{"aliases": [], "authors": ["Andrea Burns", "Kate Saenko", "Bryan A. Plummer"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "47d0ca0467a104f9b8be037ec73131f12b66a22b0a89fe975bec7680073c8f39", "submitted_at": "2024-06-12T02:43:19Z", "updated_at": "2024-08-08T02:36:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2406.07822v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:28Z", "role": "primary", "sha256": "be244bfae7f89ef9449b69d85aac44295eb7b5f7e5e1bbedd76df312696676a7", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2403.03186%2C2403.03788%2C2403.07718%2C2403.10170%2C2403.11905%2C2403.16971%2C2403.17918%2C2403.20271%2C2404.01549%2C2404.01744%2C2404.02475%2C2404.03648%2C2404.05427%2C2404.05719%2C2404.05955%2C2404.06474%2C2404.07972%2C2404.08755%2C2404.08860%2C2404.08948%2C2404.09992%2C2404.10887%2C2404.11459%2C2404.16006%2C2404.16048%2C2404.16054%2C2404.16660%2C2404.18074%2C2404.19296%2C2405.00516%2C2405.11120%2C2405.12842%2C2405.14573%2C2405.14751%2C2405.15793%2C2405.16247%2C2405.20309%2C2406.01014%2C2406.01623%2C2406.03679%2C2406.06947%2C2406.07822%2C2406.08184%2C2406.08340%2C2406.08451%2C2406.10227%2C2406.10819%2C2406.11317%2C2406.11896%2C2406.12373&max_results=50"}], "id": "arxiv:2406.07822", "identifiers": {"arxiv": "2406.07822"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:28Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2406.07822"}], "name": "Tell Me What's Next: Textual Foresight for Generic UI Representations", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-06-12", "relations": [], "review_status": "curated", "scope": "core", "summary": "Textual Foresight learns UI representations by describing the future screen expected after an action on the current interface. The OpenApp dataset supports pretraining and downstream evaluation of representations that connect individual elements with whole-screen changes.", "topics": ["screen-understanding", "world-models", "data-generation"], "updated_at": "2024-08-08", "url": "https://arxiv.org/abs/2406.07822"}
{"aliases": [], "authors": ["Hyungjoo Chae", "Namyoung Kim", "Kai Tzu-iunn Ong", "Minju Gwak", "Gwanwoo Song", "Jihoon Kim", "Sunghwan Kim", "Dongha Lee", "Jinyoung Yeo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "cae86bff4412bbfa7b969ecd60fe7b695f4b7c5ea1be6a67e97919df762d1108", "submitted_at": "2024-10-17T05:37:00Z", "updated_at": "2025-03-29T08:59:09Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2410.13232v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2410.13232", "identifiers": {"arxiv": "2410.13232"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2410.13232"}], "name": "Web Agents with World Models: Learning and Leveraging Environment Dynamics in Web Navigation", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-10-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses a learned web world model to predict important state changes and compare candidate actions before executing them.", "topics": ["world-models", "planning"], "updated_at": "2025-03-29", "url": "https://arxiv.org/abs/2410.13232"}
{"aliases": [], "authors": ["Yu Gu", "Kai Zhang", "Yuting Ning", "Boyuan Zheng", "Boyu Gou", "Tianci Xue", "Cheng Chang", "Sanjari Srivastava", "Yanan Xie", "Peng Qi", "Huan Sun", "Yu Su"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9ecaca6448aae5c8849d008d820d700bfa438ed0248f85be585b0b831408b9b6", "submitted_at": "2024-11-10T18:50:51Z", "updated_at": "2025-04-01T05:04:47Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2411.06559v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:58Z", "role": "primary", "sha256": "c87e97eb397da04806d53145cf9c13be739fc657bb54dca9318c8c4ed2a36249", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=800&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2411.06559", "identifiers": {"arxiv": "2411.06559"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:58Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2411.06559"}], "name": "Is Your LLM Secretly a World Model of the Internet? Model-Based Planning for Web Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2024-11-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebDreamer, which simulates candidate web actions with language-model world models and value estimates before acting.", "topics": ["world-models", "planning"], "updated_at": "2025-04-01", "url": "https://arxiv.org/abs/2411.06559"}
{"aliases": [], "authors": ["Jiani Zheng", "Lu Wang", "Fangkai Yang", "Chaoyun Zhang", "Lingrui Mei", "Wenjie Yin", "Qingwei Lin", "Dongmei Zhang", "Saravan Rajmohan", "Qi Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6d69b006a298a591aacfe7f29df63f508113c0ec824ede82a0bb5bbb9cb66266", "submitted_at": "2025-02-26T07:52:02Z", "updated_at": "2025-02-26T07:52:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2502.18906v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2502.18906", "identifiers": {"arxiv": "2502.18906"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2502.18906"}], "name": "VEM: Environment-Free Exploration for Training GUI Agent with Value Environment Model", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-02-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains GUI agents with a frozen value model learned from offline interactions, guiding action exploration without predicting complete next observations.", "topics": ["reinforcement-learning", "world-models"], "updated_at": "2025-02-26", "url": "https://arxiv.org/abs/2502.18906"}
{"aliases": [], "authors": ["Dezhao Luo", "Bohan Tang", "Kang Li", "Georgios Papoudakis", "Jifei Song", "Shaogang Gong", "Jianye Hao", "Jun Wang", "Kun Shao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "15e00237ef710c37c0c9f7768d699f57e9d5dcc190e8de7e20e9ec3b21b82750", "submitted_at": "2025-04-15T14:03:10Z", "updated_at": "2025-05-20T14:11:37Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.13936v2"}}, "environments": ["unspecified"], "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:2504.13936", "identifiers": {"arxiv": "2504.13936"}, "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/2504.13936"}], "name": "ViMo: A Generative Visual GUI World Model for App Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-04-15", "relations": [], "review_status": "curated", "scope": "core", "summary": "ViMo predicts the next mobile-app screenshot after a candidate action by separately modeling graphics and text. A symbolic placeholder representation preserves layout while a second predictor supplies readable text, enabling visual lookahead for GUI-agent planning.", "topics": ["world-models", "planning", "mobile-navigation"], "updated_at": "2025-05-20", "url": "https://arxiv.org/abs/2504.13936"}
{"aliases": [], "authors": ["Tianqing Fang", "Hongming Zhang", "Zhisong Zhang", "Kaixin Ma", "Wenhao Yu", "Haitao Mi", "Dong Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "86ca58c92ffc2cbbcb0297d8c7618dfab966e36a07b141c068998b0f2934d11e", "submitted_at": "2025-04-23T02:54:31Z", "updated_at": "2025-08-21T05:55:43Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2504.21024v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:54Z", "role": "primary", "sha256": "4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=700&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2504.21024", "identifiers": {"arxiv": "2504.21024"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:54Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2504.21024"}], "name": "WebEvolver: Enhancing Web Agent Self-Improvement with Coevolving World Model", "notes": "The annotation was reviewed against the complete retained primary abstract. Experiments and released artifacts have not been independently reproduced.", "published_at": "2025-04-23", "relations": [], "review_status": "curated", "scope": "core", "summary": "WebEvolver trains a co-evolving world model to predict web observations and generate additional agent-training trajectories. The same model supports look-ahead action selection at inference, with evaluation on live web-task benchmarks.", "topics": ["world-models", "data-generation", "web-navigation", "planning"], "updated_at": "2025-08-21", "url": "https://arxiv.org/abs/2504.21024"}
{"aliases": [], "authors": ["Xiaoran Yin", "Xu Luo", "Hao Wu", "Lianli Gao", "Jingkuan Song"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "93c67222fb0f581e9f9cb3ff59c8d28faec9dc037180060ad5ac962b4469144c", "submitted_at": "2025-05-22T09:08:47Z", "updated_at": "2025-09-05T02:24:00Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2505.16422v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:21:52Z", "role": "primary", "sha256": "c194d15b7e422aae48b4e05216a1f1d84471b6666718b9d3dd4c9d250c356e77", "supports": ["name", "kind", "scope", "environments", "topics", "summary", "notes"], "url": "https://export.arxiv.org/api/query?id_list=2210.03347%2C2403.04984%2C2405.07090%2C2406.06282%2C2409.00088%2C2409.10811%2C2411.10943%2C2503.09241%2C2503.20279%2C2503.23350%2C2505.05440%2C2505.10593%2C2505.16422%2C2505.21279%2C2505.24878%2C2506.02153%2C2506.02839%2C2506.05079%2C2506.07976%2C2508.02046%2C2511.07885%2C2511.11248%2C2512.21720%2C2603.20351%2C2604.19299&max_results=25"}], "id": "arxiv:2505.16422", "identifiers": {"arxiv": "2505.16422"}, "kind": "paper", "last_verified_at": "2026-10-18T04:21:52Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.16422"}], "name": "Unlocking Smarter Device Control: Foresighted Planning with a World Model-Driven Code Execution Approach", "notes": "Reviewed against the complete official arXiv abstract. Experimental results and linked artifact availability have not been independently reproduced.", "published_at": "2025-05-22", "relations": [], "review_status": "curated", "scope": "core", "summary": "FPWC builds and refines a task-oriented world model before planning mobile-device actions. It executes plans as code and evaluates the approach in simulated environments and on real mobile devices.", "topics": ["world-models", "planning", "mobile-navigation"], "updated_at": "2025-09-05", "url": "https://arxiv.org/abs/2505.16422"}
{"aliases": [], "authors": ["Runliang Niu", "Jinglong Ji", "Yi Chang", "Qi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bb782466eccc7a271c89a4161c7b142f93f8f0d585cad39a1ccad048cf514426", "submitted_at": "2025-05-25T11:13:03Z", "updated_at": "2025-05-25T11:13:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2505.19095v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T07:57:47Z", "role": "primary", "sha256": "27144160e8981bb9f36b60c391689774913ff3f026eeeffc2923154c404a6529", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2504.03206%2C2504.03601%2C2504.06821%2C2504.09723%2C2504.11301%2C2504.11543%2C2504.12609%2C2504.14233%2C2504.14348%2C2504.14603%2C2504.16073%2C2504.16419%2C2504.17892%2C2504.18580%2C2504.20965%2C2505.00019%2C2505.00693%2C2505.01578%2C2505.02094%2C2505.03570%2C2505.03574%2C2505.05026%2C2505.05849%2C2505.06507%2C2505.07003%2C2505.07473%2C2505.10468%2C2505.10887%2C2505.12053%2C2505.12490%2C2505.12728%2C2505.13291%2C2505.14030%2C2505.14454%2C2505.14646%2C2505.15801%2C2505.15816%2C2505.17020%2C2505.18746%2C2505.19095%2C2505.19713%2C2505.20023%2C2505.20732%2C2505.20795%2C2505.21055%2C2505.22304%2C2505.22429%2C2505.23450%2C2505.23559%2C2506.00641&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:20:24Z", "role": "primary", "sha256": "645bd2ecbc013dab3cbc601d5a42011d8f7306915ea2ae06baf0767b3962e9ed", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/abs/2505.19095v1"}], "id": "arxiv:2505.19095", "identifiers": {"arxiv": "2505.19095"}, "kind": "paper", "last_verified_at": "2026-10-17T11:20:24Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2505.19095"}], "name": "ScreenExplorer: Training a Vision-Language Model for Diverse Exploration in Open GUI World", "notes": "Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.", "published_at": "2025-05-25", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains ScreenExplorer with reinforcement learning in dynamic GUI environments, using a world-model curiosity reward and experience-stream distillation to support sustained exploration.", "topics": ["reinforcement-learning", "world-models", "data-generation"], "updated_at": "2025-05-25", "url": "https://arxiv.org/abs/2505.19095"}
{"aliases": [], "authors": ["Zeyi Sun", "Ziyu Liu", "Yuhang Zang", "Yuhang Cao", "Xiaoyi Dong", "Tong Wu", "Dahua Lin", "Jiaqi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "268c38e6f11f11771a1251fc522112ff052557b5753dff6cadf2ba49c0ec04b2", "submitted_at": "2025-08-06T17:58:46Z", "updated_at": "2025-08-12T15:11:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2508.04700v2"}}, "environments": ["desktop"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:50Z", "role": "primary", "sha256": "26fb4922c4c0bb5c9069935ed0aef0a9f56938cf8ae1af9724894fa1409955fb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=600&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2508.04700", "identifiers": {"arxiv": "2508.04700"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:50Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2508.04700"}], "name": "SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "SEAgent explores unfamiliar software using a world-state assessor and an automatically generated task curriculum. It learns from failed and successful interactions through adversarial imitation and GRPO, then combines specialist experience into a generalist agent evaluated on five OSWorld software environments.", "topics": ["reinforcement-learning", "data-generation", "world-models", "imitation-learning"], "updated_at": "2025-08-12", "url": "https://arxiv.org/abs/2508.04700"}
{"aliases": [], "authors": ["Kai Mei", "Jiang Guo", "Shuaichen Chang", "Mingwen Dong", "Dongkyu Lee", "Xing Niu", "Jiarong Jiang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7c37dae372655e8ef556163744e47ed42db3533175ca7d1db7560033774ed2a0", "submitted_at": "2025-10-13T19:52:04Z", "updated_at": "2026-03-09T15:25:16Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2510.11892v2"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:47Z", "role": "primary", "sha256": "ae98e77b8bb125e45277aacec768a047e7bb0e13f33ca9ff6ca7c19e29e56cca", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=500&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2510.11892", "identifiers": {"arxiv": "2510.11892"}, "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/2510.11892"}], "name": "R-WoM: Retrieval-augmented World Model For 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": "2025-10-13", "relations": [], "review_status": "curated", "scope": "core", "summary": "R-WoM evaluates how well world models predict nearby GUI states, align complete procedures, and recognize milestone transitions. It retrieves external tutorials to guide simulated future interactions, with experiments on subsets of OSWorld and WebArena.", "topics": ["world-models", "memory", "planning", "evaluation"], "updated_at": "2026-03-09", "url": "https://arxiv.org/abs/2510.11892"}
{"aliases": [], "authors": ["Shufan Li", "Konstantinos Kallidromitis", "Akash Gokul", "Yusuke Kato", "Kazuki Kozuka", "Aditya Grover"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e6228822cf64c6aeb1836bee52a6515123bd0d44d77cceb23e785bb4ff9aef47", "submitted_at": "2025-12-16T02:16:42Z", "updated_at": "2026-10-07T23:06:49Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2512.14014v2"}}, "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:2512.14014", "identifiers": {"arxiv": "2512.14014"}, "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/2512.14014"}], "name": "MobileWorldBench: Towards Semantic World Modeling For Mobile Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2025-12-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Evaluates semantic world models for mobile interfaces and studies natural-language state transitions as a tool for agent planning.", "topics": ["world-models", "planning", "evaluation"], "updated_at": "2026-10-07", "url": "https://arxiv.org/abs/2512.14014"}
{"aliases": [], "authors": ["Yilin Cao", "Yufeng Zhong", "Zhixiong Zeng", "Siran Dai", "Liming Zheng", "Jing Huang", "Haibo Qiu", "Peng Shi", "Wenji Mao"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c40e3db3f538f089c19d7c8233ea01174d0591c3211a7acc412ee4df28e902a8", "submitted_at": "2026-01-07T15:51:44Z", "updated_at": "2026-08-31T13:27:52Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.04035v2"}}, "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:2601.04035", "identifiers": {"arxiv": "2601.04035"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:47Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.04035"}], "name": "MobileDreamer: Generative Sketch World Model for GUI Agent", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-07", "relations": [], "review_status": "curated", "scope": "core", "summary": "MobileDreamer predicts post-action GUI states as task-relevant textual sketches with an order-invariant learning objective. Agents use these predictions for lookahead action selection, evaluated on AndroidWorld.", "topics": ["world-models", "planning", "mobile-navigation"], "updated_at": "2026-08-31", "url": "https://arxiv.org/abs/2601.04035"}
{"aliases": [], "authors": ["Hang Ding", "Peidong Liu", "Junqiao Wang", "Ziwei Ji", "Meng Cao", "Rongzhao Zhang", "Lynn Ai", "Eric Yang", "Tianyu Shi", "Lei Yu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aedb5ed90c91d72f62f5ab60be5374b6f431f887425bc760883ce3fb85a46d43", "submitted_at": "2026-01-29T18:59:07Z", "updated_at": "2026-04-17T18:27:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2601.22149v2"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2601.22149", "identifiers": {"arxiv": "2601.22149"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2601.22149"}], "name": "DynaWeb: Model-Based Reinforcement Learning of Web Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-01-29", "relations": [], "review_status": "curated", "scope": "core", "summary": "DynaWeb learns a world model that predicts web-page representations after actions and uses it to generate policy-training rollouts. It interleaves these rollouts with expert trajectories and evaluates the trained web agents on WebArena and WebVoyager.", "topics": ["world-models", "reinforcement-learning", "data-generation", "web-navigation"], "updated_at": "2026-04-17", "url": "https://arxiv.org/abs/2601.22149"}
{"aliases": [], "authors": ["Woosung Koh", "Sungjun Han", "Segyu Lee", "Se-Young Yun", "Jamin Shin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7c29d1404b02c4f5ef2cdaf78bdb0f9062b1b85ec1a154408bc8f49446c28d08", "submitted_at": "2026-02-02T03:12:16Z", "updated_at": "2026-05-23T02:41:44Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2602.01576v2"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:31:56Z", "role": "primary", "sha256": "c22be2d5c6a84b2c367759ce2b8f41f5a178fccd0534e832802c9d57b40918e8", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.16206%2C2601.17722%2C2601.18113%2C2601.18842%2C2601.20162%2C2601.20335%2C2601.20380%2C2601.20650%2C2601.21123%2C2601.22781%2C2602.00575%2C2602.01576%2C2602.01725%2C2602.02419%2C2602.02548%2C2602.02995%2C2602.03255%2C2602.03792%2C2602.05429%2C2602.05832%2C2602.06075%2C2602.06391%2C2602.07153%2C2602.07624%2C2602.08235%2C2602.08995%2C2602.09082%2C2602.09222%2C2602.09310%2C2602.09662%2C2602.10139%2C2602.11514%2C2602.11724%2C2602.11750%2C2602.14093%2C2602.14276%2C2602.14296%2C2602.14721%2C2602.16855%2C2602.17049%2C2602.17365%2C2602.20502%2C2602.21858%2C2602.22190%2C2603.05044%2C2603.05295%2C2603.07978%2C2603.08533%2C2603.10178%2C2603.12138&max_results=50"}], "id": "arxiv:2602.01576", "identifiers": {"arxiv": "2602.01576"}, "kind": "paper", "last_verified_at": "2026-10-16T06:31:56Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.01576"}], "name": "Generative Visual Code Mobile World Models", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "gWorld predicts future mobile GUI states as renderable web code and supplies an automated pipeline for code-based training data. The study evaluates visual state prediction and how stronger world models affect downstream mobile policies.", "topics": ["world-models", "data-generation", "screen-understanding", "mobile-navigation"], "updated_at": "2026-05-23", "url": "https://arxiv.org/abs/2602.01576"}
{"aliases": [], "authors": ["Yurun Chen", "Zeyi Liao", "Ping Yin", "Taotao Xie", "Keting Yin", "Shengyu Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d1a6f98c0ab0a221c169796d9680e4b2a7bf7a0775c0493254f47d5e0a9b3c4e", "submitted_at": "2026-02-02T07:04:06Z", "updated_at": "2026-02-02T07:04:06Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.01725v1"}}, "environments": ["desktop", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "topics", "scope"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T11:22:38+00:00", "role": "primary", "sha256": "6e1266c5a47cd667d681d3143a0331b74b6f64416b7e01f44a93a001c38ef74b", "supports": ["summary", "environments", "scope"], "url": "https://arxiv.org/html/2602.01725v1"}], "id": "arxiv:2602.01725", "identifiers": {"arxiv": "2602.01725"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.01725"}], "name": "SafePred: A Predictive Guardrail for Computer-Using Agents via World Models", "notes": "The annotation was reviewed against the complete retained official abstract and the specified sections of the same recorded paper version. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "SafePred uses world-model risk forecasts to guide action filtering and task replanning. Its desktop and web safety evaluations distinguish policy compliance, task success and success under the specified policy.", "topics": ["safety-security", "world-models", "planning", "evaluation"], "updated_at": "2026-02-02", "url": "https://arxiv.org/abs/2602.01725"}
{"aliases": [], "authors": ["Haodong Li", "Jingwei Wu", "Quan Sun", "Guopeng Li", "Juanxi Tian", "Ziyu Guo", "Donghao Zhou", "Tianfei Ren", "Huanyu Zhang", "Yanlin Lai", "Ruichuan An", "Hongbo Peng", "Yuhong Dai", "Jia Wang", "Ziyang Meng", "Zheng Ge", "Xiangyu Zhang", "Daxin Jiang", "Pheng-Ann Heng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "43a01e0eb848a57a84767836d2c3c982bbf6094d22a3e70652dd42af52760bc5", "submitted_at": "2026-02-09T18:52:02Z", "updated_at": "2026-09-26T18:28:57Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2602.09007v3"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T12:31:12Z", "role": "primary", "sha256": "c8c95278aa5dcc6a8f2eda3f3b44d61ec1f35adcb6a4c6c9ab328691a1a4f102", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2601.13462%2C2601.14232%2C2601.15331%2C2601.17548%2C2601.18418%2C2601.18491%2C2601.18827%2C2601.20171%2C2601.20613%2C2601.20742%2C2601.21233%2C2601.21253%2C2601.22037%2C2601.22154%2C2601.22511%2C2601.22607%2C2602.01550%2C2602.03117%2C2602.03361%2C2602.05665%2C2602.05746%2C2602.05877%2C2602.06052%2C2602.07398%2C2602.07458%2C2602.07918%2C2602.09007%2C2602.09463%2C2602.09514%2C2602.10122%2C2602.10525%2C2602.11224%2C2602.12268%2C2602.12430%2C2602.13318%2C2602.13543%2C2602.13692%2C2602.14038%2C2602.14337%2C2602.14364%2C2602.16313%2C2602.16317%2C2602.16666%2C2602.16901%2C2602.16930%2C2602.17547%2C2602.17646%2C2602.18846%2C2602.19320%2C2602.19594&max_results=50"}], "id": "arxiv:2602.09007", "identifiers": {"arxiv": "2602.09007"}, "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/2602.09007"}], "name": "GUI-GenBench: Evaluating Image Generation Models as Interactive GUI Environments", "notes": "Reviewed against the complete retained official abstract. Experimental findings and release availability have not been independently reproduced or verified.", "published_at": "2026-02-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "GUI-GenBench tests image-generation models as action-conditioned GUI environments across five suites, including transitions, trajectories and grounding. Its VLM-based GUI-Score separates goal completion, interaction logic, consistency and visual properties.", "topics": ["evaluation", "world-models", "grounding", "screen-understanding"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2602.09007"}
{"aliases": [], "authors": ["Yuhao Zheng", "Li'an Zhong", "Yi Wang", "Rui Dai", "Kaikui Liu", "Xiangxiang Chu", "Linyuan Lv", "Philip Torr", "Kevin Qinghong Lin"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "55287e2fe8a5e94671e4aab8e9d4a907ba9b6c83306d9b3f61b74adccd5c5059", "submitted_at": "2026-02-10T14:56:19Z", "updated_at": "2026-02-10T14:56:19Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.09856v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:32:59Z", "role": "primary", "sha256": "d494c76e512656f76a5912eac0768bbbdd50014387c779a0618a40c1722c399b", "supports": ["name", "kind", "summary", "environments", "topics", "scope"], "url": "https://export.arxiv.org/api/query?id_list=2602.07008%2C2602.07283%2C2602.07787%2C2602.09856%2C2602.10356%2C2602.10814%2C2602.10915%2C2602.11524%2C2602.11858%2C2602.12276%2C2602.12544%2C2602.12852%2C2602.13559%2C2602.13653%2C2602.14865%2C2602.15384%2C2602.17003%2C2602.17245%2C2602.17588%2C2602.24142%2C2603.00503%2C2603.00676%2C2603.02951%2C2603.04364%2C2603.04949%2C2603.07024%2C2603.07432%2C2603.08013%2C2603.08316%2C2603.08938%2C2603.10268%2C2603.10291%2C2603.10505%2C2603.10577%2C2603.12710%2C2603.14248%2C2603.14448%2C2603.17441%2C2603.17826%2C2603.20278%2C2603.20340%2C2603.22529%2C2603.23007%2C2603.23559%2C2603.24160%2C2603.24533%2C2603.26041%2C2603.26266%2C2604.00073%2C2604.00842&max_results=50"}], "id": "arxiv:2602.09856", "identifiers": {"arxiv": "2602.09856"}, "kind": "paper", "last_verified_at": "2026-10-16T06:32:59Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.09856"}], "name": "Code2World: A GUI World Model via Renderable Code Generation", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and released artifacts have not been independently reproduced.", "published_at": "2026-02-10", "relations": [], "review_status": "curated", "scope": "core", "summary": "Code2World predicts future GUI screens through renderable code, trained on AndroidCode trajectories with supervised learning and render-aware reinforcement learning. It evaluates visual prediction and downstream AndroidWorld navigation with the learned world model.", "topics": ["world-models", "data-generation", "reinforcement-learning", "mobile-navigation"], "updated_at": "2026-02-10", "url": "https://arxiv.org/abs/2602.09856"}
{"aliases": [], "authors": ["Zikai Xiao", "Jianhong Tu", "Chuhang Zou", "Yuxin Zuo", "Zhi Li", "Peng Wang", "Bowen Yu", "Fei Huang", "Junyang Lin", "Zuozhu Liu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2dca0e4219a788205bbb060911d0f004d1db4cf2ff8a4962c04fd5777ea194d7", "submitted_at": "2026-02-16T13:06:49Z", "updated_at": "2026-02-16T13:06:49Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.14721v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.14721", "identifiers": {"arxiv": "2602.14721"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.14721"}], "name": "WebWorld: A Large-Scale World Model for Web Agent Training", "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": "2026-02-16", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces WebWorld, an open-web simulator trained on interaction trajectories for generating agent training data and testing candidate actions. WebWorld-Bench evaluates simulated transitions, alongside downstream web-agent training experiments.", "topics": ["world-models", "data-generation", "evaluation"], "updated_at": "2026-02-16", "url": "https://arxiv.org/abs/2602.14721"}
{"aliases": [], "authors": ["Yiming Guan", "Rui Yu", "John Zhang", "Lu Wang", "Chaoyun Zhang", "Liqun Li", "Bo Qiao", "Si Qin", "He Huang", "Fangkai Yang", "Pu Zhao", "Lukas Wutschitz", "Samuel Kessler", "Huseyin A Inan", "Robert Sim", "Saravan Rajmohan", "Qingwei Lin", "Dongmei Zhang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "3d2237dd43f4c876ba69b68c1871ca3efd3aa501e93c86655eb2de6d9f35d2e8", "submitted_at": "2026-02-19T13:48:29Z", "updated_at": "2026-02-19T13:48:29Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2602.17365v1"}}, "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", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2602.17365", "identifiers": {"arxiv": "2602.17365"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:43Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2602.17365"}], "name": "Computer-Using World Model", "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": "2026-02-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents a desktop world model that predicts textual UI changes before synthesizing the next screenshot. It learns from Microsoft Office transitions and is evaluated by letting a frozen agent compare simulated actions before execution.", "topics": ["world-models", "planning", "action-prediction"], "updated_at": "2026-02-19", "url": "https://arxiv.org/abs/2602.17365"}
{"aliases": [], "authors": ["Mengzhou Wu", "Yuzhe Guo", "Yuan Cao", "Haochuan Lu", "Songhe Zhu", "Pingzhe Qu", "Xin Chen", "Kang Qin", "Zhongpu Wang", "Xiaode Zhang", "Xinyi Wang", "Wei Dai", "Gang Cao", "Yuetang Deng", "Zhi Gong", "Dezhi Ran", "Linyi Li", "Wei Yang", "Tao Xie"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bac86c2ff2bf18feb3edfbe76285d33718d0c6d9564aff5a9d62f8778cdc5147", "submitted_at": "2026-02-11T17:02:38Z", "updated_at": "2026-02-11T17:02:38Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.02345v1"}}, "environments": ["unspecified"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:43Z", "role": "primary", "sha256": "c432f419481a105a8ce71174ec4aea23c4f0d36ddf89cfe1d0974fa349f6cb59", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=400&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2604.02345", "identifiers": {"arxiv": "2604.02345"}, "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/2604.02345"}], "name": "UI-Oceanus: Scaling GUI Agents with Synthetic Environmental Dynamics", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-02-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "UI-Oceanus converts autonomous GUI exploration and execution feedback into supervision for predicting future interface states. It compares forward and inverse objectives during continual pretraining and examines downstream offline and online navigation as synthetic data scales.", "topics": ["world-models", "data-generation", "state-tracking", "evaluation"], "updated_at": "2026-02-11", "url": "https://arxiv.org/abs/2604.02345"}
{"aliases": [], "authors": ["Weikai Xu", "Kun Huang", "Yunren Feng", "Jiaxing Li", "Yuhan Chen", "Yuxuan Liu", "Zhizheng Jiang", "Heng Qu", "Pengzhi Gao", "Wei Liu", "Jian Luan", "Xiaolin Hu", "Bo An"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4374ec1b7c44804bb78ea56bd0aa3214083419f37780d79c5fcfb68468727bbf", "submitted_at": "2026-05-11T10:49:31Z", "updated_at": "2026-05-22T05:43:30Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.10347v2"}}, "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:2605.10347", "identifiers": {"arxiv": "2605.10347"}, "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/2605.10347"}], "name": "How Mobile World Model Guides 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": "2026-05-11", "relations": [], "review_status": "curated", "scope": "core", "summary": "Compares mobile world models that predict state changes as delta text, full text, images or renderable code. Downstream Android benchmarks examine synthetic training trajectories and test-time guidance, including limits of post-action reflection.", "topics": ["world-models", "data-generation", "evaluation"], "updated_at": "2026-05-22", "url": "https://arxiv.org/abs/2605.10347"}
{"aliases": [], "authors": ["Xue Yu", "Bo Yuan", "Kailin Zhao", "Pengshuai Yang", "Hong Hu", "Junlan Feng"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6a15963a7fb4dfd488f0e79732d9ea0c9a03d43c1e3d5e50dd1ee8e00f1072be", "submitted_at": "2026-07-17T01:45:54Z", "updated_at": "2026-09-02T01:49:21Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2607.15550v3"}}, "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:2607.15550", "identifiers": {"arxiv": "2607.15550"}, "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/2607.15550"}], "name": "SeerGuard: A Safety Framework for Mobile GUI Agents via World Model Prediction", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses instruction screening and a safety-augmented world model to assess likely consequences of mobile GUI actions before execution.", "topics": ["safety-security", "world-models"], "updated_at": "2026-09-02", "url": "https://arxiv.org/abs/2607.15550"}
{"aliases": [], "authors": ["Yaohan Yang", "Minglei Shi", "Borui Zhang", "Jie Zhou", "Jiwen Lu"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "6488798626abbec89ca739a0c5d188ae440484eb59b3be11f2b319ce742b22ec", "submitted_at": "2026-07-19T03:29:08Z", "updated_at": "2026-07-19T03:29:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.17050v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:09Z", "role": "primary", "sha256": "65cc3856c70a0fb267f9996872e077de10e9f54821ab1ffb1d82f0a92fa11dbf", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2607.05277%2C2607.06118%2C2607.06306%2C2607.07820%2C2607.11185%2C2607.11388%2C2607.12625%2C2607.12640%2C2607.13027%2C2607.14277%2C2607.14443%2C2607.14548%2C2607.15193%2C2607.15550%2C2607.17050%2C2607.18046%2C2607.19865%2C2607.21557%2C2607.22798%2C2607.23263%2C2607.24112%2C2607.24167%2C2607.25904%2C2607.26041%2C2607.27155%2C2607.28074%2C2607.28227%2C2607.28609%2C2607.29320%2C2608.01964%2C2608.02352%2C2608.02643%2C2608.03270%2C2608.03327%2C2608.04530%2C2608.04741%2C2608.05587%2C2608.05784%2C2608.05891%2C2608.06065%2C2608.06171%2C2608.06477%2C2608.09278%2C2608.09654%2C2608.10366%2C2608.10692%2C2608.11588%2C2608.12429%2C2608.15930%2C2608.17310&max_results=50"}], "id": "arxiv:2607.17050", "identifiers": {"arxiv": "2607.17050"}, "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/2607.17050"}], "name": "EvoGUI: An Evolution-Aware Benchmark for GUI State-Transition Understanding", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-19", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces EvoGUI, which derives temporal ordering, inverse action/value prediction, and successor-discrimination probes from recorded GUI trajectories. Its web-derived benchmark isolates transition understanding from the other abilities involved in end-to-end task success.", "topics": ["evaluation", "world-models", "action-prediction"], "updated_at": "2026-07-19", "url": "https://arxiv.org/abs/2607.17050"}
{"aliases": [], "authors": ["Xiangyan Liu", "Kaixin Li", "Haonan Wang", "Biao Wu", "Meng Fang", "Longxu Dou", "Chao Du", "Michael Qizhe Shieh", "Tianyu Pang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "4651964cbe45e4d740ed8f789753f82ae808e442d82c6598e7b3c997c09453c3", "submitted_at": "2026-07-27T07:54:08Z", "updated_at": "2026-07-27T07:54:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.24112v1"}}, "environments": ["desktop", "android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2607.24112", "identifiers": {"arxiv": "2607.24112"}, "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/2607.24112"}], "name": "Scaling GUI Agents with Visual State Transitions", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-07-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces State Transition Pretraining, which jointly learns to infer actions from GUI changes and predict future states from current screens and actions. Subsequent instruction-trajectory fine-tuning tests whether these representations transfer to desktop and mobile agent tasks.", "topics": ["world-models", "action-prediction", "data-generation"], "updated_at": "2026-07-27", "url": "https://arxiv.org/abs/2607.24112"}
{"aliases": [], "authors": ["Weikai Xu", "Yunren Feng", "Haoxiang Lei", "Kun Huang", "Yuxuan Liu", "Kang Zhao", "Xiaolin Hu", "Shuo Shang", "Bo An"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ce37dd8217afbaa07ddea25c0daf78cdbfb449851ca6ab17b7f3ecacd451eb2f", "submitted_at": "2026-08-06T11:15:41Z", "updated_at": "2026-08-06T11:15:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.05891v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.05891", "identifiers": {"arxiv": "2608.05891"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.05891"}], "name": "AppDeltaWorld: Transition-Grounded Delta Code World Model for Mobile GUI Agents", "notes": "The annotation was reviewed against the complete retained official abstract. Experimental findings and artifact availability have not been independently reproduced.", "published_at": "2026-08-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "AppDeltaWorld predicts executable code updates for the next mobile UI state, using transition-constrained retrieval and rendered visual assets. It evaluates reconstruction fidelity separately from policy training and adaptation using generated environments.", "topics": ["world-models", "data-generation", "imitation-learning", "reinforcement-learning", "mobile-navigation"], "updated_at": "2026-08-06", "url": "https://arxiv.org/abs/2608.05891"}
{"aliases": [], "authors": ["Yu Han", "Tianwen Qian"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "576702fae164cbcb4adc86a81115487aae7e1c40cf07a1ecf3ef7c4c7e822c3b", "submitted_at": "2026-08-27T07:09:28Z", "updated_at": "2026-08-27T07:09:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.27508v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2608.27508", "identifiers": {"arxiv": "2608.27508"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2608.27508"}], "name": "WM-R1: Training GUI Agents to Reason and leverage World Models with Reinforcement Learning", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-08-27", "relations": [], "review_status": "curated", "scope": "core", "summary": "Uses learned world models for mobile-agent training rollouts and for reasoning about the consequences of candidate GUI actions.", "topics": ["world-models", "reinforcement-learning"], "updated_at": "2026-08-27", "url": "https://arxiv.org/abs/2608.27508"}
{"aliases": [], "authors": ["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": ["Kelvin Li", "Dhruv Pendharkar", "Anish Pahilajani", "Chuyi Shang", "Leon Oks", "Leonid Karlinsky", "Rogerio Feris", "Trevor Darrell", "Roei Herzig"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8703474cb4f577b307731cf1a986f76f800470fde625b6ca4b3569d30e74f903", "submitted_at": "2026-09-02T17:59:40Z", "updated_at": "2026-09-02T17:59:40Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.02885v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:32Z", "role": "primary", "sha256": "13d3fb5811d203f2e6b80eb45d30ebc0b3cfefd88e2034ffa3b1b80687bef285", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=100&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2609.02885", "identifiers": {"arxiv": "2609.02885"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:32Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.02885"}], "name": "Discriminative World Models for Web Agents", "notes": "The annotation and scope were reviewed against the complete retained original arXiv abstract. Experiments, reported performance and artifact availability have not been independently reproduced.", "published_at": "2026-09-02", "relations": [], "review_status": "curated", "scope": "core", "summary": "Trains discriminative web world models to distinguish the true next state from outcomes of alternative actions. Branching web trajectories support training, followed by tests of predicted-state matching, process-reward action ranking and end-to-end WebArena-Lite selection.", "topics": ["world-models", "reward-models", "planning", "data-generation", "evaluation"], "updated_at": "2026-09-02", "url": "https://arxiv.org/abs/2609.02885"}
{"aliases": [], "authors": ["Dongsheng Liu", "Chao Jin", "Wenkui Yang", "Hejin Wang", "Junwei Yang", "Zeren Zhang", "Ziwei Chen", "Huaibo Huang", "Jie Cao", "Ran He"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d641819d168eedc07270607da33f3197c15f2f90cf99f538c3839eed506f3308", "submitted_at": "2026-09-26T14:35:45Z", "updated_at": "2026-09-26T14:35:45Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.32679v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:33:17Z", "role": "primary", "sha256": "5d7c9ff24833b254d33a41a7c4247e830e4d9c371881a1e16151bad5c71cc8b3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?id_list=2609.10297%2C2609.10451%2C2609.12394%2C2609.12404%2C2609.13287%2C2609.13491%2C2609.13548%2C2609.13873%2C2609.14239%2C2609.15134%2C2609.15307%2C2609.15364%2C2609.15457%2C2609.16251%2C2609.16635%2C2609.16732%2C2609.16900%2C2609.17653%2C2609.17885%2C2609.18411%2C2609.18690%2C2609.19125%2C2609.19523%2C2609.19843%2C2609.22083%2C2609.22724%2C2609.24277%2C2609.24890%2C2609.25769%2C2609.27273%2C2609.27307%2C2609.27353%2C2609.27461%2C2609.29892%2C2609.30186%2C2609.30604%2C2609.30887%2C2609.31957%2C2609.32036%2C2609.32130%2C2609.32166%2C2609.32351%2C2609.32385%2C2609.32600%2C2609.32679%2C2609.32750%2C2609.32915%2C2609.33304%2C2609.33646%2C2609.34113&max_results=50"}], "id": "arxiv:2609.32679", "identifiers": {"arxiv": "2609.32679"}, "kind": "paper", "last_verified_at": "2026-10-16T06:33:17Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2609.32679"}], "name": "The GUI Is Not the State: Diagnosing State Aliasing in GUI World Models", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-09-26", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces StateAliasBench to isolate GUI world-model ambiguities in which matching visible screens and actions can lead to different futures. It augments frozen world models with structured state recovered from interaction history and evaluates the effect on prediction and agent execution.", "topics": ["world-models", "state-tracking", "evaluation"], "updated_at": "2026-09-26", "url": "https://arxiv.org/abs/2609.32679"}
{"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": ["Cheng Yang", "Yifan Wu", "Yutao Huang", "Zhaohua Zhang", "Beiduo Chen", "Muxi Chen", "Chenchen Zhao", "Hexuan Deng", "Haolin Yang", "Geyuan Zhu", "Sa Zhu", "Jianhuan Zhuo", "Qiuyong Xiao", "Jianhao Ruan", "Yiran Peng", "Jiayi Zhang", "Tian Ye", "Xinlei Yu", "Tianwen Jiang", "Jihong Zhang", "Yuyu Luo"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "ecd40b78c6de41abb442705b993aad8ebf7b35741536a11e00ebbd60140d1f18", "submitted_at": "2026-10-01T07:22:53Z", "updated_at": "2026-10-01T07:22:53Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.01215v1"}}, "environments": ["desktop", "linux", "windows", "macos", "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:2610.01215", "identifiers": {"arxiv": "2610.01215"}, "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/2610.01215"}], "name": "AutoGUIWorld: Image Generators as Visual World Models for GUI Agent", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-01", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces AutoGUIWorld, which combines task planning and image editing to synthesize action-grounded GUI trajectories without running the depicted software. It filters generated transitions and tests whether training on them improves agents operating real desktop and scientific applications.", "topics": ["world-models", "data-generation", "grounding"], "updated_at": "2026-10-01", "url": "https://arxiv.org/abs/2610.01215"}
{"aliases": [], "authors": ["Dhananjay Ashok", "Shantanu Agarwal", "Vivek Datla", "Jonathan May", "Alfy Samuel"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "35ded32e7bac9a875a0fa41efad43fed5fb0aab54fade5cfc2a0628b1e666587", "submitted_at": "2026-10-01T22:30:08Z", "updated_at": "2026-10-01T22:30:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.02542v1"}, "license_review": {"document": {"evidence_urls": ["https://arxiv.org/html/2610.02542v1"], "reason": "The retained version-specific arXiv HTML infobox explicitly labels this paper License: CC BY 4.0. This document declaration does not establish code, training-data or model-weight licenses."}}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T15:06:37Z", "role": "primary", "sha256": "a3d753815e82190ee2de6f322cd816c32f2dd391f862845305fc7e57300303a6", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28abs%3A%22desktop+agent%22+OR+abs%3A%22computer+interaction%22+OR+abs%3A%22GUI+automation%22+OR+abs%3A%22web+navigation%22+OR+abs%3A%22browser+automation%22+OR+abs%3A%22visual+grounding%22+OR+abs%3A%22screen+understanding%22+OR+abs%3A%22mobile+agents%22%29+AND+submittedDate%3A%5B202609200000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T15:08:56Z", "role": "primary", "sha256": "d8c22c6b3f07d4156c38af9dd7ca7c06a5d0a11889a7350d821a020e21604590", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "licenses.document"], "url": "https://arxiv.org/html/2610.02542v1"}], "id": "arxiv:2610.02542", "identifiers": {"arxiv": "2610.02542"}, "kind": "paper", "last_verified_at": "2026-10-20T15:08:56Z", "licenses": {"code": null, "data": null, "document": {"basis": "The retained version-specific arXiv HTML infobox explicitly labels this paper License: CC BY 4.0. This document declaration does not establish code, training-data or model-weight licenses.", "declared": "CC-BY-4.0", "retrieved_at": "2026-10-20T15:08:56Z", "url": "https://arxiv.org/html/2610.02542v1"}, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.02542"}], "name": "How To Train Your World Model: Fine-tuning vs RAG for LM-based World Modeling", "notes": "Adjacent cross-domain study with WebShop as its web evaluation. WebShop uses text-rendered search/product pages and an explicit valid-action menu, not screenshot control or live purchases. A frozen policy proposes five actions; a world model predicts their next observations and the policy scores them before one is executed in the simulator. Utility is normalized environment reward, while fidelity uses a model judge; neither is a universal GUI task-success rate. Fine-tuning and retrieval use collected training experience; the hybrid masks same-episode retrieval, which does not establish general contamination freedom. The retained Atom and v1 HTML abstract wording differs; original bibliographic metadata and its abstract digest are preserved.", "published_at": "2026-10-01", "relations": [], "review_status": "curated", "scope": "adjacent", "summary": "Compares fine-tuned, retrieval-based and hybrid language-model world models for next-observation prediction and action selection in text environments including WebShop.", "topics": ["world-models", "planning", "evaluation", "web-navigation"], "updated_at": "2026-10-01", "url": "https://arxiv.org/abs/2610.02542"}
{"aliases": [], "authors": ["Yongxin Ning", "Runliang Niu", "Qianli Xing", "Zhiyi Duan", "Qingzu He", "Pan Wang", "Qi Wang"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "43845ad64843a0b019c1988d4ba861eccee3c48a7b2c09d0bae7794e5e86347a", "submitted_at": "2026-10-05T06:28:58Z", "updated_at": "2026-10-05T06:28:58Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.05861v1"}}, "environments": ["android"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.05861", "identifiers": {"arxiv": "2610.05861"}, "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/2610.05861"}], "name": "Imagine to Act: High-Fidelity Data Synthesis via Image Editing World Model for Scalable GUI Agent Training", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-05", "relations": [], "review_status": "curated", "scope": "core", "summary": "Introduces Infinite-Dreamer, which learns action-conditioned screenshot editing from structured descriptions of GUI changes and uses it to synthesize robustness examples and multi-step trajectories. The synthetic data trains a mobile GUI agent evaluated on AndroidWorld, MobileWorld, and AndroidControl-Curated.", "topics": ["world-models", "data-generation", "mobile-navigation"], "updated_at": "2026-10-05", "url": "https://arxiv.org/abs/2610.05861"}
{"aliases": [], "authors": ["Sarim Hashmi", "Mukul Ranjan", "Kshitij Mishra", "Mikhail Kuznetsov", "Praneeth Vepakomma", "Nils Lukas"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7113025e0dfd784d6994108c24681000625b13598e33ad0e4324e90fd2832c19", "submitted_at": "2026-10-06T17:56:43Z", "updated_at": "2026-10-06T17:56:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.08773v1"}}, "environments": ["web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-16T06:36:29Z", "role": "primary", "sha256": "74e521e46c46489e48f9bad5d774625a85b5ccd32828969fd98e1a4c453b899f", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B200001010000+TO+202610160636%5D&start=0&max_results=100&sortBy=submittedDate&sortOrder=descending"}], "id": "arxiv:2610.08773", "identifiers": {"arxiv": "2610.08773"}, "kind": "paper", "last_verified_at": "2026-10-16T06:36:29Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.08773"}], "name": "AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-06", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents AdvSim2Real, which jointly adapts web-task curricula, prompt-injection adversaries, and agent policies inside a frozen web world model. It evaluates robustness against an unseen adversary and transfer from simulated interaction to a real browser.", "topics": ["safety-security", "world-models", "reinforcement-learning"], "updated_at": "2026-10-06", "url": "https://arxiv.org/abs/2610.08773"}
{"aliases": [], "authors": ["Jiaming Zhang", "Xuan Wang", "Fuyao Zhang", "Yang Cao", "Lingjuan Lyu", "Wei Yang Bryan Lim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "53120f960c7b98e87210c93ac661a2e397b858cc9f211786b0f8233fa61c6dd3", "submitted_at": "2026-10-08T13:33:23Z", "updated_at": "2026-10-08T13:33:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.11942v1"}}, "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:2610.11942", "identifiers": {"arxiv": "2610.11942"}, "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/2610.11942"}], "name": "Right Screen, Wrong Transition: World Models as Verifiers 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": "2026-10-08", "relations": [], "review_status": "curated", "scope": "core", "summary": "Presents LGWM, a decoder-free GUI world model that predicts the next screen representation from an action and compares it with the observed state to verify transitions. Its paired diagnostic separates legitimate and hijacked transitions that lead to the same credential screen.", "topics": ["world-models", "safety-security", "evaluation"], "updated_at": "2026-10-08", "url": "https://arxiv.org/abs/2610.11942"}
{"aliases": [], "authors": ["Seungjun Yu", "Hojun Choi", "Seojeong Park", "Jaeyo Shin", "Hyunjung Shim"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "191e88c19124ddaddc2fed2f0e3f8e197a4c0c7f8adbc3997c056d3333754042", "submitted_at": "2026-10-09T18:45:11Z", "updated_at": "2026-10-09T18:45:11Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.14228v1"}}, "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:2610.14228", "identifiers": {"arxiv": "2610.14228"}, "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/2610.14228"}], "name": "Latent World Action Model for GUI Agents", "notes": "The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.", "published_at": "2026-10-09", "relations": [], "review_status": "curated", "scope": "core", "summary": "Replaces future-screenshot generation with a shared latent representation for GUI action selection and action-conditioned transition prediction. An action-aware contrastive objective encourages the transition representation to retain discriminative action-induced changes.", "topics": ["world-models", "planning", "efficiency"], "updated_at": "2026-10-09", "url": "https://arxiv.org/abs/2610.14228"}
{"aliases": [], "authors": ["Sizhe Tang", "Hanhan Zhou", "Jeffrey Svajlenko", "Rashmi Gangadharaiah"], "details": {"arxiv_snapshot": {"abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9d6808cf26a609b50b7835ae3ecece0a13580aafcd589823f0b1eeff54732f69", "submitted_at": "2026-10-17T02:03:20Z", "updated_at": "2026-10-17T02:03:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.24372v1"}}, "environments": ["android", "web"], "evidence": [{"method": "arxiv_api", "retrieved_at": "2026-10-20T12:58:15.922053Z", "role": "primary", "sha256": "b379e14eefa272f62641224fe1d839b7dfbc414d957124b8abeb103444a843ef", "supports": ["name", "authors", "published_at", "updated_at", "identifiers.arxiv", "details.arxiv_snapshot"], "url": "https://export.arxiv.org/api/query?search_query=%28ti%3AGUI+OR+ti%3A%22computer-using%22+OR+ti%3A%22agentic+browsing%22+OR+ti%3A%22visual+grounding%22+OR+ti%3A%22UI+automation%22%29+AND+submittedDate%3A%5B202608210000+TO+202610201255%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-20T13:12:51.579373Z", "role": "primary", "sha256": "9dff18935a1faa7ec16366bac6a6e9e41f0381d561568c3aba39fe783efcfbd3", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes"], "url": "https://arxiv.org/html/2610.24372v1"}], "id": "arxiv:2610.24372", "identifiers": {"arxiv": "2610.24372"}, "kind": "paper", "last_verified_at": "2026-10-20T13:12:51.579373Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://arxiv.org/abs/2610.24372"}], "name": "Beyond Next-Screen Prediction: The Decision Value of GUI World Models", "notes": "The matched augmentation experiment measures offline ranking; AndroidWorld task-success comparisons use differently trained evaluators and do not isolate prediction value. A separate web evaluator is trained on Mind2Web, and web results do not uniformly favor direct scoring. The main study concerns one-step selection. No independent reproduction.", "published_at": "2026-10-17", "relations": [], "review_status": "curated", "scope": "core", "summary": "Separates next-screen prediction from action evaluation in GUI lookahead and trains Direct Action Ranking on reference actions, comparing prediction-augmented scoring under matched offline supervision.", "topics": ["evaluation", "world-models", "action-prediction", "imitation-learning"], "updated_at": "2026-10-17", "url": "https://arxiv.org/abs/2610.24372"}
{"aliases": [], "authors": [], "details": {"component_path": "WebEvolver", "license_review": {"code": {"evidence_urls": ["https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/License.txt"], "reason": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate."}}, "repository_tree_sha": "a82450e1b5e15070bee79ee8b79f8d9a97cb85f7", "repository_url": "https://github.com/Tencent/SelfEvolvingAgent"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "447dfa1e52789f7f3f111e9e0d453fb1f9fd8bf2d7b858f28fbe21a2239b4b2e", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/WebEvolver/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-17T15:37:01Z", "role": "primary", "sha256": "4214af42bcadf1ea52a46c2008f8e741f0aa955a58f346d9880858589be7de1d", "supports": ["identity"], "url": "https://api.github.com/repos/Tencent/SelfEvolvingAgent"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:25:25Z", "role": "primary", "sha256": "ad0f4ae4b58b5a0e84feeb5df717929b97ea37cb945c17c98510ba54509c99e7", "supports": ["identity", "release_status", "details.repository_tree_sha"], "url": "https://api.github.com/repos/Tencent/SelfEvolvingAgent/git/trees/main?recursive=1"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:13Z", "role": "primary", "sha256": "873a78a46309fa3892d94b72cf73fd55542bae0e0ba8235afb26a848a64a8255", "supports": ["identity"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:19Z", "role": "primary", "sha256": "1d7a5369af7ec336bc173dba6cfee9ad0898e527f96c5b7a10cee930806c67a3", "supports": ["licenses.code"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/License.txt"}], "id": "component:tencent/selfevolvingagent/webevolver", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-18T04:27:14Z", "licenses": {"code": {"basis": "Reviewed repository license declaration for code; third-party dependencies, data, weights and executable environment images are separate.", "declared": "MIT", "retrieved_at": "2026-10-18T04:35:19Z", "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/License.txt"}, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/Tencent/SelfEvolvingAgent/tree/main/WebEvolver"}], "name": "WebEvolver code component", "notes": "Scoped to the WebEvolver directory; the repository also hosts WebCoT and VScan, which are not merged into this resource. Training and model services require separate setup. Root License.txt explicitly declares MIT for SelfEvolvingAgent and separately identifies CogKernel’s Apache-2.0 dependency terms.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/WebEvolver/README.md", "target_id": "arxiv:2504.21024", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Component for web-agent trajectory collection, supervised-data preparation and world-model-based trajectory synthesis.", "topics": ["web-navigation", "world-models", "data-generation", "self-reflection"], "updated_at": null, "url": "https://github.com/Tencent/SelfEvolvingAgent/tree/main/WebEvolver"}
{"aliases": [], "authors": [], "details": {}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T10:43:11Z", "role": "primary", "sha256": "6602cb260b52ae5b2b93db231a51a38d0463c78d8e71321620e05b4fa31caacb", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations"], "url": "https://raw.githubusercontent.com/OSU-NLP-Group/WebDreamer/HEAD/README.md"}], "id": "github:osu-nlp-group/webdreamer", "identifiers": {}, "kind": "framework", "last_verified_at": "2026-10-16T10:43:11Z", "licenses": {"code": null, "data": null, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://github.com/OSU-NLP-Group/WebDreamer"}], "name": "WebDreamer", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/OSU-NLP-Group/WebDreamer/HEAD/README.md", "target_id": "arxiv:2411.06559", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "World-model and planning components that simulate web-action outcomes, with linked training data and model checkpoints.", "topics": ["world-models", "planning"], "updated_at": null, "url": "https://github.com/OSU-NLP-Group/WebDreamer"}
{"aliases": [], "authors": [], "details": {"disabled": false, "download_url": "https://huggingface.co/GD-ML/Code2World/tree/3e48194a8228570164ecd510f9d83ac839e9e931", "gated": false, "hub_id": "GD-ML/Code2World", "hub_revision": "3e48194a8228570164ecd510f9d83ac839e9e931", "license_review": {"weights": {"evidence_urls": ["https://huggingface.co/GD-ML/Code2World/raw/3e48194a8228570164ecd510f9d83ac839e9e931/README.md"], "reason": "Publisher model-card declaration for the checkpoint; code and training-data terms are separate."}}, "repository_url": "https://github.com/AMAP-ML/Code2World"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:15Z", "role": "primary", "sha256": "befc89b09d099b86bda891141bca32b4a13b650c64d9b74ca18a9431e76f6c06", "supports": ["name", "kind", "summary", "topics", "scope", "notes", "relations", "licenses.weights"], "url": "https://huggingface.co/GD-ML/Code2World/raw/3e48194a8228570164ecd510f9d83ac839e9e931/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:22Z", "role": "primary", "sha256": "4c91d95f7533cb5550447c95bfb12d6064fd510ac01cba45ebb8d7fca305fdc9", "supports": ["identity", "release_status", "details.hub_id", "details.hub_revision", "details.gated", "details.disabled", "details.download_url"], "url": "https://huggingface.co/api/models/GD-ML/Code2World"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "26c14a14166d3b747454d1924ecdf6d96652ac857ae6ad488630dc843036a777", "supports": ["environments", "details.repository_url"], "url": "https://raw.githubusercontent.com/AMAP-ML/Code2World/HEAD/readme.md"}], "id": "url:bc3deaf8eeecf569e73b", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-18T04:35:15Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "Publisher model-card declaration for the checkpoint; code and training-data terms are separate.", "declared": "MIT", "retrieved_at": "2026-10-18T04:35:15Z", "url": "https://huggingface.co/GD-ML/Code2World/raw/3e48194a8228570164ecd510f9d83ac839e9e931/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/GD-ML/Code2World"}], "name": "Code2World-8B checkpoint", "notes": "Card identifies Code2World-8B and Qwen3-VL-8B-Instruct. Public files include four Safetensors shards, an index and rendering/prompt helper scripts. The GitHub adapter targets AndroidWorld; model-generated HTML rendering and model loading were not executed. MIT is recorded as the weight declaration, without transferring it to the separate GitHub repository’s code or all training sources.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/GD-ML/Code2World/raw/3e48194a8228570164ecd510f9d83ac839e9e931/README.md", "target_id": "arxiv:2602.09856", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Released GUI world-model checkpoint that turns a screenshot and proposed action into HTML for rendering a predicted next screen.", "topics": ["world-models", "screen-understanding"], "updated_at": null, "url": "https://huggingface.co/GD-ML/Code2World"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [{"basis": "reported", "evidence_url": "https://raw.githubusercontent.com/AMAP-ML/Code2World/HEAD/readme.md", "population": "AndroidCode corpus described by Code2World README; released JSON rows not recounted", "qualifier": "more-than", "unit": "screen–action pairs", "value": 80000}], "evidence_urls": ["https://raw.githubusercontent.com/AMAP-ML/Code2World/HEAD/readme.md", "https://huggingface.co/api/datasets/GD-ML/AndroidCode"], "formats": ["HTML", "JSON", "Python extraction/loading scripts"], "modalities": ["GUI screenshots (described input)", "GUI actions", "HTML"], "notes": "The manifest lists 58,953 HTML files, which are not converted into transition/example counts. The reported >80k pairs and the HTML-file inventory have different units. The card itself has no sample schema or count; archive/image completeness and JSON contents were not audited.", "splits": ["Android-Control train file", "Android-Control test file"], "supervision": ["action-conditioned next-state code prediction", "generated HTML supervision"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/GD-ML/AndroidCode/tree/af316a6be79bbe08c4fc09ea588b5f80f177ed25", "gated": false, "hub_id": "GD-ML/AndroidCode", "hub_revision": "af316a6be79bbe08c4fc09ea588b5f80f177ed25", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/GD-ML/AndroidCode/raw/af316a6be79bbe08c4fc09ea588b5f80f177ed25/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/AMAP-ML/Code2World"}, "environments": ["android"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "26c14a14166d3b747454d1924ecdf6d96652ac857ae6ad488630dc843036a777", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "details.dataset_profile"], "url": "https://raw.githubusercontent.com/AMAP-ML/Code2World/HEAD/readme.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:22Z", "role": "primary", "sha256": "5774a3c096264fbba58b7a16587ea65f35628edb9ca39afe8a32c74ec7fd1ce7", "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/GD-ML/AndroidCode"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:17Z", "role": "primary", "sha256": "d8d7a46d41a1a37fe4f0a5f637bf55c649310185329127d8a2204632e480be17", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/GD-ML/AndroidCode/raw/af316a6be79bbe08c4fc09ea588b5f80f177ed25/README.md"}], "id": "url:cf4d2d45a7314cb16acb", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-18T04:27:14Z", "licenses": {"code": null, "data": {"basis": "Publisher dataset-card declaration for this release; it does not establish rights to all upstream screenshots, applications or source content.", "declared": "MIT", "retrieved_at": "2026-10-18T04:35:17Z", "url": "https://huggingface.co/datasets/GD-ML/AndroidCode/raw/af316a6be79bbe08c4fc09ea588b5f80f177ed25/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/GD-ML/AndroidCode"}], "name": "AndroidCode", "notes": "The Hub card is license-only; semantic description and reported scale come from the official Code2World README. The public manifest contains HTML files, Android-Control train/test JSON files and extraction/loading scripts. It does not list standalone screenshot image files, and embedded/reference image packaging was not inspected. MIT is the data-card declaration, not a grant for every upstream source dataset.", "published_at": null, "relations": [{"evidence_url": "https://raw.githubusercontent.com/AMAP-ML/Code2World/HEAD/readme.md", "target_id": "arxiv:2602.09856", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "HTML and annotation release used for Code2World’s action-conditioned GUI-state prediction training, constructed from mobile GUI trajectories.", "topics": ["world-models", "data-generation", "screen-understanding"], "updated_at": null, "url": "https://huggingface.co/datasets/GD-ML/AndroidCode"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [], "evidence_urls": ["https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md", "https://huggingface.co/api/datasets/CognitiveKernel/WebEvolver"], "formats": ["JSONL"], "modalities": [], "notes": "Two purpose-specific files are released: webevolver_policy_sft.jsonl and webevolver_world_model_sft.jsonl. They are not assumed disjoint train/test splits, and no modality or example count is inferred from filenames alone.", "splits": [], "supervision": ["agent-policy supervised fine-tuning", "web-world-model supervised fine-tuning"]}, "disabled": false, "download_url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/tree/13f4f45b359e138a45b1c89bd4e913298c586ca9", "gated": false, "hub_id": "CognitiveKernel/WebEvolver", "hub_revision": "13f4f45b359e138a45b1c89bd4e913298c586ca9", "license_review": {"data": {"evidence_urls": ["https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md"], "reason": "Publisher dataset-card declaration for this release; it does not establish rights to all upstream screenshots, applications or source content."}}, "repository_url": "https://github.com/Tencent/SelfEvolvingAgent/tree/main/WebEvolver"}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:35:17Z", "role": "primary", "sha256": "36ccdce12c1b1512ad4b21326c5d0c0379c73e85f0b8150779d43240b54686de", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "notes", "relations", "licenses.data", "details.dataset_profile"], "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:33:29Z", "role": "primary", "sha256": "cbef3f8037405d311ed9fe8aa620ec3734f91c77ca8f4209fbe98167fb33bee8", "supports": ["identity", "release_status", "details.hub_id", "details.hub_revision", "details.gated", "details.disabled", "details.download_url", "details.dataset_profile"], "url": "https://huggingface.co/api/datasets/CognitiveKernel/WebEvolver"}, {"method": "editorial_primary_review", "retrieved_at": "2026-10-18T04:27:14Z", "role": "primary", "sha256": "447dfa1e52789f7f3f111e9e0d453fb1f9fd8bf2d7b858f28fbe21a2239b4b2e", "supports": ["details.repository_url"], "url": "https://raw.githubusercontent.com/Tencent/SelfEvolvingAgent/HEAD/WebEvolver/README.md"}], "id": "url:e126bbd9729ebdaf05ab", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-18T04:35:17Z", "licenses": {"code": null, "data": {"basis": "Publisher dataset-card declaration for this release; it does not establish rights to all upstream screenshots, applications or source content.", "declared": "MIT", "retrieved_at": "2026-10-18T04:35:17Z", "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver"}], "name": "WebEvolver SFT data", "notes": "The card and public manifest identify separate policy-model and world-model JSONL files. No row total or complete record schema is documented in the inspected card; the files were not downloaded as training samples. MIT is the publisher’s data declaration.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver/raw/13f4f45b359e138a45b1c89bd4e913298c586ca9/README.md", "target_id": "arxiv:2504.21024", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Two JSONL collections for supervised fine-tuning the WebEvolver agent policy and web world model.", "topics": ["web-navigation", "world-models", "imitation-learning"], "updated_at": null, "url": "https://huggingface.co/datasets/CognitiveKernel/WebEvolver"}
{"aliases": [], "authors": [], "details": {"base_model": "Qwen/Qwen2-VL-7B", "download_url": "https://huggingface.co/osunlp/Dreamer-7B/tree/main", "languages": ["en"]}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:10:45Z", "role": "primary", "sha256": "f15e05cc2718948e344bd1fb943402797ef2e3f2130a9ce2fd94d419210c0574", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details"], "url": "https://huggingface.co/osunlp/Dreamer-7B/raw/main/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-16T11:10:45Z", "role": "primary", "sha256": "f15e05cc2718948e344bd1fb943402797ef2e3f2130a9ce2fd94d419210c0574", "supports": ["licenses.weights"], "url": "https://huggingface.co/osunlp/Dreamer-7B/raw/main/README.md"}], "id": "url:e3c1a2a2d556e4728953", "identifiers": {}, "kind": "model", "last_verified_at": "2026-10-16T11:10:45Z", "licenses": {"code": null, "data": null, "document": null, "weights": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "apache-2.0", "retrieved_at": "2026-10-16T11:10:45Z", "url": "https://huggingface.co/osunlp/Dreamer-7B/raw/main/README.md"}}, "links": [{"rel": "primary", "url": "https://huggingface.co/osunlp/Dreamer-7B"}], "name": "Dreamer-7B", "notes": "", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/osunlp/Dreamer-7B/raw/main/README.md", "target_id": "arxiv:2411.06559", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "WebDreamer model checkpoint that predicts the consequences of web actions from screenshots and textual descriptions.", "topics": ["world-models", "planning"], "updated_at": null, "url": "https://huggingface.co/osunlp/Dreamer-7B"}
{"aliases": [], "authors": [], "details": {"dataset_profile": {"counts": [], "evidence_urls": ["https://huggingface.co/datasets/osunlp/Dreamer-V1-Data/raw/main/README.md"], "formats": [], "modalities": ["images", "prompts", "actions", "responses"], "notes": "The card reports a cleaned size of 3.12M without defining the counting unit, so no sample total is inferred. The data card declares noncommercial share-alike terms, separately from the model-weight license.", "splits": [], "supervision": ["web-world-model prediction"]}, "download_url": "https://huggingface.co/datasets/osunlp/Dreamer-V1-Data/tree/main", "modalities": ["image", "text"]}, "environments": ["web"], "evidence": [{"method": "editorial_primary_review", "retrieved_at": "2026-10-16T11:09:28Z", "role": "primary", "sha256": "e656246f09340fed08136cf41a8b79880e19e1397ddd176410688ddee4e6365c", "supports": ["name", "kind", "summary", "environments", "topics", "scope", "relations", "details.modalities", "details.download_url", "details.dataset_profile"], "url": "https://huggingface.co/datasets/osunlp/Dreamer-V1-Data/raw/main/README.md"}, {"method": "huggingface_card", "retrieved_at": "2026-10-16T11:09:28Z", "role": "primary", "sha256": "e656246f09340fed08136cf41a8b79880e19e1397ddd176410688ddee4e6365c", "supports": ["licenses.data"], "url": "https://huggingface.co/datasets/osunlp/Dreamer-V1-Data/raw/main/README.md"}], "id": "url:eadc6cba9ea46c38dd1d", "identifiers": {}, "kind": "dataset", "last_verified_at": "2026-10-16T11:09:28Z", "licenses": {"code": null, "data": {"basis": "publisher model/dataset card metadata; third-party assets retain their own terms", "declared": "cc-by-nc-sa-4.0", "retrieved_at": "2026-10-16T11:09:28Z", "url": "https://huggingface.co/datasets/osunlp/Dreamer-V1-Data/raw/main/README.md"}, "document": null, "weights": null}, "links": [{"rel": "primary", "url": "https://huggingface.co/datasets/osunlp/Dreamer-V1-Data"}], "name": "Dreamer V1 Data", "notes": "The card reports a cleaned size of 3.12M without explicitly defining its counting unit. The catalog leaves the exact sample count unknown. Its data license differs from the model-weight license.", "published_at": null, "relations": [{"evidence_url": "https://huggingface.co/datasets/osunlp/Dreamer-V1-Data/raw/main/README.md", "target_id": "arxiv:2411.06559", "type": "documented_by"}], "review_status": "curated", "scope": "core", "summary": "Web-world-model training data with prompts, images, action descriptions, and responses, released with the WebDreamer project.", "topics": ["world-models", "data-generation"], "updated_at": null, "url": "https://huggingface.co/datasets/osunlp/Dreamer-V1-Data"}