:09Z", "updated_at": "2026-06-18T08:26:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.19926v1"}}, "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:2606.19926", "identifiers": {"arxiv": "2606.19926"}, "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/2606.19926"}], "name": "MemGUI-Agent: An End-to-End Long-Horizon Mobile GUI Agent with Proactive Context Management", "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": "MemGUI-Agent makes context management an action chosen by the same policy that controls the mobile UI. Structured history, UI-state and recent-step fields are trained from annotated trajectories to retain cross-app facts while limiting context growth.", "topics": ["memory", "interaction-history", "long-horizon", "imitation-learning", "data-generation"], "updated_at": "2026-06-18", "url": "https://arxiv.org/abs/2606.19926"}
{"aliases": [], "authors": ["Vincent Siu", "Manasi Sharma", "Dawn Song", "Daniel Yue Zhang", "Chenguang Wang", "Ying Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "69e668a9491a31e8d8849bab8cc63eede9878fff10ccb0ae0a211610e90819b3", "submitted_at": "2026-06-19T18:00:14Z", "updated_at": "2026-08-19T17:29:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.21654v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.21654", "identifiers": {"arxiv": "2606.21654"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.21654"}], "name": "ChainWorld: Composing Long-Horizon Desktop Workloads from Atomic OSWorld Tasks", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "ChainWorld composes compatible atomic OSWorld tasks into multi-objective desktop chains while preserving their evaluators. It compares giving the full sequence at once with revealing goals turn by turn, exposing different state and session-management failures.", "topics": ["evaluation", "long-horizon", "state-tracking", "workflow"], "updated_at": "2026-08-19", "url": "https://arxiv.org/abs/2606.21654"}
{"aliases": [], "authors": ["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": ["Mengqi Yuan", "Zilong Zhou", "Xinzhuang Xiong", "Weiming Wu", "Jiayang Sun", "Jiamin Song", "Kaiqian Cui", "Bowen Wang", "Haoyuan Wu", "Yitong Li", "Dunjie Lu", "Haikong Lu", "Qi Zhen", "Xinyuan Wang", "Jiaqi Deng", "Yuhao Yang", "Cheng Chen", "Boyuan Zheng", "Alex Su", "Xiao Yu", "Hao Zou", "Saaket Agashe", "Xing Han Lu", "Manpreet Kaur", "Zhengyang Qi", "Vincent Sunn Chen", "Frederic Sala", "Dayiheng Liu", "Junyang Lin", "Zhou Yu", "Yu Su", "Siva Reddy", "Xin Eric Wang", "Peng Qi", "Tianbao Xie", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "03035fcbc6cd3e6bf40f244870fb35694ba985ff7f809da66e2a376cbf99d66b", "submitted_at": "2026-06-28T17:59:17Z", "updated_at": "2026-07-13T10:30:17Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.29537v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.29537", "identifiers": {"arxiv": "2606.29537"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.29537"}], "name": "OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-28", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces extended computer-use workflows with realistic artifacts, changing information, hidden state, and partial-progress and safety assessments.", "topics": ["evaluation", "long-horizon", "workflow"], "updated_at": "2026-07-13", "url": "https://arxiv.org/abs/2606.29537"}
{"aliases": [], "authors": ["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": ["Yixiong Chen", "Xinyi Bai", "Alan Yuille"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0473ced7881ed2007aa49c2767860c33719d5bf1c798dfe4771a4061b13d655e", "submitted_at": "2026-07-12T07:05:31Z", "updated_at": "2026-07-12T07:05:31Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.10608v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:32Z", "role": "primary", "sha256": "5128028eea81e3ae80715cb038513012598c0c81a7cbfe8c0d4d6adb6c00e518", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2607.03726%2C2607.05120%2C2607.05465%2C2607.05743%2C2607.06008%2C2607.06595%2C2607.07403%2C2607.07744%2C2607.08252%2C2607.08964%2C2607.10455%2C2607.10608%2C2607.12406%2C2607.12756%2C2607.13085%2C2607.13594%2C2607.13884%2C2607.13988%2C2607.15899%2C2607.16352%2C2607.16636%2C2607.16900%2C2607.17044%2C2607.20064%2C2607.20121%2C2607.21522%2C2607.22738%2C2607.23722%2C2607.24720%2C2607.26300%2C2607.26769%2C2607.27881%2C2607.28229%2C2607.28272%2C2607.28430%2C2608.00548%2C2608.02163%2C2608.03223%2C2608.03451%2C2608.03499%2C2608.05013%2C2608.05430%2C2608.06811%2C2608.07873%2C2608.09268%2C2608.10357%2C2608.10669%2C2608.11274%2C2608.11434%2C2608.15089&max_results=50"}], "id": "arxiv:2607.10608", "identifiers": {"arxiv": "2607.10608"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.10608"}], "name": "The Compliance Trap: Diagnosing How AI Agents Consume Conflicting Memory", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-07-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "The Entry–Propagation–Recovery framework traces where conflicting memory first changes an agent action, how the effect propagates and whether recovery occurs. WebArena and the controlled MemTrapBench separate memory-consumption failures across these stages.", "topics": ["memory", "safety-security", "robustness", "evaluation", "long-horizon"], "updated_at": "2026-07-12", "url": "https://arxiv.org/abs/2607.10608"}
{"aliases": [], "authors": ["Wenyi Wu", "Sibo Zhu", "Kun Zhou", "Aayush Salvi", "Zixuan Song", "Biwei Huang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "40daa1b180a0c2eaba7b907a0f61b8dd8819cb3033b9da118162330d97cfecf7", "submitted_at": "2026-07-13T10:48:25Z", "updated_at": "2026-07-13T10:48:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.11388v1"}}, "environments": ["desktop"], "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:2607.11388", "identifiers": {"arxiv": "2607.11388"}, "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/2607.11388"}], "name": "StructAgent: Harness Long-horizon Digital Agents with Unified Causal Structure", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "StructAgent maintains compact task state and regulates long-horizon execution through verified transitions and progress checkpoints. Its workflow supports evidence-based completion and targeted recovery, with desktop computer-use evaluation and a separate Minecraft extension.", "topics": ["state-tracking", "planning", "long-horizon", "robustness"], "updated_at": "2026-07-13", "url": "https://arxiv.org/abs/2607.11388"}
{"aliases": [], "authors": ["Ruei-Che Chang", "Wenqian Xu", "Dingzeyu Li", "Bryan Wang", "Anhong Guo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2069acf693e0dced49d0e10e21bf9fe28b99f50f1b932837f16137e0d3170a80", "submitted_at": "2026-07-20T04:03:38Z", "updated_at": "2026-07-20T04:03:38Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.17527v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.17527", "identifiers": {"arxiv": "2607.17527"}, "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/2607.17527"}], "name": "Sidekick: Designing Communication for Effective Multitasking with Computer Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Sidekick, a communication interface that uses ambient status cues, multimodal action summaries, and visual and verbal explanations of reasoning to help people monitor computer-use agents. A user study examines progress awareness, error traceability, and multitasking across background and foreground interaction.", "topics": ["human-in-the-loop", "observability", "long-horizon"], "updated_at": "2026-07-20", "url": "https://arxiv.org/abs/2607.17527"}
{"aliases": [], "authors": ["Zhuohang Fan", "Beichen Zhang", "Yuanfa Li", "Changqiao Wu", "Wei Liu", "Jian Luan", "Weigang Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b85b453a688543477d31a940f9d1da2857b2cb7a8cb9eea91a39a3a2b8a1e1c7", "submitted_at": "2026-07-20T15:16:59Z", "updated_at": "2026-08-18T02:32:56Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.18046v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.18046", "identifiers": {"arxiv": "2607.18046"}, "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/2607.18046"}], "name": "SEE: Structure-aware Exploring & Exploiting for Long-horizon GUI Agent Trajectory 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": "2026-07-20", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces SEE, which first explores mobile interfaces to build a screen-and-element transition graph, then composes task trajectories through graph planning and controlled sampling. It studies coverage, loop avoidance, and transfer after training agents on the generated sequences.", "topics": ["data-generation", "long-horizon", "mobile-navigation"], "updated_at": "2026-08-18", "url": "https://arxiv.org/abs/2607.18046"}
{"aliases": [], "authors": ["Zedong Yu", "Qianxing Li", "Zhi Gao", "Liuyu Xiang", "Chenrui Shi", "Yang Liu", "Huiming Wu", "Yujie Wei", "Yuhao Fei", "Yubo Fu", "Zhaofeng He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "078ff9a56b769923d15a98581685a3e2f47aac7d898ea731999cabaec3918fa7", "submitted_at": "2026-07-17T12:48:03Z", "updated_at": "2026-07-17T12:48:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.22689v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.22689", "identifiers": {"arxiv": "2607.22689"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.22689"}], "name": "Beyond Sequential Interaction: Benchmarking Parallel Execution and Coordination for GUI Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-07-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces ParaGUIBench, with tasks and efficiency measurements for coordinated GUI agents operating separate desktop instances over a shared file system. Its ParaGUI planner-worker system decomposes tasks and dispatches concurrent subtasks for comparison with serial execution.", "topics": ["evaluation", "planning", "efficiency", "long-horizon"], "updated_at": "2026-07-17", "url": "https://arxiv.org/abs/2607.22689"}
{"aliases": [], "authors": ["Yan Yang", "Xiangru Jian", "Ziyang Luo", "Zirui Zhao", "Yutong Dai", "Ziji Shi", "Hanshu Yan", "Jun Hao Liew", "Silvio Savarese", "Junnan Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a010fbef09d8359550cf156bf0dcba873f25a0d6fb8f118d53bf6e23715cf203", "submitted_at": "2026-07-24T14:17:03Z", "updated_at": "2026-07-24T14:17:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.22798v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.22798", "identifiers": {"arxiv": "2607.22798"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2607.22798"}], "name": "StateAct: Program State, before Pixels, for Long-Horizon Computer-Use Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces StateAct, a code-first computer-use harness that manipulates program state directly and delegates screenshot-and-click subgoals to a GUI agent. An independent completion gate checks saved outputs, while fresh subagents handle subgoals to limit accumulated context.", "topics": ["tool-integration", "state-tracking", "long-horizon"], "updated_at": "2026-07-24", "url": "https://arxiv.org/abs/2607.22798"}
{"aliases": [], "authors": ["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": ["Mizanur Rahman", "Mohammed Saidul Islam", "Ridwan Mahbub", "Md Tahmid Rahman Laskar", "Shafiq Joty", "Enamul Hoque Prince"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e1089e57e0f0146f29a4a411c5c6f72da3c9f45be223fbbc5c9630315be6a680", "submitted_at": "2026-08-11T01:45:56Z", "updated_at": "2026-08-11T01:45:56Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.10366v1"}}, "environments": ["desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "kind", "topics"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:06:26Z", "role": "primary", "sha256": "2158cabc9a6211e0b75e8e3307f80ebc0d42c42fccc03d0143dcf574c2f663db", "supports": ["summary", "environments", "scope"], "url": "https://github.com/vis-nlp/DSAgentBench"}], "id": "arxiv:2608.10366", "identifiers": {"arxiv": "2608.10366"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:09Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.10366"}], "name": "DSAgentBench: Can Agents Automate End-to-End Data-Science Workflows in Real Computer Environments?", "notes": "The paper abstract and official project README were reviewed. The README documents an Ubuntu desktop and GUI action interface and includes LLM-assisted evaluators; this annotation does not assert that every evaluator is deterministic. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "DSAgentBench evaluates end-to-end data-science workflows spanning analysis, modeling and visualization inside a real desktop environment. Task evaluators check produced analytical and visual artifacts, while the released interface supports screenshot or accessibility observations and GUI actions.", "topics": ["evaluation", "workflow", "science", "long-horizon"], "updated_at": "2026-08-11", "url": "https://arxiv.org/abs/2608.10366"}
{"aliases": [], "authors": ["Zihan Ding", "Longxu Dou", "Qi Gao", "Xiangwu Guo", "Shengchao Hu", "Zilong Huang", "Zihang Jiang", "Lei Ke", "Mengcheng Lan", "Weixian Lei", "Hanxuan Li", "Honglin Li", "Xiyun Li", "Zaitang Li", "Leowei Liang", "Xin Luo", "Haozhe Ma", "Jiayi Mao", "Zhoujie Pan", "Can Qin", "Tianyuan Qu", "Weiqi Wang", "Wenkai Wang", "Yonglin Wang", "Yuxin Wang", "Chenxu Wu", "Yingchen Yu", "Chenyu Zhang", "Yuhao Zheng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "efe635792c696bca73c9a436be67d8528582fb239d4d4f8d3de531219b1f4f8e", "submitted_at": "2026-08-16T20:59:59Z", "updated_at": "2026-08-16T20:59:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.15930v1"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "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:2608.15930", "identifiers": {"arxiv": "2608.15930"}, "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/2608.15930"}], "name": "UI-Mate: Advancing Open-Weight Foundation GUI Agents with In-Context Demonstrations", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Combines environment-grounded training with demonstration-conditioned workflows, and introduces OSWorkerBench for extended office-task evaluation.", "topics": ["workflow", "data-generation", "long-horizon"], "updated_at": "2026-08-16", "url": "https://arxiv.org/abs/2608.15930"}
{"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": ["Jiaxuan Luo", "Zhanfeng Liao", "Jiayao Teng", "Yuan Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2ea33a82ab7543a64b60ee3024c6f3e39f88b4e3f06eff94f8c790c56cf76715", "submitted_at": "2026-08-23T20:04:21Z", "updated_at": "2026-08-25T04:24:36Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.22577v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.22577", "identifiers": {"arxiv": "2608.22577"}, "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/2608.22577"}], "name": "CausalCache: Conditional High-Fidelity Restoration for Long-Horizon 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": "2026-08-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces CausalCache, which selects archived screenshots to restore alongside text summaries under a fixed visual-memory budget. A history-gated adapter processes restored images, with evaluation distinguishing standard step limits from longer-horizon and mobile-transfer settings.", "topics": ["memory", "long-horizon", "efficiency"], "updated_at": "2026-08-25", "url": "https://arxiv.org/abs/2608.22577"}
{"aliases": [], "authors": ["Lin Fu", "Zheyuan Yang", "Tianhui Zhang", "Jinbiao Wei", "Guo Gan", "Boxu Liu", "Yilun Zhao", "Yu Rong"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dba8eb3f30a7e88ad973080d07fbf272f63400f4f8fa858d5f2657735ce093c0", "submitted_at": "2026-08-30T05:19:43Z", "updated_at": "2026-08-30T05:19:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.00048v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:13Z", "role": "primary", "sha256": "ca895b2f0dcd74816fb5e775a7161a0932ff7120a72477791f5e4c0bb1ead88e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2608.17319%2C2608.18307%2C2608.20319%2C2608.20707%2C2608.20797%2C2608.21794%2C2608.21830%2C2608.21832%2C2608.21898%2C2608.22077%2C2608.22533%2C2608.22577%2C2608.22847%2C2608.24015%2C2608.24017%2C2608.24099%2C2608.24174%2C2608.24680%2C2608.24848%2C2608.25777%2C2608.26200%2C2608.26204%2C2608.27477%2C2608.27508%2C2608.27808%2C2608.27866%2C2608.29305%2C2608.30038%2C2608.30207%2C2609.00028%2C2609.00048%2C2609.00524%2C2609.02057%2C2609.02149%2C2609.02154%2C2609.02309%2C2609.02885%2C2609.03438%2C2609.04850%2C2609.04869%2C2609.05374%2C2609.05511%2C2609.05513%2C2609.05535%2C2609.06530%2C2609.06792%2C2609.07603%2C2609.07712%2C2609.09212%2C2609.10031&max_results=50"}], "id": "arxiv:2609.00048", "identifiers": {"arxiv": "2609.00048"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:13Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.00048"}], "name": "GUI-CC: Benchmarking Contextual Consistency of GUI World Models as Agent Environments", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces GUI-CC, a benchmark for assessing GUI world models as multi-step agent environments. Separate reference-action and agent-loop tracks test whether generated screens preserve context, plausible transitions, and task progress throughout interaction.", "topics": ["world-models", "evaluation", "long-horizon"], "updated_at": "2026-08-30", "url": "https://arxiv.org/abs/2609.00048"}
{"aliases": [], "authors": ["Zixiang Chen", "Yuheng Lu", "Zihao Cheng", "Zeming Liu", "Jizeng Bai", "Ziye Huang", "Zhiyin Lin", "Zihan Li", "Yuhang Guo", "Yunhong Wang", "Haifeng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5ec17516bce87a8e349983acac14415a6ceeb12c167fce57aa37d6157cd3cd83", "submitted_at": "2026-09-09T16:59:59Z", "updated_at": "2026-09-09T16:59:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.10451v1"}}, "environments": ["desktop", "android", "windows", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.10451", "identifiers": {"arxiv": "2609.10451"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.10451"}], "name": "JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "JarvisGUI composes typed input-output task transformations into dynamic workflows spanning Android, Windows and Ubuntu devices. Its virtual-environment benchmark tests transfer of intermediate results, shared state and cross-device dependencies.", "topics": ["evaluation", "cross-app", "planning", "long-horizon", "state-tracking"], "updated_at": "2026-09-09", "url": "https://arxiv.org/abs/2609.10451"}
{"aliases": [], "authors": ["Zihan Dong", "Yuanzhe Liu", "Zhiyuan Ma", "Qishi Zhan", "Dehan Kong", "Guohao Li", "Kaixin Li"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "0f5fd0a3024311f9aba44a2971c9c752ee2c38444139823746b90a7c9a6d5ed4", "submitted_at": "2026-09-14T19:19:05Z", "updated_at": "2026-09-21T14:30:34Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.16251v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.16251", "identifiers": {"arxiv": "2609.16251"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.16251"}], "name": "CADWorld: Computer-Use Benchmark for Long-Horizon Computer-Aided Design", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-14", "relations": [], "review_status": "curated", "scope": "core", "summary": "CADWorld evaluates screenshot-and-GUI agents on long FreeCAD workflows, from modeling and assembly to manufacturing and simulation. Executable checks inspect saved engineering artifacts for geometry, constraints, construction structure and downstream state.", "topics": ["evaluation", "long-horizon", "workflow", "science"], "updated_at": "2026-09-21", "url": "https://arxiv.org/abs/2609.16251"}
{"aliases": [], "authors": ["Bofan Chen", "Boxuan Zhang", "Fei Tang", "Zhengxi Lu", "Yong Du", "Tongbo Chen", "Weiming Lu", "Jun Xiao", "Yueting Zhuang", "Yongliang Shen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "975662f216cc9a6b89856d4457f2125b733620e80bfac1b6424775253071d2e6", "submitted_at": "2026-09-15T17:24:06Z", "updated_at": "2026-09-15T17:24:06Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.17653v1"}}, "environments": ["android", "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", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.17653", "identifiers": {"arxiv": "2609.17653"}, "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/2609.17653"}], "name": "Reflect, Revise, Reuse: Training-Free Skill Evolution for GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-09-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "EvoSkill-GUI revises reusable skill files from execution feedback through an executor–critic loop. It studies how training-free updates to localization, recovery and procedure information help subsequent mobile and desktop tasks.", "topics": ["memory", "training-free", "self-reflection", "robustness", "long-horizon", "workflow"], "updated_at": "2026-09-15", "url": "https://arxiv.org/abs/2609.17653"}
{"aliases": [], "authors": ["Alexander Gill", "Md Farhan Ishmam", "Xuyen Nguyen", "Neha Bhat", "Parker Henry DeYoung", "Fateme Hashemi Chaleshtori", "Nathan Stringham", "Kenneth Marino", "Ana Marasović"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d6ed206d8f8fec53c03956fffcafec6b55d94a47fc705bbc8b77c832b76419e6", "submitted_at": "2026-09-24T22:36:10Z", "updated_at": "2026-09-24T22:36:10Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.30604v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.30604", "identifiers": {"arxiv": "2609.30604"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.30604"}], "name": "The Hard Part Comes After Search: Benchmarking Web Agents on Synthesizing, Organizing, and Displaying Knowledge", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-24", "relations": [], "review_status": "curated", "scope": "core", "summary": "KNOWS benchmarks browser workflows that retrieve and synthesize information into documents, presentations and spreadsheets. Task-specific evaluators combine deterministic checks with model judgments, distinguishing partial progress from usable end-to-end artifact completion.", "topics": ["evaluation", "web-navigation", "workflow", "long-horizon"], "updated_at": "2026-09-24", "url": "https://arxiv.org/abs/2609.30604"}
{"aliases": [], "authors": ["Bingo Zhang", "Haochuan Lu", "Zongjie Li", "Genjian Li", "Ari Yu Zhang", "Chaozheng Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f81fd07262dccf6783729e4336fbd77835f2fcdea030eef8dc2272e7ef82b8fc", "submitted_at": "2026-09-28T09:49:25Z", "updated_at": "2026-09-28T09:49:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.34769v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.34769", "identifiers": {"arxiv": "2609.34769"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.34769"}], "name": "LongPuzzleBench: Evaluating GUI Agents on Long-Horizon Visual Puzzles", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced. Agents execute real GUI moves in six puzzle games; the tasks diagnose computer-control planning but remain game-based rather than general productivity workflows.", "published_at": "2026-09-28", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "LongPuzzleBench evaluates coherent long-horizon planning on persistent visual puzzle boards through native GUI actions. It compares GUI-only and code-assisted agents and uses controlled hints and memory diagnostics to examine decisions that preserve or eliminate future solutions.", "topics": ["evaluation", "planning", "long-horizon", "state-tracking"], "updated_at": "2026-09-28", "url": "https://arxiv.org/abs/2609.34769"}
{"aliases": [], "authors": ["Bin Kang", "Jiarui Ouyang", "Li Jiang", "Bin Chen", "Zhuotao Tian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e3af30198735ca9b9d192e15acbb7f8a37fe222fa2c0c1d2ccff448ef5e79f55", "submitted_at": "2026-09-29T07:41:34Z", "updated_at": "2026-09-29T07:41:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.36923v1"}}, "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:2609.36923", "identifiers": {"arxiv": "2609.36923"}, "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/2609.36923"}], "name": "PrecogUI: Proactive GUI Agents via Pre-cognitive Simulation and Experience Retrieval", "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": "Presents PrecogUI, which retrieves prior anomaly and success patterns and predicts symbolic UI transitions before choosing actions. It also introduces InterfereBench, generated for evaluating long GUI workflows with disruptive events and recovery requirements.", "topics": ["memory", "world-models", "robustness", "long-horizon"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.36923"}
{"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": ["Xing Han Lù", "Dheeraj Vattikonda", "Sina Hajimiri", "Fatemeh Pesaran Zadeh", "Parishad BehnamGhader", "Ghazwa Darwiche", "Amirhossein Kazemnejad", "Christopher Pal", "Alexandre Drouin", "Siva Reddy"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "37413f11bf8263f1516a8edca8c594227fcab104c0c767a7daa8eead9645ec34", "submitted_at": "2026-10-08T01:03:48Z", "updated_at": "2026-10-08T01:03:48Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.11050v1"}}, "environments": ["cross-platform"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.11050", "identifiers": {"arxiv": "2610.11050"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.11050"}], "name": "AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AgentHorizon, a benchmark for judging long computer-use trajectories using closely related instructions paired with recorded actions and screenshots. It tests whether judges can detect constraint violations and distinguish successful tasks from similar but incompatible requests.", "topics": ["evaluation", "long-horizon", "cross-app"], "updated_at": "2026-10-08", "url": "https://arxiv.org/abs/2610.11050"}
{"aliases": [], "authors": ["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": ["Yuchen Lu", "Julian McAuley", "Lei Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6f4a9b6fae506d4dd35798c1930d5df4ec7fc7a2c452ab8bd5a5714cf385d505", "submitted_at": "2026-10-13T17:35:27Z", "updated_at": "2026-10-14T04:46:42Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2610.18784v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:24Z", "role": "primary", "sha256": "9641755401e0e8ec9613f7c4509029d968ab4f10751ba4d43cab57e7d27c1da1", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2610.11655%2C2610.11707%2C2610.11942%2C2610.12611%2C2610.12823%2C2610.12866%2C2610.12984%2C2610.13268%2C2610.13416%2C2610.13703%2C2610.14157%2C2610.14228%2C2610.15138%2C2610.15765%2C2610.16234%2C2610.16319%2C2610.16720%2C2610.16778%2C2610.17457%2C2610.17710%2C2610.17792%2C2610.17895%2C2610.17928%2C2610.18508%2C2610.18784%2C2610.18868%2C2610.19100%2C2610.19128%2C2610.20584&max_results=29"}], "id": "arxiv:2610.18784", "identifiers": {"arxiv": "2610.18784"}, "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/2610.18784"}], "name": "Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon 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": "adjacent", "summary": "Introduces CarryOn, which maintains evidence-linked task objects and remaining-work context across history compression, with evaluations on desktop tasks and web research.", "topics": ["memory", "long-horizon", "efficiency"], "updated_at": "2026-10-14", "url": "https://arxiv.org/abs/2610.18784"}
{"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": ["Zhaoyang Wang", "Qianhui Wu", "Haorui Wang", "Simon Yu", "Baolin Peng", "Canwen Xu", "Hao Cheng", "Wenlin Yao", "Boyi Liu", "Yite Wang", "Jianfeng Gao", "Huaxiu Yao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1f1ff33859352b61f017709fff91fd8e32f08636e9423fb044e7ca8e28f06905", "submitted_at": "2026-10-19T17:55:35Z", "updated_at": "2026-10-19T17:55:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.26828v1"}}, "environments": ["web", "desktop", "linux"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T05:52:00Z", "role": "primary", "sha256": "af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:27:26.236797+00:00", "role": "primary", "sha256": "80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T08:29:05.337543+00:00", "role": "primary", "sha256": "b680e07a5a883e7bb0210a1734a628be6c47d91b7df0487a2044d56d1b7e227e", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.26828v1"}], "id": "arxiv:2610.26828", "identifiers": {"arxiv": "2610.26828"}, "kind": "paper", "last_verified_at": "2026-10-20T08:27:26.236797+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.26828"}], "name": "Agent World Model in the Wild: Computer-Use Agent Environment Synthesis", "notes": "Reviewed the complete abstract and selected v1 synthesis, training and evaluation sections. The system uses hybrid GUI/MCP/CLI interfaces. Evaluation includes benchmark subsets and model-based judgments, so reported scores are not treated as directly comparable GUI-only results. No independent reproduction.", "published_at": "2026-10-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "WildAWM synthesizes resettable GUI/MCP/CLI environments and cross-interface tasks, using execution-based checks of solvability, difficulty and verifier consistency before supervised and reinforcement learning.", "topics": ["data-generation", "sandbox", "tool-integration", "long-horizon", "imitation-learning", "reinforcement-learning"], "updated_at": "2026-10-19", "url": "https://arxiv.org/abs/2610.26828"}
{"aliases": [], "authors": [], "details": {"availability": "Public inference/scoring source; benchmark data access remains gated.", "evaluation_caution": "Missing screenshots are marked AUTO_PASS and credited as correct according to the README.", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/LICENSE"], "reason": "Repository code license only; no data-license or gated-access permission is inferred."}}, "repository_revision": "0c2da7002baa5b0ee71f80dd6fbd90e33071e259"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:21Z", "role": "primary", "sha256": "6681ce1bd98c856fed4193e239e234a69eb2a74fa937e9c38db89f48db40f452", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:20Z", "role": "primary", "sha256": "f2b7c762383746a56e62273f756ca7ed3223726535763a41fda3ac5e3467aa82", "supports": ["identity"], "url": "https://api.github.com/repos/CVC2233/AndroTMem"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:59Z", "role": "primary", "sha256": "302c804dc65ebb3e8558651e0c7cd0a79c1caf29fc9fb11868d26fba36afdf48", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/CVC2233/AndroTMem/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:00Z", "role": "primary", "sha256": "6d572ccef64746baac78441c0b26488f525c70c2b7453f985bd4f19cbec00634", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:07Z", "role": "primary", "sha256": "e82623b7dc889eb7bff45c33f15a47c8c5caac3de0a3c56a7a9975ca9b2d6694", "supports": ["notes", "release_status"], "url": "https://huggingface.co/api/datasets/CVC2233/AndroTMem-Bench"}], "id": "github:cvc2233/androtmem", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-17T11:18:21Z", "licenses": {"code": {"basis": "Repository code license only; no data-license or gated-access permission is inferred.", "declared": "MIT", "retrieved_at": "2026-10-17T11:23:00Z", "url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/CVC2233/AndroTMem"}], "name": "AndroTMem inference and evaluation code", "notes": "The source tree releases inference backends and evaluation scripts under evaluation/. The separately hosted benchmark is gated=auto and its raw card returned 401; no access request was made and it is not approved here as a directly accessible dataset. The README explicitly credits missing screenshots as AUTO_PASS/correct, so incomplete local image inventories can inflate scores. Its TCR is an offline anchor/dependency score and is not presented here as independently verified live task success.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/CVC2233/AndroTMem/HEAD/README.md", "target_id": "arxiv:2603.18429", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A public Android GUI evaluation framework with action-history, summary and anchored-memory prompt modes, plus step-action and dependency-aware task scoring.", "topics": ["memory", "long-horizon", "evaluation"], "updated_at": null, "url": "https://github.com/CVC2233/AndroTMem"}
{"aliases": ["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": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/LICENSE", "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/README.md"], "reason": "The README declares Apache-2.0, but the inspected LICENSE is a short notice and link rather than the complete terms; the code license remains unconfirmed."}}, "repository_id": 1278022559}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:54:57.854138+00:00", "role": "primary", "sha256": "3f7545f466407ed5336f47d18b12e8354ec352ff83dcf973b1545f81d9a629ac", "supports": ["name", "kind", "summary", "topics", "scope", "relations", "notes", "licenses.code"], "url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:26.237759+00:00", "role": "primary", "sha256": "4c92c584b6885a3bde8ee89d04ccf48287e7f6119889f75cde0c034c90bd8217", "supports": ["summary", "environments", "topics", "scope", "notes"], "url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/evaluation/memgui3k_offline_eval/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T03:55:01.061844+00:00", "role": "primary", "sha256": "d4e344274155be1d0a192c8ae2940169abb0bd8546d045b96b45810593860636", "supports": ["licenses.code", "notes"], "url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/LICENSE"}, {"method": "github_repository_identity", "retrieved_at": "2026-10-19T19:10:26.954787+00:00", "role": "primary", "sha256": "0ae4477b885b0915e43874399c2a6b91f7d07954697370204550df931f7db82c", "supports": ["details.repository_id"], "url": "https://api.github.com/search/repositories?q=topic%3Agui-agent+is%3Apublic+fork%3Afalse+created%3A2008-01-01..2026-10-18&per_page=100&page=1&sort=updated&order=asc"}], "id": "github:kwai/memgui-agent", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-20T03:54:57.854138+00:00", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/kwai/MemGUI-Agent"}], "name": "MemGUI-Agent", "notes": "The inspected release documentation describes dataset utilities, a LoRA SFT template and gold-step offline evaluation; live Android rollouts are explicitly outside that evaluator and linked to separate benchmark repositories. Executable source files, external datasets/checkpoints and reported results were not independently verified in this bounded review.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/kwai/MemGUI-Agent/main/README.md", "target_id": "arxiv:2606.19926", "type": "introduced_by"}], "review_status": "curated", "scope": "core", "summary": "MemGUI-Agent training and offline evaluation tools provide trajectory conversion, LoRA SFT templates, and step-level checks of GUI actions, memory updates and context folding.", "topics": ["memory", "interaction-history", "long-horizon", "imitation-learning", "evaluation"], "updated_at": null, "url": "https://github.com/kwai/MemGUI-Agent"}
{"aliases": [], "authors": [], "details": {"availability": "Public graph/task manifests, evaluation source and ungated screenshot archive manifest.", "benchmark_protocol": {"actions": ["mobile GUI operations along graph transitions"], "counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "tasks", "value": 175}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "cross-app tasks", "value": 101}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "single-app tasks", "value": 74}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "population": "reported ColorBench benchmark", "qualifier": "exact", "unit": "mobile applications", "value": 21}], "evaluation": "The publisher reports success rate, completion rate and atomic-capability diagnostics using graph-based evaluation.", "evidence_urls": ["https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md"], "focus": ["long-horizon mobile task completion", "multi-path graph interaction"], "notes": "The interaction environment replays graph states rather than running real user accounts. The 101 and 74 subsets partition the reported 175 tasks; do not add the three counts.", "observations": ["mobile screenshots represented as graph states"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/LICENSE"], "reason": "Repository source LICENSE."}, "data": {"evidence_urls": ["https://huggingface.co/datasets/Norsheep/ColorBench/raw/main/README.md"], "reason": "Publisher metadata for the separately hosted screenshot archive; underlying third-party content retains applicable rights."}}, "repository_revision": "1f26a84230d26f99d84eab97d05032582b39018e"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:16Z", "role": "primary", "sha256": "5c0190fffb68b91c1855a408f6a5e252871d94000b95fa258504a3c100600cec", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:18:16Z", "role": "primary", "sha256": "2f6940503cc0d6cb97bdf142d6b8c6ab116de82227f2fa6d97fd56d7e5aebec0", "supports": ["identity"], "url": "https://api.github.com/repos/MadeAgents/ColorBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:54Z", "role": "primary", "sha256": "6a00c093a6794ec4a172d30b70fdabc17400584a9e25a157a47a7661dbdd20fa", "supports": ["release_status", "details.repository_revision"], "url": "https://api.github.com/repos/MadeAgents/ColorBench/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:06Z", "role": "primary", "sha256": "c0c3286b5150e128e55034ba0393ef85f7fcf73e4b73f4276389a04c9bcec919", "supports": ["identity", "release_status"], "url": "https://huggingface.co/api/datasets/Norsheep/ColorBench"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:23:06Z", "role": "primary", "sha256": "4bcf87ecfbbb8e07a01b21415a970c8b53a5283bf6872b657040d3f45c9241f7", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/Norsheep/ColorBench/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:54Z", "role": "primary", "sha256": "130976c026bcafa9dcc479697fe6595dbefb73cdb375fc47b8478acc16e6ea73", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/LICENSE"}], "id": "github:madeagents/colorbench", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T11:18:16Z", "licenses": {"code": {"basis": "Repository source LICENSE.", "declared": "Apache-2.0", "retrieved_at": "2026-10-17T11:22:54Z", "url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/LICENSE"}, "data": {"basis": "Publisher metadata for the separately hosted screenshot archive; underlying third-party content retains applicable rights.", "declared": "apache-2.0", "retrieved_at": "2026-10-17T11:23:06Z", "url": "https://huggingface.co/datasets/Norsheep/ColorBench/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/MadeAgents/ColorBench"}], "name": "ColorBench", "notes": "The source tree contains graph and task JSON files plus construction/evaluation code; the linked ungated Hub provides the screenshot archive. The environment is static but interactive and graph-based, not a live-device deployment. Counts are publisher-reported task/app inventories. Apache-2.0 declarations for repository code and the Hub dataset are separately retained; third-party application visuals are not granted additional rights by this catalog.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/MadeAgents/ColorBench/HEAD/README.md", "target_id": "arxiv:2510.14621", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A graph-based mobile GUI benchmark that evaluates long tasks through screenshot states and multiple valid or erroneous transitions, including reflection, replanning and backtracking.", "topics": ["evaluation", "mobile-navigation", "long-horizon", "cross-app"], "updated_at": null, "url": "https://github.com/MadeAgents/ColorBench"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/MobileLLM/AgentProg/HEAD/LICENSE"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate."}}, "repository_tree_sha": "ef5e2998ac2defa86e0dc4c13e7e00d595e1e7cf", "repository_url": "https://github.com/MobileLLM/AgentProg"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:13Z", "role": "primary", "sha256": "becc4961d93f0170fab29120afa4b09b2999f2cd0e74b13ccb7f28cb9623761b", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/MobileLLM/AgentProg/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T15:37:01Z", "role": "primary", "sha256": "d14eecf7c041933cefecc03b62bbea45c1617a183fd8c4ec1334e22a644a7f67", "supports": ["identity"], "url": "https://api.github.com/repos/MobileLLM/AgentProg"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:25:22Z", "role": "primary", "sha256": "dd1cf9cd0aed31af63d7046ffa98d1be2c36258706951173ad8c28a64fcf9f7a", "supports": ["identity", "release_status", "details.repository_tree_sha"], "url": "https://api.github.com/repos/MobileLLM/AgentProg/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:19Z", "role": "primary", "sha256": "751123fccb2fd04b5c08b8c712a68a8bc69f56391593711efbfa533eaebbdb10", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/MobileLLM/AgentProg/HEAD/LICENSE"}], "id": "github:mobilellm/agentprog", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-18T04:27:13Z", "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/MobileLLM/AgentProg/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/MobileLLM/AgentProg"}], "name": "AgentProg", "notes": "Public tree contains the agentprog package, CLI, program/context-management code and AndroidWorld evaluation integration. README announces AW-Extend, but this record is the framework rather than a separately audited benchmark release. External model services and Android device/emulator setup are required. Apache-2.0 covers code, not external model weights, apps or VM images.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/MobileLLM/AgentProg/HEAD/README.md", "target_id": "arxiv:2512.10371", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Android GUI-agent framework that keeps long-running task context in structured programs, variables and control flow, with a CLI and AndroidWorld integration.", "topics": ["long-horizon", "memory", "planning", "state-tracking"], "updated_at": null, "url": "https://github.com/MobileLLM/AgentProg"}
{"aliases": [], "authors": [], "details": {"availability": "Public task definitions, execution harness and an ungated VM-file manifest.", "benchmark_protocol": {"actions": ["actions executed by the OSWorld-derived DesktopEnv"], "counts": [{"basis": "computed", "evidence_url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/all.json", "population": "evaluation_examples/all.json", "qualifier": "exact", "unit": "category/task references", "value": 416}], "evaluation": "Task configurations provide evaluators for application/system state, using an OSWorld-derived runtime.", "evidence_urls": ["https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md", "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/all.json", "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/README.md"], "focus": ["breadth of desktop task coverage", "depth of desktop task execution"], "notes": "416 category/task pairs are unique, but only 412 bare ID strings are unique across categories. This is a manifest-reference count, not a claim of 416 semantically distinct tasks. Main runtime documentation specifies Ubuntu; the subdirectory README retains older Windows-only wording.", "observations": ["desktop observations from an Ubuntu virtual machine"], "setting": "interactive"}, "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/LICENSE"], "reason": "Repository code LICENSE; no license is inferred for the VM image or bundled third-party applications."}}, "repository_revision": "191efd0e6766acab444a31685b437565e1a94ac2"}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:14Z", "role": "primary", "sha256": "fc9f3c8d9fffdff8b51e312b013be0d9fdfc10e57fdc2ec4a8a8d012c4b42475", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:12:14Z", "role": "primary", "sha256": "d6c91b4910e92f40cc694b4cb6467c1a20f42c0c14047b95a1e777aca0892677", "supports": ["identity"], "url": "https://api.github.com/repos/OS-Copilot/OS-Map"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:07Z", "role": "primary", "sha256": "41342157c53f28ad2729137491930225ac11da922e7b16762e84ac958a2e9d28", "supports": ["release_status"], "url": "https://api.github.com/repos/OS-Copilot/OS-Map/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:07Z", "role": "primary", "sha256": "1dd924135cbf0a85ce72d75025259adb7b849970d400af4ed854327a34e7ee41", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/LICENSE"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:17:22Z", "role": "primary", "sha256": "4199af4c8bc60451079d8a2e7598ab7386bd82b0d5c9ad7e6927d6cb8e03ad6f", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/all.json"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:17:22Z", "role": "primary", "sha256": "a6b60d5721cfdd26416e7d0ae3192360896427a97ccb3793812cc150e9cd7126", "supports": ["details.benchmark_protocol", "notes"], "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/evaluation_examples/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T09:14:16Z", "role": "primary", "sha256": "d2da837872b123f2927a069a020aeb14fb4b22148599c9f2e8b6689ee444ec88", "supports": ["release_status", "details"], "url": "https://huggingface.co/api/datasets/os-map/OSMAP"}], "id": "github:os-copilot/os-map", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T09:12:14Z", "licenses": {"code": {"basis": "Repository code LICENSE; no license is inferred for the VM image or bundled third-party applications.", "declared": "MIT", "retrieved_at": "2026-10-17T09:14:07Z", "url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OS-Copilot/OS-Map"}], "name": "OS-Map", "notes": "Public task manifests, agent runner and evaluator code are present, and the linked ungated Hub manifest lists Ubuntu0.zip and task files. The main setup supports x86 VMware Workstation and says ARM/Docker alternatives are still in development. An inherited evaluation-examples README still mentions Windows 10 and under-construction status; those statements are retained as a documentation inconsistency, not interpreted as verified Windows support. The manifest has 416 category/task references but 412 globally distinct ID strings.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md", "target_id": "github:xlang-ai/osworld", "type": "uses"}, {"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/OS-Map/HEAD/README.md", "target_id": "arxiv:2507.19132", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "An OSWorld-derived desktop benchmark with an Ubuntu virtual-machine environment and expanded task configurations for measuring breadth and depth of computer interaction.", "topics": ["evaluation", "long-horizon", "cross-app"], "updated_at": null, "url": "https://github.com/OS-Copilot/OS-Map"}
{"aliases": [], "authors": [], "details": {"leaderboard_url": "https://osu-nlp-group.github.io/Mind2Web-2/#leaderboard"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T17:18:21Z", "role": "primary", "sha256": "5cc72b992cfc11094dcf341c284d5baadc6c66a302fdfd474b27913fe1c91f6f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web-2/HEAD/README.md"}, {"method": "license_header", "retrieved_at": "2026-10-17T17:18:56Z", "role": "primary", "sha256": "93b43d692b033b76129504448695bfe76ef22d18b11e51352bfab4b5622e5aaa", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web-2/HEAD/LICENSE"}], "id": "github:osu-nlp-group/mind2web-2", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-17T17:18:21Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-17T17:18:56Z", "url": "https://raw.githubusercontent.com/OSU-NLP-Group/Mind2Web-2/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/Mind2Web-2"}], "name": "Mind2Web 2", "notes": "", "published_at": null, "relations": [], "review_status": "curated", "scope": "core", "summary": "Long-horizon agentic web search benchmark whose Agent-as-a-Judge framework scores time-varying, citation-backed answers, with evaluation scripts for its dev and test sets.", "topics": ["evaluation", "web-navigation", "long-horizon"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/Mind2Web-2"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "abc848e7eb5181be22168dbc076446c5eb0f065fd57c7e10367284ca6e50bf5c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/shulin16/MMInA/HEAD/README.md"}], "id": "github:shulin16/mmina", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/shulin16/MMInA"}], "name": "MMInA", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/shulin16/MMInA/HEAD/README.md", "target_id": "arxiv:2404.09992", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Multimodal web tasks in a WebArena-based environment that chain from 1 to 10 hops across sites such as a shopping store and Wikipedia, testing long-chain reasoning.", "topics": ["evaluation", "long-horizon"], "updated_at": null, "url": "https://github.com/shulin16/MMInA"}
{"aliases": [], "authors": [], "details": {"license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/sunblaze-ucb/AgentSynth/HEAD/LICENSE"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate."}}, "repository_tree_sha": "cd42a4bd158fe3f0a1b0f376b9428d6a5fcb81b2", "repository_url": "https://github.com/sunblaze-ucb/AgentSynth"}, "environments": ["desktop", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:12Z", "role": "primary", "sha256": "42e8b8dead7caea3f8b21e93a075e8c895dfc30a25e1470bf97a90940bbf0106", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/sunblaze-ucb/AgentSynth/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T15:37:00Z", "role": "primary", "sha256": "e08f08a99b8cb3a4bb857a9fac976b2820f9ad406d66675b0286e880dfeeceaa", "supports": ["identity"], "url": "https://api.github.com/repos/sunblaze-ucb/AgentSynth"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:25:20Z", "role": "primary", "sha256": "b0998652a9b6b465444c2bbeef015bdbef50e83984a3bd15725f0e811645ca62", "supports": ["identity", "release_status", "details.repository_tree_sha"], "url": "https://api.github.com/repos/sunblaze-ucb/AgentSynth/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:19Z", "role": "primary", "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/sunblaze-ucb/AgentSynth/HEAD/LICENSE"}], "id": "github:sunblaze-ucb/agentsynth", "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/sunblaze-ucb/AgentSynth/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/sunblaze-ucb/AgentSynth"}], "name": "AgentSynth task-generation pipeline", "notes": "Released code includes task-generation and training-format conversion scripts. OSWorld/InSTA environments and model services require separate setup; the repository is not a bundled VM or model checkpoint. Code and the separately published dataset have distinct license declarations.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/sunblaze-ucb/AgentSynth/HEAD/README.md", "target_id": "arxiv:2506.14205", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Generates synthetic computer-use tasks and records trajectories in OSWorld desktop environments and InSTA browser environments.", "topics": ["data-generation", "long-horizon", "automation"], "updated_at": null, "url": "https://github.com/sunblaze-ucb/AgentSynth"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": [], "counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md", "population": "Hub test split at revision b8eca829", "qualifier": "exact", "unit": "task-definition records", "value": 25}], "evaluation": "The published task schema contains rubrics and human_reference_operations. No automated scoring protocol or single numeric agent metric is established by the inspected card; execution uses separately configured Android tooling.", "evidence_urls": ["https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md", "https://huggingface.co/api/datasets/mikewang/mobile_eval_e", "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/Mobile-Agent-E/README.md", "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"], "focus": ["long-horizon multi-app Android task execution"], "notes": "Features are task_id, instruction, type, apps, scenario, rubrics and human-reference operation strings. A task record is not an action count or recorded screenshot trajectory. Public Parquet inventory was checked but no sample rows were downloaded; companion README limits the tested agent platform to Android.", "observations": ["task instructions", "app/scenario metadata"], "setting": "interactive"}, "disabled": false, "download_url": "https://huggingface.co/datasets/mikewang/mobile_eval_e/tree/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83", "gated": false, "hub_id": "mikewang/mobile_eval_e", "hub_revision": "b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md"], "reason": "No data-license declaration is present in the inspected task card; companion code MIT does not establish a dataset grant."}}, "repository_url": "https://github.com/X-PLUG/MobileAgent/tree/main/Mobile-Agent-E"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:15Z", "role": "primary", "sha256": "f8515fbce73ea0ef3d1d558d4567bb949244e6384e61c956ba3ddbc27294f0ae", "supports": ["name", "kind", "environments", "scope", "notes", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/Mobile-Agent-E/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:29Z", "role": "primary", "sha256": "f9b10eada9e0a64e26f1a7d477768f218b807103199855d893b3ab07feddc755", "supports": ["identity", "release_status", "details.hub_id", "details.hub_revision", "details.gated", "details.disabled", "details.download_url", "details.benchmark_protocol"], "url": "https://huggingface.co/api/datasets/mikewang/mobile_eval_e"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:15Z", "role": "primary", "sha256": "9fdf43753fbfac9868d8bec57fa5601db3e03ed5c30fb981f9da88f6b75ffcc5", "supports": ["summary", "details.benchmark_protocol", "licenses.data"], "url": "https://huggingface.co/datasets/mikewang/mobile_eval_e/raw/b8eca82988ed8d4f9dec01eaa9d51abf4b35aa83/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:38Z", "role": "primary", "sha256": "e96b86015c763c780e0430a8d0867b96cae947f50e888b3441c85abcaa42de78", "supports": ["summary", "topics", "details.benchmark_protocol"], "url": "https://export.arxiv.org/api/query?id_list=2412.04454%2C2412.04626%2C2412.05467%2C2412.09362%2C2412.09605%2C2412.10342%2C2412.10742%2C2412.10840%2C2412.13103%2C2412.13194%2C2412.13501%2C2412.14161%2C2412.14470%2C2412.16256%2C2412.17589%2C2412.18116%2C2412.18426%2C2412.19723%2C2501.01149%2C2501.04575%2C2501.07572%2C2501.11733%2C2501.12326%2C2501.13896%2C2501.16150%2C2502.01600%2C2502.01977%2C2502.02982%2C2502.06395%2C2502.07949%2C2502.08047%2C2502.08226%2C2502.11357%2C2502.13053%2C2502.13923%2C2502.14282%2C2502.18906%2C2503.00401%2C2503.02268%2C2503.03196%2C2503.04730%2C2503.05143%2C2503.06470%2C2503.09263%2C2503.11069%2C2503.11170%2C2503.12687%2C2503.15661%2C2503.15937%2C2503.16465&max_results=50"}], "id": "url:342b081c3668a8bbe155", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-18T04:27:15Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/mikewang/mobile_eval_e"}], "name": "Mobile-Eval-E task definitions", "notes": "The public Hub manifest contains one test Parquet file. This resource is the benchmark’s task-definition package, not a bundled phone environment or agent. Card metadata supplies rubrics/reference-operation fields but no data-license declaration or verified automated scorer; the code repository’s MIT license is not transferred to these records.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/X-PLUG/MobileAgent/HEAD/Mobile-Agent-E/README.md", "target_id": "arxiv:2501.11733", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Task definitions for Mobile-Agent-E’s long-horizon, multi-app Android evaluation, packaged as a 25-record test split with rubrics and human-reference operation strings.", "topics": ["evaluation", "long-horizon", "cross-app"], "updated_at": null, "url": "https://huggingface.co/datasets/mikewang/mobile_eval_e"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/tree/main", "gated": false, "hub_id": "xTRam1/plan-and-act-planner-70b", "license_review": {"weights": {"evidence_urls": ["https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md"], "reason": "The model card declares MIT in its metadata header (license: mit), beside a template comment asking the publisher to update it if needed. It is recorded as the publisher's declaration and does not override the Llama-3.3-70B-Instruct base model terms."}}}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "4ef2d84d2b5d9d2972907e6135ea49f1ef088beaf6ef6042dc7d8fb4297a9a3a", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.weights"], "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:28Z", "role": "primary", "sha256": "d2f757790ded81244e7c9b9e7350ec9fdc304609e387a95c7918db705bda535c", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/xTRam1/plan-and-act-planner-70b"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T02:02:06Z", "role": "primary", "sha256": "feee31aa4d0801cb364578099a13e4f9cdcaa4eb5fbf4312b3fc3f8edd479a14", "supports": ["environments"], "url": "https://raw.githubusercontent.com/SqueezeAILab/plan-and-act/HEAD/README.md"}], "id": "url:3932d69c6b6f9b2a7ba8", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-17T02:02:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "mit", "retrieved_at": "2026-10-17T02:02:28Z", "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b"}], "name": "Plan-and-Act Planner 70B", "notes": "The Hub manifest contains 17 safetensors shards and tokenizer/configuration files. Card metadata declares MIT and Llama-3.3-70B-Instruct as the base, but retains template comments asking the publisher to update those fields if needed. These are recorded as publisher declarations, without claiming they override upstream model terms.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b/raw/main/README.md", "target_id": "arxiv:2503.09572", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "The planner component of Plan-and-Act, generating structured high-level plans that an executor turns into web actions.", "topics": ["planning", "web-navigation", "long-horizon"], "updated_at": null, "url": "https://huggingface.co/xTRam1/plan-and-act-planner-70b"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://huggingface.co/datasets/sunblaze-ucb/AgentSynth/raw/9ec3f3139f9786e81aee14b7fff1eb487cc7e991/README.md", "population": "collection described in the dataset card; not an independently counted release-row total", "qualifier": "more-than", "unit": "tasks", "value": 6000}], "evidence_urls": ["https://huggingface.co/datasets/sunblaze-ucb/AgentSynth/raw/9ec3f3139f9786e81aee14b7fff1eb487cc7e991/README.md", "https://huggingface.co/api/datasets/sunblaze-ucb/AgentSynth"], "formats": ["JSONL", "ZIP trajectory packages"], "modalities": ["screenshots", "task instructions", "LLM thoughts", "commands"], "notes": "Card maps osworld/train to all_tasks.jsonl and insta/train to task_sequences.jsonl. Six difficulty levels and trajectory screenshots/thoughts/commands are documented. ZIP file labels are not trajectory counts; configurations and composed-task levels are not assumed disjoint.", "splits": ["train (osworld configuration)", "train (insta configuration)"], "supervision": ["synthetic tasks", "agent trajectories"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/sunblaze-ucb/AgentSynth/tree/9ec3f3139f9786e81aee14b7fff1eb487cc7e991", "gated": false, "hub_id": "sunblaze-ucb/AgentSynth", "hub_revision": "9ec3f3139f9786e81aee14b7fff1eb487cc7e991", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/sunblaze-ucb/AgentSynth/raw/9ec3f3139f9786e81aee14b7fff1eb487cc7e991/README.md"], "reason": "Publisher dataset-card declaration for this release; it does not establish rights to all upstream screenshots, applications or source content."}}, "repository_url": "https://github.com/sunblaze-ucb/AgentSynth"}, "environments": ["desktop", "web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:17Z", "role": "primary", "sha256": "c1669d8e3caa27f8cd47c17b243b6a8d42807438d8a6a36074668adc57d33757", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "licenses.data", "details.dataset_profile"], "url": "https://huggingface.co/datasets/sunblaze-ucb/AgentSynth/raw/9ec3f3139f9786e81aee14b7fff1eb487cc7e991/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:20Z", "role": "primary", "sha256": "0d3797d75dd4089ece8c8d4624f2a726024d25be255255d8770f78de21d75906", "supports": ["identity", "release_status", "details.hub_id", "details.hub_revision", "details.gated", "details.disabled", "details.download_url", "details.dataset_profile"], "url": "https://huggingface.co/api/datasets/sunblaze-ucb/AgentSynth"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:12Z", "role": "primary", "sha256": "42e8b8dead7caea3f8b21e93a075e8c895dfc30a25e1470bf97a90940bbf0106", "supports": ["details.repository_url"], "url": "https://raw.githubusercontent.com/sunblaze-ucb/AgentSynth/HEAD/README.md"}], "id": "url:58a209f2b3b970916396", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-18T04:35:17Z", "licenses": {"code": null, "data": {"basis": "Publisher dataset-card declaration for this release; it does not establish rights to all upstream screenshots, applications or source content.", "declared": "Apache-2.0", "retrieved_at": "2026-10-18T04:35:17Z", "url": "https://huggingface.co/datasets/sunblaze-ucb/AgentSynth/raw/9ec3f3139f9786e81aee14b7fff1eb487cc7e991/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/sunblaze-ucb/AgentSynth"}], "name": "AgentSynth tasks and trajectories", "notes": "Public Hub manifest contains JSONL task files and trajectory ZIP packages. Counts describe the publisher-reported task collection, not independently counted archive rows or numbers inferred from ZIP filenames.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/sunblaze-ucb/AgentSynth/raw/9ec3f3139f9786e81aee14b7fff1eb487cc7e991/README.md", "target_id": "arxiv:2506.14205", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Published AgentSynth task definitions and recorded GUI trajectories, with OSWorld and InSTA training configurations and packaged screenshots, thoughts and commands.", "topics": ["data-generation", "long-horizon", "imitation-learning"], "updated_at": null, "url": "https://huggingface.co/datasets/sunblaze-ucb/AgentSynth"}