{"aliases": [], "authors": ["Ruisheng Cao", "Fangyu Lei", "Haoyuan Wu", "Jixuan Chen", "Yeqiao Fu", "Hongcheng Gao", "Xinzhuang Xiong", "Hanchong Zhang", "Yuchen Mao", "Wenjing Hu", "Tianbao Xie", "Hongshen Xu", "Danyang Zhang", "Sida Wang", "Ruoxi Sun", "Pengcheng Yin", "Caiming Xiong", "Ansong Ni", "Qian Liu", "Victor Zhong", "Lu Chen", "Kai Yu", "Tao Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7922a723142ac62018064fe5fc9357fb77c08e316ac4d6d6ba728f1b337df690", "submitted_at": "2024-07-15T17:54:37Z", "updated_at": "2024-07-15T17:54:37Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.10956v1"}, "publication": {"authors": ["Ruisheng Cao", "Fangyu Lei", "Haoyuan Wu", "Jixuan Chen", "Yeqiao Fu", "Hongcheng Gao", "Xinzhuang Xiong", "Hanchong Zhang", "Yuchen Mao", "Wenjing Hu", "Tianbao Xie", "Hongshen Xu", "Danyang Zhang", "Sida Wang", "Ruoxi Sun", "Pengcheng Yin", "Caiming Xiong", "Ansong Ni", "Qian Liu", "Victor Zhong", "Lu Chen", "Kai Yu", "Tao Yu"], "date_basis": "Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.", "date_parts": [2024, 12, 16], "doi": "10.52202/079017-3421", "event_dates": null, "evidence_urls": ["https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html"], "identity_basis": "Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity. The normalized publisher abstract matches the retained arXiv abstract SHA-256.", "issue": null, "notes": "Publisher proceedings edition; retained separately from the arXiv snapshot.", "pages": "107703-107744", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/c2f71567cd53464161cab3336e8fc865-Paper-Datasets_and_Benchmarks_Track.pdf", "publisher": null, "series": null, "title": "Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?", "type": "conference-paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html", "venue": "Advances in Neural Information Processing Systems", "volume": "37"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:10:04Z", "role": "primary", "sha256": "872286a0210616be9737de716cc6069a1e9cbee1df3182d45c718582e185f62f", "supports": ["details.publication.authors", "details.publication.date_basis", "details.publication.date_parts", "details.publication.doi", "details.publication.identity_basis", "details.publication.notes", "details.publication.pages", "details.publication.pdf_url", "details.publication.title", "details.publication.type", "details.publication.url", "details.publication.venue", "details.publication.volume"], "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:32Z", "role": "primary", "sha256": "b9a83994d8ac36ccdb606a5219511f8bb35ad3948ade4f9f3e28d386221aedfb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2406.12814%2C2406.13719%2C2406.14056%2C2406.14250%2C2406.14314%2C2406.18082%2C2406.19263%2C2407.00993%2C2407.01476%2C2407.01489%2C2407.01511%2C2407.01558%2C2407.03037%2C2407.03913%2C2407.04346%2C2407.05291%2C2407.07061%2C2407.08850%2C2407.09018%2C2407.09295%2C2407.10956%2C2407.13032%2C2407.13063%2C2407.15673%2C2407.15711%2C2407.16741%2C2407.17490%2C2407.18901%2C2407.19056%2C2407.20183%2C2408.00203%2C2408.02544%2C2408.06327%2C2408.07199%2C2408.11824%2C2408.15978%2C2409.02453%2C2409.04566%2C2409.07429%2C2409.08264%2C2409.09354%2C2409.10741%2C2409.11295%2C2409.13711%2C2409.14337%2C2409.14818%2C2409.15441%2C2409.15637%2C2409.17140%2C2409.20566&max_results=50"}], "id": "arxiv:2407.10956", "identifiers": {"arxiv": "2407.10956"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2407.10956"}, {"rel": "publication", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/c2f71567cd53464161cab3336e8fc865-Abstract-Datasets_and_Benchmarks_Track.html"}, {"rel": "publication_pdf", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/c2f71567cd53464161cab3336e8fc865-Paper-Datasets_and_Benchmarks_Track.pdf"}], "name": "Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-07-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "Benchmarks multimodal agents on data-science and engineering workflows involving enterprise applications, code, and GUI operations in real computer environments.", "topics": ["evaluation", "science", "workflow"], "updated_at": "2024-07-15", "url": "https://arxiv.org/abs/2407.10956"}
{"aliases": [], "authors": ["Qiushi Sun", "Zhoumianze Liu", "Chang Ma", "Zichen Ding", "Fangzhi Xu", "Zhangyue Yin", "Haiteng Zhao", "Zhenyu Wu", "Kanzhi Cheng", "Zhaoyang Liu", "Jianing Wang", "Qintong Li", "Xiangru Tang", "Tianbao Xie", "Xiachong Feng", "Xiang Li", "Ben Kao", "Wenhai Wang", "Biqing Qi", "Lingpeng Kong", "Zhiyong Wu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4743f6bfd0ca48717184cf55545e33a229dd5ab2d454385d33ea2f1b25c15b0c", "submitted_at": "2025-05-26T12:27:27Z", "updated_at": "2026-04-20T16:29:04Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2505.19897v3"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:42Z", "role": "primary", "sha256": "afe4e75ffb8a5c44f4c088a90ffa5f53553036f514aa768f5aee0c5aa38d731e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2503.16788%2C2503.17709%2C2503.18492%2C2503.19537%2C2503.20978%2C2503.21620%2C2503.23434%2C2504.01382%2C2504.07981%2C2504.10445%2C2504.10458%2C2504.11257%2C2504.11281%2C2504.12679%2C2504.13805%2C2504.13865%2C2504.14239%2C2504.18575%2C2504.19838%2C2504.20464%2C2505.10924%2C2505.11891%2C2505.12299%2C2505.12632%2C2505.13227%2C2505.13887%2C2505.13909%2C2505.14289%2C2505.15810%2C2505.16282%2C2505.16827%2C2505.18829%2C2505.19897%2C2505.21496%2C2505.21499%2C2505.21936%2C2505.22648%2C2505.23762%2C2505.24838%2C2506.00618%2C2506.01391%2C2506.01968%2C2506.02456%2C2506.03143%2C2506.03569%2C2506.04135%2C2506.04614%2C2506.07672%2C2506.08012%2C2506.08972&max_results=50"}], "id": "arxiv:2505.19897", "identifiers": {"arxiv": "2505.19897"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:42Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.19897"}], "name": "ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-05-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "ScienceBoard combines an interactive environment with professional scientific software and 169 human-curated tasks in areas including biochemistry, astronomy, and geoinformatics. It evaluates computer-use agents on visually rich research workflows and analyzes their execution limitations.", "topics": ["evaluation", "science", "workflow"], "updated_at": "2026-04-20", "url": "https://arxiv.org/abs/2505.19897"}
{"aliases": [], "authors": ["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": ["Zeyi Sun", "Yuhang Cao", "Jianze Liang", "Qiushi Sun", "Ziyu Liu", "Zhixiong Zhang", "Yuhang Zang", "Xiaoyi Dong", "Kai Chen", "Dahua Lin", "Jiaqi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ac401f8d5c56050e72839f3bffba4f2c78bbdf34a00cab05fe504f03efd96a19", "submitted_at": "2025-08-27T17:59:50Z", "updated_at": "2025-08-27T17:59:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2508.20096v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.20096", "identifiers": {"arxiv": "2508.20096"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.20096"}], "name": "CODA: Coordinating the Cerebrum and Cerebellum for a Dual-Brain Computer Use Agent with Decoupled Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "CODA combines a generalist planner with a specialist GUI executor and trains application-specific planners through decoupled GRPO. Successful specialist trajectories then support supervised training of a consolidated planner, evaluated across four ScienceBoard applications.", "topics": ["reinforcement-learning", "planning", "science", "imitation-learning"], "updated_at": "2025-08-27", "url": "https://arxiv.org/abs/2508.20096"}
{"aliases": [], "authors": ["Jiaxin Ai", "Yukang Feng", "Fanrui Zhang", "Jianwen Sun", "Zizhen Li", "Chuanhao Li", "Yifan Chang", "Wenxiao Wu", "Ruoxi Wang", "Mingliang Zhai", "Kaipeng Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "afab3dcd2444bf8333f2a2fea887819f6fca55c41209d8c449f6211d1c62b819", "submitted_at": "2025-12-30T01:49:46Z", "updated_at": "2025-12-30T01:49:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.02399v1"}}, "environments": ["windows", "desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:52Z", "role": "primary", "sha256": "8623b0ff7ef9d2e8ba484cd67359e7b8817e78e9ca22b8f27432c2533f5b90a0", "supports": ["name", "kind", "topics"], "url": "https://export.arxiv.org/api/query?id_list=2510.01670%2C2510.04673%2C2510.10073%2C2510.12200%2C2510.14621%2C2510.16051%2C2510.17790%2C2510.18488%2C2510.18596%2C2510.19488%2C2510.19949%2C2510.20286%2C2510.22009%2C2510.24168%2C2510.24411%2C2510.24563%2C2511.00810%2C2511.04307%2C2511.07332%2C2511.09157%2C2511.12254%2C2511.13087%2C2511.15567%2C2511.16590%2C2511.19663%2C2511.20597%2C2511.20766%2C2512.08529%2C2512.10371%2C2512.10962%2C2512.12634%2C2512.14014%2C2512.15431%2C2512.16295%2C2512.16501%2C2512.19432%2C2512.22894%2C2512.24965%2C2601.02399%2C2601.06031%2C2601.07181%2C2601.07262%2C2601.07779%2C2601.08406%2C2601.09923%2C2601.11044%2C2601.11868%2C2601.12822%2C2601.13060%2C2601.15876&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:06:36Z", "role": "primary", "sha256": "f9a44f8090369a296a584a9b94c4491662ee25bb8f7239ef75dfd8f775be8887", "supports": ["summary", "environments", "scope"], "url": "https://arxiv.org/html/2601.02399v1"}], "id": "arxiv:2601.02399", "identifiers": {"arxiv": "2601.02399"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.02399"}], "name": "ProSoftArena: Benchmarking Hierarchical Capabilities of Multimodal Agents in Professional Software Environments", "notes": "The paper abstract and official full-text sections were reviewed. The full text supports Windows 11 virtual-machine execution, screenshot/state observations, and mouse/keyboard actions. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "ProSoftArena evaluates 436 professional-software tasks across six disciplines and 13 applications in a Windows 11 virtual-machine environment. Its capability hierarchy, execution-based checks, and human-in-the-loop evaluation cover precise GUI operations and multi-software workflows.", "topics": ["evaluation", "industry", "science", "cross-app"], "updated_at": "2025-12-30", "url": "https://arxiv.org/abs/2601.02399"}
{"aliases": [], "authors": ["Jackson Vonderhorst", "Kuangshi Ai", "Haichao Miao", "Shusen Liu", "Chaoli Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8e7b07a811df7e74a28f91a5b6c59c69b7ba066bd6c64452a299e678a3187733", "submitted_at": "2026-04-30T15:22:28Z", "updated_at": "2026-06-26T20:31:32Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2604.27996v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:19Z", "role": "primary", "sha256": "05ec5d668ac75d3d7c5b71ea7d6e76a3fbf5ebeb742e0843db2329e5e41aa208", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2604.03598%2C2604.04759%2C2604.05547%2C2604.06170%2C2604.07340%2C2604.10547%2C2604.10938%2C2604.11978%2C2604.12282%2C2604.13018%2C2604.13346%2C2604.13954%2C2604.15579%2C2604.15671%2C2604.15715%2C2604.16520%2C2604.18240%2C2604.18934%2C2604.19572%2C2604.19667%2C2604.19742%2C2604.20841%2C2604.20987%2C2604.22436%2C2604.23781%2C2604.24479%2C2604.26148%2C2604.27621%2C2604.27996%2C2604.28093%2C2604.28139%2C2604.28181%2C2605.00425%2C2605.01462%2C2605.02240%2C2605.03042%2C2605.03242%2C2605.04811%2C2605.05191%2C2605.05811%2C2605.06110%2C2605.07073%2C2605.07358%2C2605.08374%2C2605.08985%2C2605.10912%2C2605.10966%2C2605.11363%2C2605.11732%2C2605.11817&max_results=50"}], "id": "arxiv:2604.27996", "identifiers": {"arxiv": "2604.27996"}, "kind": "paper", "last_verified_at": "2026-10-17T12:31:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2604.27996"}], "name": "Exploring LLM Agent Designs and Interaction Modalities for Scientific Visualization", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-04-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "This study compares specialized, computer-use and coding agents on scientific visualization workflows. It examines GUI, CLI, code and tool-call interaction modes, including how persistent memory changes performance across repeated trials.", "topics": ["science", "evaluation", "tool-integration", "memory"], "updated_at": "2026-06-26", "url": "https://arxiv.org/abs/2604.27996"}
{"aliases": [], "authors": ["Jia Yu", "Zilong Wang", "Xinyang Jiang", "Dongsheng Li", "Shuo Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "b23c12cafdd2d4a3ef63027c914ed3894c61c555edd89468c56ae40c3ec6cea9", "submitted_at": "2026-06-02T06:02:35Z", "updated_at": "2026-06-02T06:02:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.03203v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:36Z", "role": "primary", "sha256": "aeacc1a41bb90f3d296823356174fe77d1547ffbe7cd56652cf9d98fca09206c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=200&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2606.03203", "identifiers": {"arxiv": "2606.03203"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:36Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.03203"}], "name": "MedCUA-Bench: A Screenshot-Only Benchmark for Clinical Computer-Use Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-06-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "MedCUA-Bench evaluates screenshot-only clinical interface workflows using paired intent- and step-level goals. It combines reconstructed medical interfaces with completion and clinical-safety checks, and separately reports results on a real OpenEMR system.", "topics": ["evaluation", "safety-security", "science", "workflow"], "updated_at": "2026-06-02", "url": "https://arxiv.org/abs/2606.03203"}
{"aliases": [], "authors": ["Jiaxin Ai", "Tao Hu", "Xuemeng Yang", "Shu Zou", "Hairong Zhang", "Daocheng Fu", "Yu Yang", "Hongbin Zhou", "Nianchen Deng", "Pinlong Cai", "Zhongyuan Wang", "Botian Shi", "Kaipeng Zhang", "Licheng Wen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "edf7e63803efba89551aa297bff1da9b0babf8ec898a2971fa6c8a20cc58dea8", "submitted_at": "2026-06-11T11:53:32Z", "updated_at": "2026-06-30T03:31:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.13239v2"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:05Z", "role": "primary", "sha256": "2e5722527f8da4b2cf06372d1e2daccd33ef7fdc7151704b411a0bddd734beae", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2606.10423%2C2606.10522%2C2606.10742%2C2606.11078%2C2606.11520%2C2606.12485%2C2606.12666%2C2606.13239%2C2606.13385%2C2606.14106%2C2606.14832%2C2606.15017%2C2606.15034%2C2606.16659%2C2606.16748%2C2606.16802%2C2606.17321%2C2606.17645%2C2606.17929%2C2606.18448%2C2606.18671%2C2606.18890%2C2606.19388%2C2606.19926%2C2606.19930%2C2606.20363%2C2606.20636%2C2606.20717%2C2606.20724%2C2606.20785%2C2606.21654%2C2606.22864%2C2606.22948%2C2606.23449%2C2606.24515%2C2606.24525%2C2606.25705%2C2606.25760%2C2606.27330%2C2606.27595%2C2606.29445%2C2606.29537%2C2606.29705%2C2606.30697%2C2606.31154%2C2606.31270%2C2606.31410%2C2606.31612%2C2607.00502%2C2607.04425&max_results=50"}], "id": "arxiv:2606.13239", "identifiers": {"arxiv": "2606.13239"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:05Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2606.13239"}], "name": "ComAct: Reframing Professional Software Manipulation via COM-as-Action Paradigm", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-06-11", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "ComAct proposes COM program synthesis as an alternative interface for manipulating professional software and introduces ComCADBench with GUI-based comparison agents. ComActor adds a staged training and correction approach, supported by a Windows-container training platform.", "topics": ["automation", "evaluation", "tool-integration", "science"], "updated_at": "2026-06-30", "url": "https://arxiv.org/abs/2606.13239"}
{"aliases": [], "authors": ["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": ["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 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": ["Mei Wu", "Rui Xie", "Runyu Zhang", "Yuqiang Li", "Tianfan Fu", "Bo Chen", "Kai Yu", "Xin Chen", "Lu Chen"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e032f67257a52dbc59310d921d802e4dc85af3af110d22bb6d005d202bbaa6b2", "submitted_at": "2026-09-29T08:58:51Z", "updated_at": "2026-09-29T08:58:51Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.37053v1"}}, "environments": ["desktop", "windows"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}], "id": "arxiv:2609.37053", "identifiers": {"arxiv": "2609.37053"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.37053"}], "name": "MatToolBench: Benchmarking Multimodal Agents in Real-World Materials Science Workflows", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces MatToolBench, a Windows-based materials-science benchmark combining GUI operation, scientific-software scripting, and database queries. Expert-defined subcriteria support partial-credit scoring, while the evaluation examines domain knowledge gaps and cross-tool handoff failures.", "topics": ["science", "evaluation", "cross-app"], "updated_at": "2026-09-29", "url": "https://arxiv.org/abs/2609.37053"}
{"aliases": [], "authors": ["Hongcheng Gao", "Hailong Qu", "Yu Lei", "Henghui Sun", "Haoyang Li", "Yipeng Wei", "Naihao Xue", "Xiaohan Yu", "Zhuo Tao", "Yihe Zang", "Yajiao Wang", "Jingyi Tang", "Yi Li", "Jingjing Zhou", "Jie Luo", "Bohan Zeng", "Chengyu Shen", "Hao Jiang", "Chong Chen", "Bowen Qu", "Olive Huang", "Zeqiang Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aab7f586cb58a1f2d20267fd810ba34d6d4d166e3da715d2b3ff0b77e464473e", "submitted_at": "2026-09-29T14:32:35Z", "updated_at": "2026-10-09T08:15:21Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2609.37686v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:20Z", "role": "primary", "sha256": "c420ee40cc87a574ae2554556843f27f2eda43af553f88e95affd6319b1dae33", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.34321%2C2609.34769%2C2609.34974%2C2609.35026%2C2609.35671%2C2609.35814%2C2609.35904%2C2609.36201%2C2609.36488%2C2609.36923%2C2609.36927%2C2609.36939%2C2609.37053%2C2609.37176%2C2609.37686%2C2609.38008%2C2609.38397%2C2609.38778%2C2609.38867%2C2609.39306%2C2609.39547%2C2609.39903%2C2609.40253%2C2609.40284%2C2610.00948%2C2610.01215%2C2610.01491%2C2610.01787%2C2610.02320%2C2610.02932%2C2610.03036%2C2610.03089%2C2610.03574%2C2610.03818%2C2610.04116%2C2610.04800%2C2610.04906%2C2610.05185%2C2610.05437%2C2610.05861%2C2610.06829%2C2610.07118%2C2610.07127%2C2610.07444%2C2610.07972%2C2610.08699%2C2610.08773%2C2610.09240%2C2610.09469%2C2610.11050&max_results=50"}], "id": "arxiv:2609.37686", "identifiers": {"arxiv": "2609.37686"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:20Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.37686"}], "name": "EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments?", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces professional engineering tasks across software domains, with GUI and CLI access and artifact checks for geometric, physical, and design validity.", "topics": ["evaluation", "science", "cross-app"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2609.37686"}
{"aliases": [], "authors": ["Dingyuan Dai", "Heli Qi", "Lei Liu", "Yinxi Li", "Baiding Chen", "Zijun Dou", "Qingcheng Zeng", "Qi Kang", "Oliver Sun", "Eric Wang", "Bo Zhou", "Haixin Wang", "Yufan Du", "Shi Bo", "Ruihan Lin", "Mengqi Yuan", "Dunjie Lu", "Steven Dillmann", "Yiming Shi", "Tina Su", "Amy Xin", "Minghao Liu", "Xi Wang", "Xu Huang", "Ge Zhang", "Pengyu Nie", "Zhen Yang", "Jie Tang", "Juanzi Li", "Weihao Xuan", "Tianyu Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "690cd75934bae65311240cd4e6db7b8fd607056bd60799bf70cddd03efda9261", "submitted_at": "2026-09-30T15:03:05Z", "updated_at": "2026-09-30T15:03:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.39903v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.39903", "identifiers": {"arxiv": "2609.39903"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.39903"}], "name": "OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-30", "relations": [], "review_status": "curated", "scope": "core", "summary": "Benchmarks computer agents on scientific-software workflows with task-specific checks of application states and generated research artifacts.", "topics": ["evaluation", "science"], "updated_at": "2026-09-30", "url": "https://arxiv.org/abs/2609.39903"}
{"aliases": [], "authors": ["Ziyang Long", "Xinqi Li", "Lujing Xing", "Hsin-Jung Yang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "baacd8b4fad7b146bb9152a54d60a617fdbed5d0142c0c0768dacea2e7f18e60", "submitted_at": "2026-10-03T22:56:09Z", "updated_at": "2026-10-03T22:56:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.04800v1"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.04800", "identifiers": {"arxiv": "2610.04800"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.04800"}], "name": "MedImageOSWorld: Benchmarking GUI Agents for Medical Image Consoles", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-03", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates GUI agents in simulated medical-imaging consoles, separately checking workflow completion and acquisition outcomes against hidden anatomical targets.", "topics": ["evaluation", "science", "planning"], "updated_at": "2026-10-03", "url": "https://arxiv.org/abs/2610.04800"}
{"aliases": [], "authors": [], "details": {"benchmark_protocol": {"actions": ["PyAutoGUI desktop actions", "code/tool commands and answer/stop actions according to configured agent mode"], "counts": [{"basis": "computed", "evidence_url": "https://api.github.com/repos/OS-Copilot/ScienceBoard/git/trees/main?recursive=1", "population": "tasks/VM/**/*.json at tree 21b5eb0f; excludes tasks/Raw and tasks/Dual", "qualifier": "exact", "unit": "VM task definitions", "value": 169}, {"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/verifier-audit.md", "population": "VM suite: Celestia, ChimeraX, GrassGIS, KAlgebra, Lean, TeXstudio", "qualifier": "exact", "unit": "application families", "value": 6}], "evaluation": "Application-specific verifiers inspect state, files, database values, equations, compilation or answer/stop conditions. The official audit records 242 verifier entries across its VM task suite. This profile covers the released VM task definitions; Docker smoke results and Raw/Dual alternatives are separate populations.", "evidence_urls": ["https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/verifier-audit.md", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/known-issues.md", "https://api.github.com/repos/OS-Copilot/ScienceBoard/git/trees/main?recursive=1", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/sci/base/task.py"], "focus": ["scientific-software desktop workflows", "GUI and code-assisted scientific tasks"], "notes": "Manifest count 169 matches the official verifier audit. Other Raw/Dual definitions can represent alternative execution modes of the same problems and are not added. Official Docker migration documents a 20-task acceptance subset, not a full-suite performance result; legacy VMware remains. External service sign-in/rate limits, amd64 binaries and ChimeraX redistribution terms limit portability. Only code MIT is established; VM, scientific data and bundled apps need their own terms.", "observations": ["desktop screenshots", "accessibility-tree observations", "set-of-marks with accessibility tree"], "setting": "interactive"}}, "environments": ["desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "4c3fe93aa740dbf4a279a246ddd836f2590a2a36f39b820abe166e38cb484577", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "d276e11127abc5826bd1b2aa2caf543fcd45a8a60477774e8c8b1054067997ec", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/verifier-audit.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:21:10Z", "role": "primary", "sha256": "4f90270617c0c1d58f523c0a7f2e4603fa67a49560d68e2f88cfa080d764313b", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/docs/known-issues.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:14:07Z", "role": "primary", "sha256": "3398670a517a5d1d71f420dd866493d3f8a77853e3933fd4928103f9a8be1e4f", "supports": ["details.benchmark_protocol"], "url": "https://api.github.com/repos/OS-Copilot/ScienceBoard/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T14:35:24Z", "role": "primary", "sha256": "d6a1481b453c074e34739fd29def56efa35a1dc10e856418f557f067e9ff08c5", "supports": ["details.benchmark_protocol"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/sci/base/task.py"}, {"method": "license_header", "retrieved_at": "2026-10-16T06:41:59Z", "role": "primary", "sha256": "cbb1d28b515a39c294963d4bfc24077e669a483fc5ebba9b84a727afd4187426", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/LICENSE"}], "id": "github:os-copilot/scienceboard", "identifiers": {}, "kind": "benchmark", "last_verified_at": "2026-10-16T06:35:58Z", "licenses": {"code": {"basis": "repository LICENSE header; not a weights/data license", "declared": "MIT", "retrieved_at": "2026-10-16T06:41:59Z", "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/LICENSE"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OS-Copilot/ScienceBoard"}], "name": "ScienceBoard", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md", "target_id": "arxiv:2505.19897", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Scientific-workflow tasks for multimodal agents in desktop software such as ChimeraX, KAlgebra, Celestia, Grass GIS and Lean 4, with apps modified to expose state for evaluation.", "topics": ["evaluation", "science"], "updated_at": null, "url": "https://github.com/OS-Copilot/ScienceBoard"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [], "evidence_urls": ["https://huggingface.co/api/datasets/OS-Copilot/ScienceBoard-Traj", "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md"], "formats": ["ZIP"], "modalities": ["screenshots", "accessibility trees", "text"], "notes": "Observation modes follow publisher archive labels rather than inspected record schemas. No count or training-use permission is inferred.", "splits": [], "supervision": ["evaluation rollout logs"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/OS-Copilot/ScienceBoard-Traj/tree/main", "gated": false, "hub_id": "OS-Copilot/ScienceBoard-Traj", "languages": [], "modalities": ["image", "text"], "size_category_reported": null}, "environments": ["desktop", "linux"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "0a97233bd9132fdddc6ddb26fd3cf9f2af06ae086429f367a61cfab0c16c4f44", "supports": ["name", "kind", "identity", "release_status", "details"], "url": "https://huggingface.co/api/datasets/OS-Copilot/ScienceBoard-Traj"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:35:58Z", "role": "primary", "sha256": "4c3fe93aa740dbf4a279a246ddd836f2590a2a36f39b820abe166e38cb484577", "supports": ["summary", "environments", "relations", "topics", "scope"], "url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "8ee436f67e708bb917992764bc60a017c9db2edb21bdb99356b35c69f73724d7", "supports": ["environments", "notes"], "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md"}, {"method": "huggingface_api", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "0a97233bd9132fdddc6ddb26fd3cf9f2af06ae086429f367a61cfab0c16c4f44", "supports": ["details", "licenses.data"], "url": "https://huggingface.co/api/datasets/OS-Copilot/ScienceBoard-Traj"}], "id": "url:4784f63cd7117f64a2a3", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-17T04:07:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/OS-Copilot/ScienceBoard-Traj"}], "name": "ScienceBoard evaluation trajectories", "notes": "The manifest exposes six ZIP archives but no README card or data-license declaration. The official project announces evaluation-result trajectories; this record does not label them as a training corpus or infer a trajectory count from archive names. Internal schemas and complete payload contents were not inspected.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OS-Copilot/ScienceBoard/HEAD/README.md", "target_id": "arxiv:2505.19897", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Archives of ScienceBoard agent evaluation runs, separated by agent configuration and screenshot, accessibility-tree or set-of-marks observation mode.", "topics": ["evaluation", "science", "workflow"], "updated_at": null, "url": "https://huggingface.co/datasets/OS-Copilot/ScienceBoard-Traj"}
{"aliases": [], "authors": [], "details": {"artifact_type": "Ubuntu VMware image", "disabled": false, "download_url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/tree/main", "gated": false, "hub_id": "OS-Copilot/ScienceBoard-Env", "license_review": {"weights": {"evidence_urls": ["https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md", "https://huggingface.co/api/models/OS-Copilot/ScienceBoard-Env"], "reason": "This resource is a prebuilt Ubuntu VMware environment archive (VM.zip), not neural model weights. MIT Hub metadata in a model namespace cannot establish a weights license or relicense the guest OS and bundled applications."}}}, "environments": ["linux", "desktop"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:19Z", "role": "primary", "sha256": "8ee436f67e708bb917992764bc60a017c9db2edb21bdb99356b35c69f73724d7", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details", "licenses.weights"], "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T04:07:18Z", "role": "primary", "sha256": "9ffaf62b54a610afe186b3576b55fdd8287289e0cdfb88371bd9849a616f8460", "supports": ["identity", "release_status", "details", "licenses.weights"], "url": "https://huggingface.co/api/models/OS-Copilot/ScienceBoard-Env"}], "id": "url:75917f2c2b2802aafd3e", "identifiers": {}, "kind": "infrastructure", "last_verified_at": "2026-10-17T04:07:19Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env"}], "name": "ScienceBoard VMware environment", "notes": "The public Hub manifest contains VM.zip. This is an environment image, not neural model weights despite its Hub namespace. The card declares MIT metadata for the hosted release, but that does not relicense Ubuntu or included applications. The archive was not downloaded or booted; VMware’s own terms are separate.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env/raw/main/README.md", "target_id": "arxiv:2505.19897", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "A prebuilt Ubuntu virtual-machine archive containing the software environment used for ScienceBoard scientific computer-use tasks.", "topics": ["sandbox", "science", "evaluation"], "updated_at": null, "url": "https://huggingface.co/OS-Copilot/ScienceBoard-Env"}