[ { "URL": "https://arxiv.org/abs/2404.16054v2", "archive": "arXiv", "archive_location": "2404.16054", "author": [ { "literal": "Li Zhang" }, { "literal": "Shihe Wang" }, { "literal": "Xianqing Jia" }, { "literal": "Zhihan Zheng" }, { "literal": "Yunhe Yan" }, { "literal": "Longxi Gao" }, { "literal": "Yuanchun Li" }, { "literal": "Mengwei Xu" } ], "citation-key": "arxiv_2404_16054", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "28c3ad086106ec4199dcc9da2b8aa3b20eac9347d9354b9dfb373e50ea4c3f50", "submitted_at": "2024-04-12T15:39:09Z", "updated_at": "2024-08-02T13:49:32Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2404.16054v2" }, "canonical_source_url": "https://arxiv.org/abs/2404.16054", "catalog_id": "arxiv:2404.16054", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2024-04-12", "identifiers": { "arxiv": "2404.16054" }, "record_updated_at": "2024-08-02" }, "genre": "preprint", "id": "arxiv:2404.16054", "issued": { "date-parts": [ [ 2024, 4, 12 ] ] }, "note": "First published: 2024-04-12; Recorded arXiv version: v2; Recorded version updated: 2024-08-02T13:49:32Z.", "title": "LlamaTouch: A Faithful and Scalable Testbed for Mobile UI Task Automation", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2405.11120v1", "archive": "arXiv", "archive_location": "2405.11120", "author": [ { "literal": "William E Bishop" }, { "literal": "Alice Li" }, { "literal": "Christopher Rawles" }, { "literal": "Oriana Riva" } ], "citation-key": "arxiv_2405_11120", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "81f62326dddc5c512fb05472eb1f387064c0a5aad77453e9d2e85415bb627fa9", "submitted_at": "2024-05-17T23:27:33Z", "updated_at": "2024-05-17T23:27:33Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2405.11120v1" }, "canonical_source_url": "https://arxiv.org/abs/2405.11120", "catalog_id": "arxiv:2405.11120", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2024-05-17", "identifiers": { "arxiv": "2405.11120" }, "record_updated_at": "2024-05-17" }, "genre": "preprint", "id": "arxiv:2405.11120", "issued": { "date-parts": [ [ 2024, 5, 17 ] ] }, "note": "First published: 2024-05-17; Recorded arXiv version: v1; Recorded version updated: 2024-05-17T23:27:33Z.", "title": "Latent State Estimation Helps UI Agents to Reason", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2407.03037v2", "archive": "arXiv", "archive_location": "2407.03037", "author": [ { "literal": "Zhe Liu" }, { "literal": "Cheng Li" }, { "literal": "Chunyang Chen" }, { "literal": "Junjie Wang" }, { "literal": "Mengzhuo Chen" }, { "literal": "Boyu Wu" }, { "literal": "Yawen Wang" }, { "literal": "Jun Hu" }, { "literal": "Qing Wang" } ], "citation-key": "arxiv_2407_03037", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "73b706ac6089a7d19836268437f3fea58d10b9d170ac596ac0d5a12f0d3a9668", "submitted_at": "2024-07-03T11:58:09Z", "updated_at": "2024-12-04T08:00:42Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2407.03037v2" }, "canonical_source_url": "https://arxiv.org/abs/2407.03037", "catalog_id": "arxiv:2407.03037", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2024-07-03", "identifiers": { "arxiv": "2407.03037" }, "record_updated_at": "2024-12-04" }, "genre": "preprint", "id": "arxiv:2407.03037", "issued": { "date-parts": [ [ 2024, 7, 3 ] ] }, "note": "First published: 2024-07-03; Recorded arXiv version: v2; Recorded version updated: 2024-12-04T08:00:42Z.", "title": "Seeing is Believing: Vision-driven Non-crash Functional Bug Detection for Mobile Apps", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2407.13032v1", "archive": "arXiv", "archive_location": "2407.13032", "author": [ { "literal": "Tamer Abuelsaad" }, { "literal": "Deepak Akkil" }, { "literal": "Prasenjit Dey" }, { "literal": "Ashish Jagmohan" }, { "literal": "Aditya Vempaty" }, { "literal": "Ravi Kokku" } ], "citation-key": "arxiv_2407_13032", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "18cfd29fcb66d703d90f747c162db1b3fc6a43f400b4eacbddf081bb83ed7947", "submitted_at": "2024-07-17T21:44:28Z", "updated_at": "2024-07-17T21:44:28Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2407.13032v1" }, "canonical_source_url": "https://arxiv.org/abs/2407.13032", "catalog_id": "arxiv:2407.13032", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2024-07-17", "identifiers": { "arxiv": "2407.13032" }, "record_updated_at": "2024-07-17" }, "genre": "preprint", "id": "arxiv:2407.13032", "issued": { "date-parts": [ [ 2024, 7, 17 ] ] }, "note": "First published: 2024-07-17; Recorded arXiv version: v1; Recorded version updated: 2024-07-17T21:44:28Z.", "title": "Agent-E: From Autonomous Web Navigation to Foundational Design Principles in Agentic Systems", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2410.18963v1", "archive": "arXiv", "archive_location": "2410.18963", "author": [ { "literal": "Xiaoqiang Wang" }, { "literal": "Bang Liu" } ], "citation-key": "arxiv_2410_18963", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "66fa51b108ac79cc43a895ba757ca463d5008bfe6db5cde4bdf3a3bb38c444eb", "submitted_at": "2024-10-24T17:58:08Z", "updated_at": "2024-10-24T17:58:08Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2410.18963v1" }, "canonical_source_url": "https://arxiv.org/abs/2410.18963", "catalog_id": "arxiv:2410.18963", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2024-10-24", "identifiers": { "arxiv": "2410.18963" }, "record_updated_at": "2024-10-24" }, "genre": "preprint", "id": "arxiv:2410.18963", "issued": { "date-parts": [ [ 2024, 10, 24 ] ] }, "note": "First published: 2024-10-24; Recorded arXiv version: v1; Recorded version updated: 2024-10-24T17:58:08Z.", "title": "OSCAR: Operating System Control via State-Aware Reasoning and Re-Planning", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2410.23555v1", "archive": "arXiv", "archive_location": "2410.23555", "author": [ { "literal": "Nalin Tiwary" }, { "literal": "Vardhan Dongre" }, { "literal": "Sanil Arun Chawla" }, { "literal": "Ashwin Lamani" }, { "literal": "Dilek Hakkani-Tür" } ], "citation-key": "arxiv_2410_23555", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "1c67b1559410152607fc9d54abf25c37ef939c9f755f089bdf87d4f39393c085", "submitted_at": "2024-10-31T01:51:41Z", "updated_at": "2024-10-31T01:51:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2410.23555v1" }, "canonical_source_url": "https://arxiv.org/abs/2410.23555", "catalog_id": "arxiv:2410.23555", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2024-10-31", "identifiers": { "arxiv": "2410.23555" }, "record_updated_at": "2024-10-31" }, "genre": "preprint", "id": "arxiv:2410.23555", "issued": { "date-parts": [ [ 2024, 10, 31 ] ] }, "note": "First published: 2024-10-31; Recorded arXiv version: v1; Recorded version updated: 2024-10-31T01:51:41Z.", "title": "From Context to Action: Analysis of the Impact of State Representation and Context on the Generalization of Multi-Turn Web Navigation Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2506.18158v1", "archive": "arXiv", "archive_location": "2506.18158", "author": [ { "literal": "Xinzge Gao" }, { "literal": "Chuanrui Hu" }, { "literal": "Bin Chen" }, { "literal": "Teng Li" } ], "citation-key": "arxiv_2506_18158", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "aaa62e24a7a269fea49471f17028a3d5d8068d934e450f53bc002ae80fb7cb03", "submitted_at": "2025-06-22T20:17:46Z", "updated_at": "2025-06-22T20:17:46Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2506.18158v1" }, "canonical_source_url": "https://arxiv.org/abs/2506.18158", "catalog_id": "arxiv:2506.18158", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2025-06-22", "identifiers": { "arxiv": "2506.18158" }, "record_updated_at": "2025-06-22" }, "genre": "preprint", "id": "arxiv:2506.18158", "issued": { "date-parts": [ [ 2025, 6, 22 ] ] }, "note": "First published: 2025-06-22; Recorded arXiv version: v1; Recorded version updated: 2025-06-22T20:17:46Z.", "title": "Chain-of-Memory: Enhancing GUI Agents for Cross-Application Navigation", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2509.13615v5", "archive": "arXiv", "archive_location": "2509.13615", "author": [ { "literal": "Zongru Wu" }, { "literal": "Rui Mao" }, { "literal": "Zhiyuan Tian" }, { "literal": "Pengzhou Cheng" }, { "literal": "Tianjie Ju" }, { "literal": "Zheng Wu" }, { "literal": "Lingzhong Dong" }, { "literal": "Haiyue Sheng" }, { "literal": "Zhuosheng Zhang" }, { "literal": "Gongshen Liu" } ], "citation-key": "arxiv_2509_13615", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "be8620b021bd9218c1143282f1c7a7ea06be2abe6fc0928a310cfc8ac3172a18", "submitted_at": "2025-09-17T01:14:14Z", "updated_at": "2026-03-18T08:12:54Z", "version": 5, "versioned_url": "https://arxiv.org/abs/2509.13615v5" }, "canonical_source_url": "https://arxiv.org/abs/2509.13615", "catalog_id": "arxiv:2509.13615", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2025-09-17", "identifiers": { "arxiv": "2509.13615" }, "record_updated_at": "2026-03-18" }, "genre": "preprint", "id": "arxiv:2509.13615", "issued": { "date-parts": [ [ 2025, 9, 17 ] ] }, "note": "First published: 2025-09-17; Recorded arXiv version: v5; Recorded version updated: 2026-03-18T08:12:54Z.", "title": "See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles", "type": "article", "version": "5" }, { "URL": "https://arxiv.org/abs/2510.14621v1", "archive": "arXiv", "archive_location": "2510.14621", "author": [ { "literal": "Yuanyi Song" }, { "literal": "Heyuan Huang" }, { "literal": "Qiqiang Lin" }, { "literal": "Yin Zhao" }, { "literal": "Xiangmou Qu" }, { "literal": "Jun Wang" }, { "literal": "Xingyu Lou" }, { "literal": "Weiwen Liu" }, { "literal": "Zhuosheng Zhang" }, { "literal": "Jun Wang" }, { "literal": "Yong Yu" }, { "literal": "Weinan Zhang" }, { "literal": "Zhaoxiang Wang" } ], "citation-key": "arxiv_2510_14621", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fc45dc62068c971cd831fc3db64bd3d8995d4e5fd60e9c83953476d4fb24f7f3", "submitted_at": "2025-10-16T12:30:05Z", "updated_at": "2025-10-16T12:30:05Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2510.14621v1" }, "canonical_source_url": "https://arxiv.org/abs/2510.14621", "catalog_id": "arxiv:2510.14621", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2025-10-16", "identifiers": { "arxiv": "2510.14621" }, "record_updated_at": "2025-10-16" }, "genre": "preprint", "id": "arxiv:2510.14621", "issued": { "date-parts": [ [ 2025, 10, 16 ] ] }, "note": "First published: 2025-10-16; Recorded arXiv version: v1; Recorded version updated: 2025-10-16T12:30:05Z.", "title": "ColorBench: Benchmarking Mobile Agents with Graph-Structured Framework for Complex Long-Horizon Tasks", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2510.24168v3", "archive": "arXiv", "archive_location": "2510.24168", "author": [ { "literal": "Weihua Cheng" }, { "literal": "Junming Liu" }, { "literal": "Yifei Sun" }, { "literal": "Botian Shi" }, { "literal": "Yirong Chen" }, { "literal": "Ding Wang" } ], "citation-key": "arxiv_2510_24168", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d80312fbd8f57cdc53cc10e03f2c68e2152d06945e5089967dceda50c02b9d4b", "submitted_at": "2025-10-28T08:19:58Z", "updated_at": "2026-04-14T04:34:20Z", "version": 3, "versioned_url": "https://arxiv.org/abs/2510.24168v3" }, "canonical_source_url": "https://arxiv.org/abs/2510.24168", "catalog_id": "arxiv:2510.24168", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2025-10-28", "identifiers": { "arxiv": "2510.24168" }, "record_updated_at": "2026-04-14" }, "genre": "preprint", "id": "arxiv:2510.24168", "issued": { "date-parts": [ [ 2025, 10, 28 ] ] }, "note": "First published: 2025-10-28; Recorded arXiv version: v3; Recorded version updated: 2026-04-14T04:34:20Z.", "title": "MGA: Memory-Driven GUI Agent for Observation-Centric Interaction", "type": "article", "version": "3" }, { "URL": "https://arxiv.org/abs/2511.22235v2", "archive": "arXiv", "archive_location": "2511.22235", "author": [ { "literal": "Zehao Deng" }, { "literal": "Tianjie Ju" }, { "literal": "Zheng Wu" }, { "literal": "Zhuosheng Zhang" }, { "literal": "Gongshen Liu" } ], "citation-key": "arxiv_2511_22235", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "9acb395dfaad18b84ea4b52d97be194a42eb42d6ac922c55f2179e38cc258a31", "submitted_at": "2025-11-27T09:01:38Z", "updated_at": "2026-03-04T12:15:06Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2511.22235v2" }, "canonical_source_url": "https://arxiv.org/abs/2511.22235", "catalog_id": "arxiv:2511.22235", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2025-11-27", "identifiers": { "arxiv": "2511.22235" }, "record_updated_at": "2026-03-04" }, "genre": "preprint", "id": "arxiv:2511.22235", "issued": { "date-parts": [ [ 2025, 11, 27 ] ] }, "note": "First published: 2025-11-27; Recorded arXiv version: v2; Recorded version updated: 2026-03-04T12:15:06Z.", "title": "Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2512.01174v1", "archive": "arXiv", "archive_location": "2512.01174", "author": [ { "literal": "Hyunjun Kim" }, { "literal": "Sooyoung Ryu" } ], "citation-key": "arxiv_2512_01174", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7e445d0ca79f727a6c1d51e83381c3fd1d8e1fddeff7de99fe70ca06bc1bff49", "submitted_at": "2025-12-01T01:18:21Z", "updated_at": "2025-12-01T01:18:21Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2512.01174v1" }, "canonical_source_url": "https://arxiv.org/abs/2512.01174", "catalog_id": "arxiv:2512.01174", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2025-12-01", "identifiers": { "arxiv": "2512.01174" }, "record_updated_at": "2025-12-01" }, "genre": "preprint", "id": "arxiv:2512.01174", "issued": { "date-parts": [ [ 2025, 12, 1 ] ] }, "note": "First published: 2025-12-01; Recorded arXiv version: v1; Recorded version updated: 2025-12-01T01:18:21Z.", "title": "DrawingBench: Evaluating Spatial Reasoning and UI Interaction Capabilities of Large Language Models through Mouse-Based Drawing Tasks", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2601.21352v1", "archive": "arXiv", "archive_location": "2601.21352", "author": [ { "literal": "Ziyu Lu" }, { "literal": "Tengjin Weng" }, { "literal": "Yiying Yang" }, { "literal": "Yuhang Zhao" }, { "literal": "Xinxin Huang" }, { "literal": "Wenhao Jiang" } ], "citation-key": "arxiv_2601_21352", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "fa36a9fbd0b189717b2a5311f06b6127d54f2eca09961cedd2a0b9d2a78b1b8b", "submitted_at": "2026-01-29T07:22:50Z", "updated_at": "2026-01-29T07:22:50Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2601.21352v1" }, "canonical_source_url": "https://arxiv.org/abs/2601.21352", "catalog_id": "arxiv:2601.21352", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-01-29", "identifiers": { "arxiv": "2601.21352" }, "record_updated_at": "2026-01-29" }, "genre": "preprint", "id": "arxiv:2601.21352", "issued": { "date-parts": [ [ 2026, 1, 29 ] ] }, "note": "First published: 2026-01-29; Recorded arXiv version: v1; Recorded version updated: 2026-01-29T07:22:50Z.", "title": "BEAP-Agent: Backtrackable Execution and Adaptive Planning for GUI Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2603.00476v1", "archive": "arXiv", "archive_location": "2603.00476", "author": [ { "literal": "Linxi Jiang" }, { "literal": "Zhijie Liu" }, { "literal": "Haotian Luo" }, { "literal": "Zhiqiang Lin" } ], "citation-key": "arxiv_2603_00476", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d3b24ac636d86124a70b44e84657b963db4c5af3367ae70e5c35e0bebb3f600f", "submitted_at": "2026-02-28T05:25:03Z", "updated_at": "2026-02-28T05:25:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.00476v1" }, "canonical_source_url": "https://arxiv.org/abs/2603.00476", "catalog_id": "arxiv:2603.00476", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-02-28", "identifiers": { "arxiv": "2603.00476" }, "record_updated_at": "2026-02-28" }, "genre": "preprint", "id": "arxiv:2603.00476", "issued": { "date-parts": [ [ 2026, 2, 28 ] ] }, "note": "First published: 2026-02-28; Recorded arXiv version: v1; Recorded version updated: 2026-02-28T05:25:03Z.", "title": "Atomicity for Agents: Exposing, Exploiting, and Mitigating TOCTOU Vulnerabilities in Browser-Use Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2603.08013v1", "archive": "arXiv", "archive_location": "2603.08013", "author": [ { "literal": "Yuxiang Chai" }, { "literal": "Shunye Tang" }, { "literal": "Han Xiao" }, { "literal": "Rui Liu" }, { "literal": "Hongsheng Li" } ], "citation-key": "arxiv_2603_08013", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7334e01b1fbf5cdd02c275c73d2a1b5c2931718d6019fcf6a0ea72b449704d94", "submitted_at": "2026-03-09T06:41:32Z", "updated_at": "2026-03-09T06:41:32Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2603.08013v1" }, "canonical_source_url": "https://arxiv.org/abs/2603.08013", "catalog_id": "arxiv:2603.08013", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-03-09", "identifiers": { "arxiv": "2603.08013" }, "record_updated_at": "2026-03-09" }, "genre": "preprint", "id": "arxiv:2603.08013", "issued": { "date-parts": [ [ 2026, 3, 9 ] ] }, "note": "First published: 2026-03-09; Recorded arXiv version: v1; Recorded version updated: 2026-03-09T06:41:32Z.", "title": "PIRA-Bench: A Transition from Reactive GUI Agents to GUI-based Proactive Intent Recommendation Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2604.02345v1", "archive": "arXiv", "archive_location": "2604.02345", "author": [ { "literal": "Mengzhou Wu" }, { "literal": "Yuzhe Guo" }, { "literal": "Yuan Cao" }, { "literal": "Haochuan Lu" }, { "literal": "Songhe Zhu" }, { "literal": "Pingzhe Qu" }, { "literal": "Xin Chen" }, { "literal": "Kang Qin" }, { "literal": "Zhongpu Wang" }, { "literal": "Xiaode Zhang" }, { "literal": "Xinyi Wang" }, { "literal": "Wei Dai" }, { "literal": "Gang Cao" }, { "literal": "Yuetang Deng" }, { "literal": "Zhi Gong" }, { "literal": "Dezhi Ran" }, { "literal": "Linyi Li" }, { "literal": "Wei Yang" }, { "literal": "Tao Xie" } ], "citation-key": "arxiv_2604_02345", "custom": { "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" }, "canonical_source_url": "https://arxiv.org/abs/2604.02345", "catalog_id": "arxiv:2604.02345", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-02-11", "identifiers": { "arxiv": "2604.02345" }, "record_updated_at": "2026-02-11" }, "genre": "preprint", "id": "arxiv:2604.02345", "issued": { "date-parts": [ [ 2026, 2, 11 ] ] }, "note": "First published: 2026-02-11; Recorded arXiv version: v1; Recorded version updated: 2026-02-11T17:02:38Z.", "title": "UI-Oceanus: Scaling GUI Agents with Synthetic Environmental Dynamics", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2604.04399v1", "archive": "arXiv", "archive_location": "2604.04399", "author": [ { "literal": "Yuwen Zhai" }, { "literal": "Runze Li" }, { "literal": "Liang Wang" }, { "literal": "Nian Shi" }, { "literal": "Liwu Xu" }, { "literal": "Wei Zhang" }, { "literal": "Ran Lin" }, { "literal": "Bo Xu" }, { "literal": "Benlei Cui" } ], "citation-key": "arxiv_2604_04399", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "856e9920aea7d821061b3258b31aa32fd37b5579d628f5507f45ba83f237b77b", "submitted_at": "2026-04-06T03:58:43Z", "updated_at": "2026-04-06T03:58:43Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.04399v1" }, "canonical_source_url": "https://arxiv.org/abs/2604.04399", "catalog_id": "arxiv:2604.04399", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-06", "identifiers": { "arxiv": "2604.04399" }, "record_updated_at": "2026-04-06" }, "genre": "preprint", "id": "arxiv:2604.04399", "issued": { "date-parts": [ [ 2026, 4, 6 ] ] }, "note": "First published: 2026-04-06; Recorded arXiv version: v1; Recorded version updated: 2026-04-06T03:58:43Z.", "title": "GUIDE: Interpretable GUI Agent Evaluation via Hierarchical Diagnosis", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2604.05477v1", "archive": "arXiv", "archive_location": "2604.05477", "author": [ { "literal": "Yuzhe Zhang" }, { "literal": "Xianwei Xue" }, { "literal": "Xingyong Wu" }, { "literal": "Mengke Chen" }, { "literal": "Chen Liu" }, { "literal": "Xinran He" }, { "literal": "Run Shao" }, { "literal": "Feiran Liu" }, { "literal": "Huanmin Xu" }, { "literal": "Qiutong Pan" }, { "literal": "Haiwei Wang" } ], "citation-key": "arxiv_2604_05477", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "7c9ef656cbee786bec0d37ac4915bfd04adb36410eda8e647fab0d8605f90753", "submitted_at": "2026-04-07T06:16:23Z", "updated_at": "2026-04-07T06:16:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.05477v1" }, "canonical_source_url": "https://arxiv.org/abs/2604.05477", "catalog_id": "arxiv:2604.05477", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-07", "identifiers": { "arxiv": "2604.05477" }, "record_updated_at": "2026-04-07" }, "genre": "preprint", "id": "arxiv:2604.05477", "issued": { "date-parts": [ [ 2026, 4, 7 ] ] }, "note": "First published: 2026-04-07; Recorded arXiv version: v1; Recorded version updated: 2026-04-07T06:16:23Z.", "title": "Don't Act Blindly: Robust GUI Automation via Action-Effect Verification and Self-Correction", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2604.06367v2", "archive": "arXiv", "archive_location": "2604.06367", "author": [ { "literal": "Guruprasad Viswanathan Ramesh" }, { "literal": "Asmit Nayak" }, { "literal": "Basieem Siddique" }, { "literal": "Kassem Fawaz" } ], "citation-key": "arxiv_2604_06367", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "400d8a9f7aed4f150dd5198d9a1561d2df0454069495594ca165ad4254a1be59", "submitted_at": "2026-04-07T18:43:21Z", "updated_at": "2026-06-16T22:37:53Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.06367v2" }, "canonical_source_url": "https://arxiv.org/abs/2604.06367", "catalog_id": "arxiv:2604.06367", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-07", "identifiers": { "arxiv": "2604.06367" }, "record_updated_at": "2026-06-16" }, "genre": "preprint", "id": "arxiv:2604.06367", "issued": { "date-parts": [ [ 2026, 4, 7 ] ] }, "note": "First published: 2026-04-07; Recorded arXiv version: v2; Recorded version updated: 2026-06-16T22:37:53Z.", "title": "WebSP-Eval: Evaluating Web Agents on Website Security and Privacy Tasks", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2604.18240v1", "archive": "arXiv", "archive_location": "2604.18240", "author": [ { "literal": "Wentao Shi" }, { "literal": "Yu Wang" }, { "literal": "Yuyang Zhao" }, { "literal": "Yuxin Chen" }, { "literal": "Fuli Feng" }, { "literal": "Xueyuan Hao" }, { "literal": "Xi Su" }, { "literal": "Qi Gu" }, { "literal": "Hui Su" }, { "literal": "Xunliang Cai" }, { "literal": "Xiangnan He" } ], "citation-key": "arxiv_2604_18240", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "91a9cdb6c8cb124d2e6033f0decc85e0c73bd972984d973ff687bbdc2be21176", "submitted_at": "2026-04-20T13:23:38Z", "updated_at": "2026-04-20T13:23:38Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.18240v1" }, "canonical_source_url": "https://arxiv.org/abs/2604.18240", "catalog_id": "arxiv:2604.18240", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-20", "identifiers": { "arxiv": "2604.18240" }, "record_updated_at": "2026-04-20" }, "genre": "preprint", "id": "arxiv:2604.18240", "issued": { "date-parts": [ [ 2026, 4, 20 ] ] }, "note": "First published: 2026-04-20; Recorded arXiv version: v1; Recorded version updated: 2026-04-20T13:23:38Z.", "title": "AJ-Bench: Benchmarking Agent-as-a-Judge for Environment-Aware Evaluation", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2604.18860v1", "archive": "arXiv", "archive_location": "2604.18860", "author": [ { "literal": "Wenpeng Xu" } ], "citation-key": "arxiv_2604_18860", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "8eb499729268ef287a9b9bf73275f526d47dabf2d7de399beaead2fe54cfd74e", "submitted_at": "2026-04-20T21:36:16Z", "updated_at": "2026-04-20T21:36:16Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.18860v1" }, "canonical_source_url": "https://arxiv.org/abs/2604.18860", "catalog_id": "arxiv:2604.18860", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-20", "identifiers": { "arxiv": "2604.18860" }, "record_updated_at": "2026-04-20" }, "genre": "preprint", "id": "arxiv:2604.18860", "issued": { "date-parts": [ [ 2026, 4, 20 ] ] }, "note": "First published: 2026-04-20; Recorded arXiv version: v1; Recorded version updated: 2026-04-20T21:36:16Z.", "title": "Temporal UI State Inconsistency in Desktop GUI Agents: Formalizing and Defending Against TOCTOU Attacks on Computer-Use Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2604.24441v1", "archive": "arXiv", "archive_location": "2604.24441", "author": [ { "literal": "Hongxin Li" }, { "literal": "Xiping Wang" }, { "literal": "Jingran Su" }, { "literal": "Zheng Ju" }, { "literal": "Yuntao Chen" }, { "literal": "Qing Li" }, { "literal": "Zhaoxiang Zhang" } ], "citation-key": "arxiv_2604_24441", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "65325912365effe8adbc39301a378d7526c60d1185691f3a89cb8acd4aa53c3f", "submitted_at": "2026-04-27T13:06:27Z", "updated_at": "2026-04-27T13:06:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.24441v1" }, "canonical_source_url": "https://arxiv.org/abs/2604.24441", "catalog_id": "arxiv:2604.24441", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-27", "identifiers": { "arxiv": "2604.24441" }, "record_updated_at": "2026-04-27" }, "genre": "preprint", "id": "arxiv:2604.24441", "issued": { "date-parts": [ [ 2026, 4, 27 ] ] }, "note": "First published: 2026-04-27; Recorded arXiv version: v1; Recorded version updated: 2026-04-27T13:06:27Z.", "title": "AutoGUI-v2: A Comprehensive Multi-Modal GUI Functionality Understanding Benchmark", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2604.25380v2", "archive": "arXiv", "archive_location": "2604.25380", "author": [ { "literal": "Enqi Liu" }, { "literal": "Liyuan Pan" }, { "literal": "Zhi Gao" }, { "literal": "Yan Yang" }, { "literal": "Chenrui Shi" }, { "literal": "Yang Liu" }, { "literal": "Jingrong Wu" }, { "literal": "Qing Li" } ], "citation-key": "arxiv_2604_25380", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "54b49f54319922268ba880a1c1e8a27ffa7597e228b18be04caa1a6952a6456a", "submitted_at": "2026-04-28T08:43:58Z", "updated_at": "2026-05-08T04:29:20Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2604.25380v2" }, "canonical_source_url": "https://arxiv.org/abs/2604.25380", "catalog_id": "arxiv:2604.25380", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-28", "identifiers": { "arxiv": "2604.25380" }, "record_updated_at": "2026-05-08" }, "genre": "preprint", "id": "arxiv:2604.25380", "issued": { "date-parts": [ [ 2026, 4, 28 ] ] }, "note": "First published: 2026-04-28; Recorded arXiv version: v2; Recorded version updated: 2026-05-08T04:29:20Z.", "title": "Benchmarking and Improving GUI Agents in High-Dynamic Environments", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2604.27151v1", "archive": "arXiv", "archive_location": "2604.27151", "author": [ { "literal": "Jinbiao Wei" }, { "literal": "Kangqi Ni" }, { "literal": "Yilun Zhao" }, { "literal": "Guo Gan" }, { "literal": "Arman Cohan" } ], "citation-key": "arxiv_2604_27151", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "68183be37dbd78314aca35f7f1420e5f2f8da8d8c0042a258b45ae03c90606cf", "submitted_at": "2026-04-29T19:59:36Z", "updated_at": "2026-04-29T19:59:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2604.27151v1" }, "canonical_source_url": "https://arxiv.org/abs/2604.27151", "catalog_id": "arxiv:2604.27151", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-04-29", "identifiers": { "arxiv": "2604.27151" }, "record_updated_at": "2026-04-29" }, "genre": "preprint", "id": "arxiv:2604.27151", "issued": { "date-parts": [ [ 2026, 4, 29 ] ] }, "note": "First published: 2026-04-29; Recorded arXiv version: v1; Recorded version updated: 2026-04-29T19:59:36Z.", "title": "Step-level Optimization for Efficient Computer-use Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2605.12493v1", "archive": "arXiv", "archive_location": "2605.12493", "author": [ { "literal": "Di Wu" }, { "literal": "Zixiang Ji" }, { "literal": "Asmi Kawatkar" }, { "literal": "Bryan Kwan" }, { "literal": "Jia-Chen Gu" }, { "literal": "Nanyun Peng" }, { "literal": "Kai-Wei Chang" } ], "citation-key": "arxiv_2605_12493", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5893d1d9205776667a1e1d89720ec3baf55fafe80cc9c6201ca97088f44a10ac", "submitted_at": "2026-05-12T17:59:34Z", "updated_at": "2026-05-12T17:59:34Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.12493v1" }, "canonical_source_url": "https://arxiv.org/abs/2605.12493", "catalog_id": "arxiv:2605.12493", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-05-12", "identifiers": { "arxiv": "2605.12493" }, "record_updated_at": "2026-05-12" }, "genre": "preprint", "id": "arxiv:2605.12493", "issued": { "date-parts": [ [ 2026, 5, 12 ] ] }, "note": "First published: 2026-05-12; Recorded arXiv version: v1; Recorded version updated: 2026-05-12T17:59:34Z.", "title": "LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2605.15777v2", "archive": "arXiv", "archive_location": "2605.15777", "author": [ { "literal": "Kean Shi" }, { "literal": "Zihang Li" }, { "literal": "Tianyi Ma" }, { "literal": "Zengji Tu" }, { "literal": "Jialong Wu" }, { "literal": "Xinbo Xu" }, { "literal": "Qingyao Yang" }, { "literal": "Ruoyu Wu" }, { "literal": "Weichu Xie" }, { "literal": "Ming Wu" }, { "literal": "Jason Zeng" }, { "literal": "Michael Heinrich" }, { "literal": "Elvis Zhang" }, { "literal": "Liang Chen" }, { "literal": "Kuan Li" }, { "literal": "Baobao Chang" } ], "citation-key": "arxiv_2605_15777", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "c4ee2439dd31a375b6f64cbc7f1d0ff75c87e1346b1627a53db145ae42682f8a", "submitted_at": "2026-05-15T09:35:15Z", "updated_at": "2026-05-24T16:35:56Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2605.15777v2" }, "canonical_source_url": "https://arxiv.org/abs/2605.15777", "catalog_id": "arxiv:2605.15777", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-05-15", "identifiers": { "arxiv": "2605.15777" }, "record_updated_at": "2026-05-24" }, "genre": "preprint", "id": "arxiv:2605.15777", "issued": { "date-parts": [ [ 2026, 5, 15 ] ] }, "note": "First published: 2026-05-15; Recorded arXiv version: v2; Recorded version updated: 2026-05-24T16:35:56Z.", "title": "SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows?", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2605.29534v1", "archive": "arXiv", "archive_location": "2605.29534", "author": [ { "literal": "Yuxiang Chai" }, { "literal": "Han Xiao" }, { "literal": "Xinyu Fu" }, { "literal": "Jinpeng Chen" }, { "literal": "Rui Liu" }, { "literal": "Hongsheng Li" } ], "citation-key": "arxiv_2605_29534", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "461790f351a2816004027cc58d07f79ee7c670ed5064c59b1651a4efb84ff2bf", "submitted_at": "2026-05-28T07:49:09Z", "updated_at": "2026-05-28T07:49:09Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2605.29534v1" }, "canonical_source_url": "https://arxiv.org/abs/2605.29534", "catalog_id": "arxiv:2605.29534", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-05-28", "identifiers": { "arxiv": "2605.29534" }, "record_updated_at": "2026-05-28" }, "genre": "preprint", "id": "arxiv:2605.29534", "issued": { "date-parts": [ [ 2026, 5, 28 ] ] }, "note": "First published: 2026-05-28; Recorded arXiv version: v1; Recorded version updated: 2026-05-28T07:49:09Z.", "title": "UI-KOBE: Knowledge-Oriented Behavior Exploration for Lightweight Graph-Guided GUI Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2606.01533v1", "archive": "arXiv", "archive_location": "2606.01533", "author": [ { "literal": "Jing Yu Koh" }, { "literal": "Ruslan Salakhutdinov" }, { "literal": "Daniel Fried" } ], "citation-key": "arxiv_2606_01533", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "acc6e8af676b7de5393469d379746e674ccd662f570399622d0d5b1b0e882939", "submitted_at": "2026-06-01T01:29:36Z", "updated_at": "2026-06-01T01:29:36Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.01533v1" }, "canonical_source_url": "https://arxiv.org/abs/2606.01533", "catalog_id": "arxiv:2606.01533", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-06-01", "identifiers": { "arxiv": "2606.01533" }, "record_updated_at": "2026-06-01" }, "genre": "preprint", "id": "arxiv:2606.01533", "issued": { "date-parts": [ [ 2026, 6, 1 ] ] }, "note": "First published: 2026-06-01; Recorded arXiv version: v1; Recorded version updated: 2026-06-01T01:29:36Z.", "title": "Multi-Agent Computer Use", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2606.07027v2", "archive": "arXiv", "archive_location": "2606.07027", "author": [ { "literal": "Haojie Hao" }, { "literal": "Longkun Hao" }, { "literal": "Yihang Lou" }, { "literal": "Yan Bai" }, { "literal": "Zhenyang Li" }, { "literal": "Zhichao Yang" }, { "literal": "Dongshuo Huang" }, { "literal": "Hongyu Lin" }, { "literal": "Lanqing Hong" }, { "literal": "Jiakai Wang" }, { "literal": "Xianglong Liu" } ], "citation-key": "arxiv_2606_07027", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "2d530152b210a69cb458fecb036c10621c72daccfbcac7c8306c7e248c87cf14", "submitted_at": "2026-06-05T08:17:28Z", "updated_at": "2026-06-12T07:28:19Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.07027v2" }, "canonical_source_url": "https://arxiv.org/abs/2606.07027", "catalog_id": "arxiv:2606.07027", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-06-05", "identifiers": { "arxiv": "2606.07027" }, "record_updated_at": "2026-06-12" }, "genre": "preprint", "id": "arxiv:2606.07027", "issued": { "date-parts": [ [ 2026, 6, 5 ] ] }, "note": "First published: 2026-06-05; Recorded arXiv version: v2; Recorded version updated: 2026-06-12T07:28:19Z.", "title": "StainFlow: Entity-Stain Tracking and Evidence Linking for Process Rewards in GUI Agents", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2606.21654v2", "archive": "arXiv", "archive_location": "2606.21654", "author": [ { "literal": "Vincent Siu" }, { "literal": "Manasi Sharma" }, { "literal": "Dawn Song" }, { "literal": "Daniel Yue Zhang" }, { "literal": "Chenguang Wang" }, { "literal": "Ying Liu" } ], "citation-key": "arxiv_2606_21654", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "69e668a9491a31e8d8849bab8cc63eede9878fff10ccb0ae0a211610e90819b3", "submitted_at": "2026-06-19T18:00:14Z", "updated_at": "2026-08-19T17:29:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2606.21654v2" }, "canonical_source_url": "https://arxiv.org/abs/2606.21654", "catalog_id": "arxiv:2606.21654", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-06-19", "identifiers": { "arxiv": "2606.21654" }, "record_updated_at": "2026-08-19" }, "genre": "preprint", "id": "arxiv:2606.21654", "issued": { "date-parts": [ [ 2026, 6, 19 ] ] }, "note": "First published: 2026-06-19; Recorded arXiv version: v2; Recorded version updated: 2026-08-19T17:29:04Z.", "title": "ChainWorld: Composing Long-Horizon Desktop Workloads from Atomic OSWorld Tasks", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2606.24525v1", "archive": "arXiv", "archive_location": "2606.24525", "author": [ { "literal": "Jiachen Qian" } ], "citation-key": "arxiv_2606_24525", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a3d0341f41a8ecba57fb7ed9e4b266bb4a176639c742be2b4fc98555754f5744", "submitted_at": "2026-06-23T12:57:02Z", "updated_at": "2026-06-23T12:57:02Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2606.24525v1" }, "canonical_source_url": "https://arxiv.org/abs/2606.24525", "catalog_id": "arxiv:2606.24525", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-06-23", "identifiers": { "arxiv": "2606.24525" }, "record_updated_at": "2026-06-23" }, "genre": "preprint", "id": "arxiv:2606.24525", "issued": { "date-parts": [ [ 2026, 6, 23 ] ] }, "note": "First published: 2026-06-23; Recorded arXiv version: v1; Recorded version updated: 2026-06-23T12:57:02Z.", "title": "VisCritic: Visual State Comparison as Process Reward for GUI Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2607.04334v2", "archive": "arXiv", "archive_location": "2607.04334", "author": [ { "literal": "Guijia Zhang" }, { "literal": "Yuxun Chen" }, { "literal": "Yuheng Qi" }, { "literal": "Harry Yang" } ], "citation-key": "arxiv_2607_04334", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e7fdae5d6e5ea17639fe0e2b8e2d655df7201a438d84acdeb3e0d02a78051ae2", "submitted_at": "2026-07-05T14:33:29Z", "updated_at": "2026-07-18T16:56:38Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.04334v2" }, "canonical_source_url": "https://arxiv.org/abs/2607.04334", "catalog_id": "arxiv:2607.04334", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-07-05", "identifiers": { "arxiv": "2607.04334" }, "record_updated_at": "2026-07-18" }, "genre": "preprint", "id": "arxiv:2607.04334", "issued": { "date-parts": [ [ 2026, 7, 5 ] ] }, "note": "First published: 2026-07-05; Recorded arXiv version: v2; Recorded version updated: 2026-07-18T16:56:38Z.", "title": "Do GUI Agents Believe Their Eyes? Diagnosing State-Belief Reliance on Pixels versus Structure", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2607.11388v1", "archive": "arXiv", "archive_location": "2607.11388", "author": [ { "literal": "Wenyi Wu" }, { "literal": "Sibo Zhu" }, { "literal": "Kun Zhou" }, { "literal": "Aayush Salvi" }, { "literal": "Zixuan Song" }, { "literal": "Biwei Huang" } ], "citation-key": "arxiv_2607_11388", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "40daa1b180a0c2eaba7b907a0f61b8dd8819cb3033b9da118162330d97cfecf7", "submitted_at": "2026-07-13T10:48:25Z", "updated_at": "2026-07-13T10:48:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.11388v1" }, "canonical_source_url": "https://arxiv.org/abs/2607.11388", "catalog_id": "arxiv:2607.11388", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-07-13", "identifiers": { "arxiv": "2607.11388" }, "record_updated_at": "2026-07-13" }, "genre": "preprint", "id": "arxiv:2607.11388", "issued": { "date-parts": [ [ 2026, 7, 13 ] ] }, "note": "First published: 2026-07-13; Recorded arXiv version: v1; Recorded version updated: 2026-07-13T10:48:25Z.", "title": "StructAgent: Harness Long-horizon Digital Agents with Unified Causal Structure", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2607.22798v1", "archive": "arXiv", "archive_location": "2607.22798", "author": [ { "literal": "Yan Yang" }, { "literal": "Xiangru Jian" }, { "literal": "Ziyang Luo" }, { "literal": "Zirui Zhao" }, { "literal": "Yutong Dai" }, { "literal": "Ziji Shi" }, { "literal": "Hanshu Yan" }, { "literal": "Jun Hao Liew" }, { "literal": "Silvio Savarese" }, { "literal": "Junnan Li" } ], "citation-key": "arxiv_2607_22798", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a010fbef09d8359550cf156bf0dcba873f25a0d6fb8f118d53bf6e23715cf203", "submitted_at": "2026-07-24T14:17:03Z", "updated_at": "2026-07-24T14:17:03Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2607.22798v1" }, "canonical_source_url": "https://arxiv.org/abs/2607.22798", "catalog_id": "arxiv:2607.22798", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-07-24", "identifiers": { "arxiv": "2607.22798" }, "record_updated_at": "2026-07-24" }, "genre": "preprint", "id": "arxiv:2607.22798", "issued": { "date-parts": [ [ 2026, 7, 24 ] ] }, "note": "First published: 2026-07-24; Recorded arXiv version: v1; Recorded version updated: 2026-07-24T14:17:03Z.", "title": "StateAct: Program State, before Pixels, for Long-Horizon Computer-Use Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2607.26041v2", "archive": "arXiv", "archive_location": "2607.26041", "author": [ { "literal": "Abhishek Pillai" }, { "literal": "Samir Kumar Nayak" }, { "literal": "Yuan Chen" } ], "citation-key": "arxiv_2607_26041", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "352c892a52692301e14125e126a52508feb39ef7cd034d7cb0c48a27fb54b8ac", "submitted_at": "2026-07-28T17:49:51Z", "updated_at": "2026-07-29T19:07:50Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2607.26041v2" }, "canonical_source_url": "https://arxiv.org/abs/2607.26041", "catalog_id": "arxiv:2607.26041", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-07-28", "identifiers": { "arxiv": "2607.26041" }, "record_updated_at": "2026-07-29" }, "genre": "preprint", "id": "arxiv:2607.26041", "issued": { "date-parts": [ [ 2026, 7, 28 ] ] }, "note": "First published: 2026-07-28; Recorded arXiv version: v2; Recorded version updated: 2026-07-29T19:07:50Z.", "title": "Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2608.02643v2", "archive": "arXiv", "archive_location": "2608.02643", "author": [ { "literal": "Weijia Zhang" }, { "literal": "Kunlun Zhu" }, { "literal": "Zeyi Liu" }, { "literal": "Yinting Chen" }, { "literal": "Tianyi Ma" }, { "literal": "Jiateng Liu" }, { "literal": "Jiaxun Zhang" }, { "literal": "Bingxuan Li" }, { "literal": "Xiangru Tang" }, { "literal": "Heng Ji" }, { "literal": "Jiaxuan You" } ], "citation-key": "arxiv_2608_02643", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "a11e065bef4cf860c36ca2f682801ca21b552fbd40bd1b6f5d9ae39b56e7d77a", "submitted_at": "2026-07-31T13:59:45Z", "updated_at": "2026-09-06T04:43:23Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.02643v2" }, "canonical_source_url": "https://arxiv.org/abs/2608.02643", "catalog_id": "arxiv:2608.02643", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-07-31", "identifiers": { "arxiv": "2608.02643" }, "record_updated_at": "2026-09-06" }, "genre": "preprint", "id": "arxiv:2608.02643", "issued": { "date-parts": [ [ 2026, 7, 31 ] ] }, "note": "First published: 2026-07-31; Recorded arXiv version: v2; Recorded version updated: 2026-09-06T04:43:23Z.", "title": "CUADebug: Diagnosing and Repairing Computer-Use Agent Failures", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2608.05587v1", "archive": "arXiv", "archive_location": "2608.05587", "author": [ { "literal": "Linqiang Guo" }, { "literal": "Wei Liu" }, { "literal": "Li Gu" }, { "literal": "Yang Wang" }, { "literal": "Tse-Hsun" }, { "literal": "Chen" } ], "citation-key": "arxiv_2608_05587", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "dc7934927c82a43e6ae6a246ade1d50c4551cc49d263ca67cfb32906cbec9ea4", "submitted_at": "2026-08-06T04:17:27Z", "updated_at": "2026-08-06T04:17:27Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.05587v1" }, "canonical_source_url": "https://arxiv.org/abs/2608.05587", "catalog_id": "arxiv:2608.05587", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-08-06", "identifiers": { "arxiv": "2608.05587" }, "record_updated_at": "2026-08-06" }, "genre": "preprint", "id": "arxiv:2608.05587", "issued": { "date-parts": [ [ 2026, 8, 6 ] ] }, "note": "First published: 2026-08-06; Recorded arXiv version: v1; Recorded version updated: 2026-08-06T04:17:27Z.", "title": "StepReflect: Structured UI Transition Reflection for Mobile GUI Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2608.12429v1", "archive": "arXiv", "archive_location": "2608.12429", "author": [ { "literal": "Ruitao Wang" }, { "literal": "Yuwen Hao" }, { "literal": "Menglin Yang" } ], "citation-key": "arxiv_2608_12429", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "bf543ef0b77eb6fe430c8cef5382932f13e4e4136aaba9c14c44b5a591d6612a", "submitted_at": "2026-08-12T12:03:20Z", "updated_at": "2026-08-12T12:03:20Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2608.12429v1" }, "canonical_source_url": "https://arxiv.org/abs/2608.12429", "catalog_id": "arxiv:2608.12429", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-08-12", "identifiers": { "arxiv": "2608.12429" }, "record_updated_at": "2026-08-12" }, "genre": "preprint", "id": "arxiv:2608.12429", "issued": { "date-parts": [ [ 2026, 8, 12 ] ] }, "note": "First published: 2026-08-12; Recorded arXiv version: v1; Recorded version updated: 2026-08-12T12:03:20Z.", "title": "SynWeaver: Website-Prior Task and Trajectory Co-Synthesis for Web Agents", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2608.25777v2", "archive": "arXiv", "archive_location": "2608.25777", "author": [ { "literal": "Weiming Li" }, { "literal": "Helen Paik" }, { "literal": "Yulei Sui" } ], "citation-key": "arxiv_2608_25777", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "af6b60c6a39a990e950cf4b98f905f2d3aa5ae62d4055630c25a650508ec8bf8", "submitted_at": "2026-08-26T13:20:27Z", "updated_at": "2026-09-25T23:54:04Z", "version": 2, "versioned_url": "https://arxiv.org/abs/2608.25777v2" }, "canonical_source_url": "https://arxiv.org/abs/2608.25777", "catalog_id": "arxiv:2608.25777", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-08-26", "identifiers": { "arxiv": "2608.25777" }, "record_updated_at": "2026-09-25" }, "genre": "preprint", "id": "arxiv:2608.25777", "issued": { "date-parts": [ [ 2026, 8, 26 ] ] }, "note": "First published: 2026-08-26; Recorded arXiv version: v2; Recorded version updated: 2026-09-25T23:54:04Z.", "title": "LocalLSTC: A Long Short-Term Control Architecture for Locally Deployed GUI Agents", "type": "article", "version": "2" }, { "URL": "https://arxiv.org/abs/2609.10451v1", "archive": "arXiv", "archive_location": "2609.10451", "author": [ { "literal": "Zixiang Chen" }, { "literal": "Yuheng Lu" }, { "literal": "Zihao Cheng" }, { "literal": "Zeming Liu" }, { "literal": "Jizeng Bai" }, { "literal": "Ziye Huang" }, { "literal": "Zhiyin Lin" }, { "literal": "Zihan Li" }, { "literal": "Yuhang Guo" }, { "literal": "Yunhong Wang" }, { "literal": "Haifeng Wang" } ], "citation-key": "arxiv_2609_10451", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "5ec17516bce87a8e349983acac14415a6ceeb12c167fce57aa37d6157cd3cd83", "submitted_at": "2026-09-09T16:59:59Z", "updated_at": "2026-09-09T16:59:59Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.10451v1" }, "canonical_source_url": "https://arxiv.org/abs/2609.10451", "catalog_id": "arxiv:2609.10451", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-09-09", "identifiers": { "arxiv": "2609.10451" }, "record_updated_at": "2026-09-09" }, "genre": "preprint", "id": "arxiv:2609.10451", "issued": { "date-parts": [ [ 2026, 9, 9 ] ] }, "note": "First published: 2026-09-09; Recorded arXiv version: v1; Recorded version updated: 2026-09-09T16:59:59Z.", "title": "JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2609.15457v1", "archive": "arXiv", "archive_location": "2609.15457", "author": [ { "literal": "Shengjie Jin" }, { "literal": "Zelong Sun" }, { "literal": "Hengbo Xu" }, { "literal": "Yanbiao Ma" }, { "literal": "Zhiwu Lu" } ], "citation-key": "arxiv_2609_15457", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "e8a98337941972fadb725906b9d1315a0ca5921859bd04780cacbabb0500e6a5", "submitted_at": "2026-09-14T12:22:26Z", "updated_at": "2026-09-14T12:22:26Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.15457v1" }, "canonical_source_url": "https://arxiv.org/abs/2609.15457", "catalog_id": "arxiv:2609.15457", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-09-14", "identifiers": { "arxiv": "2609.15457" }, "record_updated_at": "2026-09-14" }, "genre": "preprint", "id": "arxiv:2609.15457", "issued": { "date-parts": [ [ 2026, 9, 14 ] ] }, "note": "First published: 2026-09-14; Recorded arXiv version: v1; Recorded version updated: 2026-09-14T12:22:26Z.", "title": "AnchorGUI: Asymmetric Memory for Dual-Scale Learning in GUI Navigation", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2609.25769v1", "archive": "arXiv", "archive_location": "2609.25769", "author": [ { "literal": "Yan Zhang" }, { "literal": "Pei Fu" }, { "literal": "Daiqing Wu" }, { "literal": "Huawen Shen" }, { "literal": "Ruoceng Zhang" }, { "literal": "Shaojie Zhang" }, { "literal": "Jiahui Yang" }, { "literal": "Yu Zhou" }, { "literal": "Can Ma" }, { "literal": "Zhenbo Luo" }, { "literal": "Jian Luan" } ], "citation-key": "arxiv_2609_25769", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "d184b111aa2a9ad65dd57391b72bcef369c2bcd01ceb8e4f00f7a2320db856d8", "submitted_at": "2026-09-22T07:02:35Z", "updated_at": "2026-09-22T07:02:35Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.25769v1" }, "canonical_source_url": "https://arxiv.org/abs/2609.25769", "catalog_id": "arxiv:2609.25769", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-09-22", "identifiers": { "arxiv": "2609.25769" }, "record_updated_at": "2026-09-22" }, "genre": "preprint", "id": "arxiv:2609.25769", "issued": { "date-parts": [ [ 2026, 9, 22 ] ] }, "note": "First published: 2026-09-22; Recorded arXiv version: v1; Recorded version updated: 2026-09-22T07:02:35Z.", "title": "Towards Omni-dimensional GUI Agent Navigation with Masked Trajectory Prediction", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2609.32679v1", "archive": "arXiv", "archive_location": "2609.32679", "author": [ { "literal": "Dongsheng Liu" }, { "literal": "Chao Jin" }, { "literal": "Wenkui Yang" }, { "literal": "Hejin Wang" }, { "literal": "Junwei Yang" }, { "literal": "Zeren Zhang" }, { "literal": "Ziwei Chen" }, { "literal": "Huaibo Huang" }, { "literal": "Jie Cao" }, { "literal": "Ran He" } ], "citation-key": "arxiv_2609_32679", "custom": { "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" }, "canonical_source_url": "https://arxiv.org/abs/2609.32679", "catalog_id": "arxiv:2609.32679", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-09-26", "identifiers": { "arxiv": "2609.32679" }, "record_updated_at": "2026-09-26" }, "genre": "preprint", "id": "arxiv:2609.32679", "issued": { "date-parts": [ [ 2026, 9, 26 ] ] }, "note": "First published: 2026-09-26; Recorded arXiv version: v1; Recorded version updated: 2026-09-26T14:35:45Z.", "title": "The GUI Is Not the State: Diagnosing State Aliasing in GUI World Models", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2609.34769v1", "archive": "arXiv", "archive_location": "2609.34769", "author": [ { "literal": "Bingo Zhang" }, { "literal": "Haochuan Lu" }, { "literal": "Zongjie Li" }, { "literal": "Genjian Li" }, { "literal": "Ari Yu Zhang" }, { "literal": "Chaozheng Wang" } ], "citation-key": "arxiv_2609_34769", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "f81fd07262dccf6783729e4336fbd77835f2fcdea030eef8dc2272e7ef82b8fc", "submitted_at": "2026-09-28T09:49:25Z", "updated_at": "2026-09-28T09:49:25Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.34769v1" }, "canonical_source_url": "https://arxiv.org/abs/2609.34769", "catalog_id": "arxiv:2609.34769", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "adjacent", "first_published_at": "2026-09-28", "identifiers": { "arxiv": "2609.34769" }, "record_updated_at": "2026-09-28" }, "genre": "preprint", "id": "arxiv:2609.34769", "issued": { "date-parts": [ [ 2026, 9, 28 ] ] }, "note": "First published: 2026-09-28; Recorded arXiv version: v1; Recorded version updated: 2026-09-28T09:49:25Z.", "title": "LongPuzzleBench: Evaluating GUI Agents on Long-Horizon Visual Puzzles", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2609.35814v1", "archive": "arXiv", "archive_location": "2609.35814", "author": [ { "literal": "Xunjian Yin" }, { "literal": "Tianchen Guan" }, { "literal": "Jinao Wang" }, { "literal": "Weili Cao" }, { "literal": "Daisy Xinlei Lin" }, { "literal": "Royce Cheng-Yue" }, { "literal": "Keagan Long" }, { "literal": "Kyle Wong" }, { "literal": "Bhuwan Dhingra" }, { "literal": "Xiangjun Wang" }, { "literal": "Shuyan Zhou" } ], "citation-key": "arxiv_2609_35814", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "197cfb8424584955bc85602d6559037ad31ee815f89b45dc8f9f18c8636552ad", "submitted_at": "2026-09-20T16:16:23Z", "updated_at": "2026-09-20T16:16:23Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2609.35814v1" }, "canonical_source_url": "https://arxiv.org/abs/2609.35814", "catalog_id": "arxiv:2609.35814", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-09-20", "identifiers": { "arxiv": "2609.35814" }, "record_updated_at": "2026-09-20" }, "genre": "preprint", "id": "arxiv:2609.35814", "issued": { "date-parts": [ [ 2026, 9, 20 ] ] }, "note": "First published: 2026-09-20; Recorded arXiv version: v1; Recorded version updated: 2026-09-20T16:16:23Z.", "title": "Constructing Challenging Browser-Use Tasks by Controlled Environment Interventions", "type": "article", "version": "1" }, { "URL": "https://arxiv.org/abs/2610.22408v1", "archive": "arXiv", "archive_location": "2610.22408", "author": [ { "literal": "Tengda Guo" }, { "literal": "Jiawang Hu" }, { "literal": "Yifei Tao" }, { "literal": "Yin Wang" }, { "literal": "Haotian Hu" }, { "literal": "Fanyi Wang" }, { "literal": "Liujuan Tang" } ], "citation-key": "arxiv_2610_22408", "custom": { "arxiv_snapshot": { "abstract_hash_method": "sha256-utf8-whitespace-collapsed-v1", "abstract_sha256": "025c7018c86ae4de79a659242070b2f91c5913333e9ecd4bd7fd357d13d27a34", "submitted_at": "2026-10-14T14:39:41Z", "updated_at": "2026-10-14T14:39:41Z", "version": 1, "versioned_url": "https://arxiv.org/abs/2610.22408v1" }, "canonical_source_url": "https://arxiv.org/abs/2610.22408", "catalog_id": "arxiv:2610.22408", "catalog_kind": "paper", "catalog_review_status": "curated", "catalog_scope": "core", "first_published_at": "2026-10-14", "identifiers": { "arxiv": "2610.22408" }, "record_updated_at": "2026-10-14" }, "genre": "preprint", "id": "arxiv:2610.22408", "issued": { "date-parts": [ [ 2026, 10, 14 ] ] }, "note": "First published: 2026-10-14; Recorded arXiv version: v1; Recorded version updated: 2026-10-14T14:39:41Z.", "title": "ECHOES: Seeing Beyond the Current Image via Recursive History Compression and Privileged On-Policy Distillation", "type": "article", "version": "1" } ]