{"active_bibliographic_resources":2599,"active_kind_counts":{"agent":41,"benchmark":83,"company":16,"dataset":85,"framework":72,"infrastructure":14,"model":98,"paper":2584,"product":20,"survey":15,"tool":20},"active_resources":3048,"asset_linked_paper_ids":["arxiv:1802.08802","arxiv:2101.09465","arxiv:2105.13231","arxiv:2107.13731","arxiv:2202.02312","arxiv:2205.11029","arxiv:2207.01206","arxiv:2209.08199","arxiv:2303.17491","arxiv:2305.08144","arxiv:2306.06070","arxiv:2306.07863","arxiv:2307.10088","arxiv:2307.13854","arxiv:2308.03688","arxiv:2308.15272","arxiv:2309.11436","arxiv:2310.10634","arxiv:2310.11441","arxiv:2311.01767","arxiv:2312.03003","arxiv:2312.08914","arxiv:2312.11190","arxiv:2312.13771","arxiv:2312.15820","arxiv:2401.01614","arxiv:2401.04124","arxiv:2401.10935","arxiv:2401.13649","arxiv:2401.13919","arxiv:2401.16158","arxiv:2402.05930","arxiv:2402.06596","arxiv:2402.07456","arxiv:2402.07939","arxiv:2402.07945","arxiv:2402.11941","arxiv:2402.15057","arxiv:2402.17553","arxiv:2403.02713","arxiv:2403.03186","arxiv:2403.07718","arxiv:2404.05719","arxiv:2404.05955","arxiv:2404.07972","arxiv:2404.09992","arxiv:2404.16054","arxiv:2405.14573","arxiv:2406.01014","arxiv:2406.03679","arxiv:2406.06947","arxiv:2406.08451","arxiv:2406.10227","arxiv:2406.10819","arxiv:2406.11317","arxiv:2406.11896","arxiv:2406.12373","arxiv:2406.12814","arxiv:2406.13264","arxiv:2406.19263","arxiv:2407.01511","arxiv:2407.05291","arxiv:2407.09018","arxiv:2407.10956","arxiv:2407.13032","arxiv:2407.17490","arxiv:2407.18901","arxiv:2408.00203","arxiv:2408.06327","arxiv:2409.07429","arxiv:2409.08264","arxiv:2409.14337","arxiv:2409.14818","arxiv:2410.00467","arxiv:2410.02810","arxiv:2410.02907","arxiv:2410.05243","arxiv:2410.08164","arxiv:2410.13757","arxiv:2410.13824","arxiv:2410.14803","arxiv:2410.15164","arxiv:2410.23218","arxiv:2410.24024","arxiv:2411.02337","arxiv:2411.06559","arxiv:2411.10323","arxiv:2411.17465","arxiv:2412.00402","arxiv:2412.04454","arxiv:2412.05467","arxiv:2412.09605","arxiv:2412.10840","arxiv:2412.13194","arxiv:2412.14161","arxiv:2412.16256","arxiv:2412.17589","arxiv:2412.18116","arxiv:2412.19723","arxiv:2501.11425","arxiv:2501.11733","arxiv:2501.12326","arxiv:2502.01977","arxiv:2502.08047","arxiv:2502.14282","arxiv:2502.15760","arxiv:2502.18906","arxiv:2503.00401","arxiv:2503.02268","arxiv:2503.02950","arxiv:2503.03196","arxiv:2503.03743","arxiv:2503.04957","arxiv:2503.05143","arxiv:2503.09263","arxiv:2503.09572","arxiv:2503.09780","arxiv:2503.12532","arxiv:2503.15661","arxiv:2503.16465","arxiv:2503.21620","arxiv:2504.00906","arxiv:2504.01382","arxiv:2504.07079","arxiv:2504.07981","arxiv:2504.08942","arxiv:2504.10127","arxiv:2504.10458","arxiv:2504.12679","arxiv:2504.14239","arxiv:2504.21024","arxiv:2505.11891","arxiv:2505.12632","arxiv:2505.13227","arxiv:2505.13909","arxiv:2505.15277","arxiv:2505.15810","arxiv:2505.16282","arxiv:2505.16421","arxiv:2505.19897","arxiv:2505.21496","arxiv:2505.21936","arxiv:2505.22942","arxiv:2505.23762","arxiv:2505.24838","arxiv:2506.00618","arxiv:2506.01391","arxiv:2506.02865","arxiv:2506.03143","arxiv:2506.03533","arxiv:2506.07672","arxiv:2506.14205","arxiv:2506.14477","arxiv:2506.14866","arxiv:2506.16042","arxiv:2506.23491","arxiv:2507.00008","arxiv:2507.15846","arxiv:2507.16853","arxiv:2507.19132","arxiv:2507.19478","arxiv:2507.22025","arxiv:2507.23779","arxiv:2508.05615","arxiv:2508.05731","arxiv:2508.09123","arxiv:2508.10833","arxiv:2508.14040","arxiv:2508.15144","arxiv:2508.18689","arxiv:2509.11543","arxiv:2509.15221","arxiv:2509.15738","arxiv:2509.18119","arxiv:2510.14621","arxiv:2510.20286","arxiv:2510.24411","arxiv:2511.04307","arxiv:2511.19663","arxiv:2511.20766","arxiv:2512.10371","arxiv:2512.14014","arxiv:2512.16295","arxiv:2512.19432","arxiv:2512.22047","arxiv:2512.24965","arxiv:2601.07181","arxiv:2601.15876","arxiv:2602.06391","arxiv:2602.09082","arxiv:2602.09662","arxiv:2602.09856","arxiv:2602.16855","arxiv:2603.08533","arxiv:2603.18429","arxiv:2603.25226","arxiv:2604.06240","arxiv:2604.11784","arxiv:2604.13019","arxiv:2604.27776","arxiv:2605.15777","arxiv:2605.25624","arxiv:2605.26114","arxiv:2605.28534","arxiv:2606.20785","arxiv:2606.22557","arxiv:2607.11185","arxiv:2609.00028","arxiv:2609.07712","arxiv:2609.12394","arxiv:2610.00948","arxiv:2610.04116","arxiv:2610.12984","arxiv:2610.14157","arxiv:2610.16234","arxiv:2610.17928","arxiv:2610.19100","url:6262c757080f8838e62c","url:a4cdb2ee0d07d3980651","url:b560b5d6abaab98483f6"],"asset_linked_papers":220,"benchmark_profiles":60,"benchmarks":83,"boundary":"These statistics describe this selected catalog snapshot, not field-wide prevalence, completeness, publication quality, or independently reproduced results. Topic and environment memberships overlap.","catalog_sha256":"7f71a1683ab013e587a20086c3e357956981f517a24ae8bc8e81791108c66e4f","dataset_profiles":66,"datasets":85,"metadata_only_bibliographic_resources":1476,"paper_asset_pairs":311,"primary_discovery_sources":47,"record_fingerprint_method":"SHA256 of the original full record serialized as UTF-8 JSON with sorted keys, no ASCII escaping, and compact separators.","release":"v2026.10.18","reviewed_bibliography_environment_memberships":{"android":175,"cross-platform":110,"desktop":205,"harmonyos":2,"ios":9,"linux":18,"macos":14,"unspecified":334,"web":426,"windows":37},"reviewed_bibliography_scope":{"adjacent":52,"core":1071},"reviewed_bibliography_topic_memberships":{"action-prediction":26,"automation":83,"cross-app":35,"data-generation":258,"efficiency":138,"evaluation":548,"grounding":244,"human-in-the-loop":71,"imitation-learning":44,"industry":11,"interaction-history":45,"long-horizon":88,"memory":140,"mobile-navigation":181,"observability":33,"planning":206,"recording":15,"reinforcement-learning":180,"reward-models":42,"robustness":162,"safety-security":158,"sandbox":38,"science":15,"screen-understanding":160,"self-reflection":27,"state-tracking":44,"survey":16,"tool-integration":90,"training-free":28,"ui-design":19,"video":29,"web-navigation":198,"workflow":78,"world-models":35},"reviewed_bibliography_year":{"1991":1,"2007":2,"2009":1,"2010":1,"2016":2,"2017":3,"2018":5,"2019":1,"2020":4,"2021":8,"2022":13,"2023":41,"2024":180,"2025":355,"2026":506},"reviewed_nonbibliographic_resources":449,"reviewed_papers":1108,"reviewed_papers_and_surveys":1123,"reviewed_records":[{"environments":["web"],"id":"arxiv:1602.02261","kind":"paper","record_sha256":"f86c73e02bd0aa44b2fff5312837e5903b8dbc16ae0e05eab9c088a6c3774e9f","scope":"core","source_url":"https://arxiv.org/abs/1602.02261","title":"End-to-End Goal-Driven Web Navigation","topics":["web-navigation","evaluation","planning"],"year":"2016"},{"environments":["web"],"id":"arxiv:1802.08802","kind":"paper","record_sha256":"8e68e80f6e809fbc7f0d4ab55042c7758019e4f49b46fc91c996f305204783f0","scope":"core","source_url":"https://arxiv.org/abs/1802.08802","title":"Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration","topics":["reinforcement-learning","workflow"],"year":"2018"},{"environments":["web"],"id":"arxiv:1808.09132","kind":"paper","record_sha256":"9801b6dc68f704d00437dc919fa03a4d57f34a7df1d63a240b86f405e9e46d66","scope":"core","source_url":"https://arxiv.org/abs/1808.09132","title":"Mapping Natural Language Commands to Web Elements","topics":["grounding","data-generation","evaluation"],"year":"2018"},{"environments":["unspecified"],"id":"arxiv:1909.00031","kind":"paper","record_sha256":"f43c1273f7d55aab61cf4db5854032a1742697227814a89bac2dbef1cf95ed90","scope":"core","source_url":"https://arxiv.org/abs/1909.00031","title":"Interactive Task and Concept Learning from Natural Language Instructions and GUI Demonstrations","topics":["human-in-the-loop","mobile-navigation","imitation-learning"],"year":"2019"},{"environments":["unspecified"],"id":"arxiv:2005.03776","kind":"paper","record_sha256":"95eb6614cbedee6c08b36740a4060735f90e0e43b35264941f52b9cdf335d0b4","scope":"core","source_url":"https://arxiv.org/abs/2005.03776","title":"Mapping Natural Language Instructions to Mobile UI Action Sequences","topics":["mobile-navigation","grounding","data-generation","action-prediction"],"year":"2020"},{"environments":["unspecified"],"id":"arxiv:2008.05132","kind":"paper","record_sha256":"7731d93e5abe0b67cb10ed4e7a0a3a7c306dd8b8f46a9ce993cf4961bf3171b8","scope":"core","source_url":"https://arxiv.org/abs/2008.05132","title":"Object Detection for Graphical User Interface: Old Fashioned or Deep Learning or a Combination?","topics":["grounding","screen-understanding","evaluation"],"year":"2020"},{"environments":["unspecified"],"id":"arxiv:2010.04295","kind":"paper","record_sha256":"94a14799f767a59f5075eded64f7643ea8fba7f605eb92a91f29c4b859ea6d99","scope":"core","source_url":"https://arxiv.org/abs/2010.04295","title":"Widget Captioning: Generating Natural Language Description for Mobile User Interface Elements","topics":["screen-understanding","grounding"],"year":"2020"},{"environments":["unspecified"],"id":"arxiv:2012.12350","kind":"paper","record_sha256":"bfd9d45ad03f3a44eedba18e7adac45aced9f8bd0a1503182434462a9705616b","scope":"core","source_url":"https://arxiv.org/abs/2012.12350","title":"ActionBert: Leveraging User Actions for Semantic Understanding of User Interfaces","topics":["screen-understanding","grounding","action-prediction"],"year":"2020"},{"environments":["ios"],"id":"arxiv:2101.04893","kind":"paper","record_sha256":"8ba3dbd3966fe5ba1f07a81805f4698c70686b7bc6cbeb444b57345b1b789761","scope":"adjacent","source_url":"https://arxiv.org/abs/2101.04893","title":"Screen Recognition: Creating Accessibility Metadata for Mobile Applications from Pixels","topics":["screen-understanding"],"year":"2021"},{"environments":["web"],"id":"arxiv:2101.09465","kind":"paper","record_sha256":"3abb63a2bcfb82fe82e437c3f3faac4ac3a1421608746545688483a138ffc766","scope":"core","source_url":"https://arxiv.org/abs/2101.09465","title":"WebSRC: A Dataset for Web-Based Structural Reading Comprehension","topics":["screen-understanding","evaluation"],"year":"2021"},{"environments":["web"],"id":"arxiv:2103.16057","kind":"paper","record_sha256":"1e8cc9a0efb8876e5bd4b4101428b142cda49eb56652ce8dafd0058dffd06bb9","scope":"core","source_url":"https://arxiv.org/abs/2103.16057","title":"Grounding Open-Domain Instructions to Automate Web Support Tasks","topics":["grounding","web-navigation"],"year":"2021"},{"environments":["unspecified"],"id":"arxiv:2105.11941","kind":"paper","record_sha256":"da36b1613c7d1c2b644d40d0a8d067cfa95d07cc06fb834eb46f1dacf1fa2db3","scope":"core","source_url":"https://arxiv.org/abs/2105.11941","title":"Understanding Mobile GUI: from Pixel-Words to Screen-Sentences","topics":["screen-understanding","grounding","data-generation"],"year":"2021"},{"environments":["android"],"id":"arxiv:2105.13231","kind":"paper","record_sha256":"84e6c9f39712bac72c3146752340d4446bef22d62394fe8cd831e3cce8d4ff7c","scope":"core","source_url":"https://arxiv.org/abs/2105.13231","title":"AndroidEnv: A Reinforcement Learning Platform for Android","topics":["reinforcement-learning","sandbox"],"year":"2021"},{"environments":["unspecified"],"id":"arxiv:2107.13731","kind":"paper","record_sha256":"b60252cc522ba35bb0e4a392691a99a4b4289fb8c8cc279f52fe0b30a9966f8a","scope":"core","source_url":"https://arxiv.org/abs/2107.13731","title":"UIBert: Learning Generic Multimodal Representations for UI Understanding","topics":["screen-understanding"],"year":"2021"},{"environments":["unspecified"],"id":"arxiv:2108.03353","kind":"paper","record_sha256":"0f47f3213629f5c152ee4ccbf4ce8ba9a1c5846476726b0d6be2e6fff7ee6f31","scope":"core","source_url":"https://arxiv.org/abs/2108.03353","title":"Screen2Words: Automatic Mobile UI Summarization with Multimodal Learning","topics":["screen-understanding"],"year":"2021"},{"environments":["unspecified"],"id":"arxiv:2112.05692","kind":"paper","record_sha256":"aec1a1fe833e87904337319ee6960e72e93bc8d249807fd85a2a7bcfaa21926e","scope":"core","source_url":"https://arxiv.org/abs/2112.05692","title":"VUT: Versatile UI Transformer for Multi-Modal Multi-Task User Interface Modeling","topics":["screen-understanding","grounding","efficiency"],"year":"2021"},{"environments":["unspecified"],"id":"arxiv:2201.04100","kind":"paper","record_sha256":"839ee39b2e596374a6d0ffe292fc5bbca63e9eaed26f793cd15609aa18d1862b","scope":"core","source_url":"https://arxiv.org/abs/2201.04100","title":"Learning to Denoise Raw Mobile UI Layouts for Improving Datasets at Scale","topics":["screen-understanding","data-generation"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2202.02312","kind":"paper","record_sha256":"1d5739dd8be5acb7d574b3fe692a651d12ce2eed971c41c1473e835e7eb3698e","scope":"core","source_url":"https://arxiv.org/abs/2202.02312","title":"A Dataset for Interactive Vision-Language Navigation with Unknown Command Feasibility","topics":["mobile-navigation","human-in-the-loop","evaluation"],"year":"2022"},{"environments":["web"],"id":"arxiv:2202.08137","kind":"paper","record_sha256":"aaf3cdd8e0b4413bbb67edbed75d7c92086e8c12a469e344acb010c7fa3da1fb","scope":"core","source_url":"https://arxiv.org/abs/2202.08137","title":"A data-driven approach for learning to control computers","topics":["reinforcement-learning","automation"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2205.11029","kind":"paper","record_sha256":"07826d35bb0b927b20783543a9340bea78e33b5f0c4df81d258f64b5e621da7a","scope":"core","source_url":"https://arxiv.org/abs/2205.11029","title":"META-GUI: Towards Multi-modal Conversational Agents on Mobile GUI","topics":["mobile-navigation","human-in-the-loop"],"year":"2022"},{"environments":["web"],"id":"arxiv:2207.01206","kind":"paper","record_sha256":"45fd768c98211666474f1a363c4f821ce818507243f0e092a121d830c88b57fc","scope":"core","source_url":"https://arxiv.org/abs/2207.01206","title":"WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents","topics":["evaluation","web-navigation"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2209.08199","kind":"paper","record_sha256":"f47b4a25e626240951e2f49a2c9a6a362ebf5aa7f3cd453c47eb87720f247011","scope":"core","source_url":"https://arxiv.org/abs/2209.08199","title":"ScreenQA: Large-Scale Question-Answer Pairs over Mobile App Screenshots","topics":["screen-understanding","evaluation"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2209.08655","kind":"paper","record_sha256":"7ba80b557bd6a41b5b16f32a895d322f2759cd51d63da874ac8ba093384c1e31","scope":"core","source_url":"https://arxiv.org/abs/2209.08655","title":"Enabling Conversational Interaction with Mobile UI using Large Language Models","topics":["mobile-navigation","screen-understanding","training-free"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2209.14927","kind":"paper","record_sha256":"2b8b2d9a7d7d219e8cb0f6949ab0bda713c5e7cc1eea8a7c7c5869bfb4e40a54","scope":"core","source_url":"https://arxiv.org/abs/2209.14927","title":"Spotlight: Mobile UI Understanding using Vision-Language Models with a Focus","topics":["screen-understanding","grounding"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2209.15099","kind":"paper","record_sha256":"b12fb7d77ed5ab0f08bba1561fdb8509aae61272b47f3f292dc295c187d80e7e","scope":"core","source_url":"https://arxiv.org/abs/2209.15099","title":"MUG: Interactive Multimodal Grounding on User Interfaces","topics":["grounding","human-in-the-loop"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2210.02663","kind":"paper","record_sha256":"05e2b5b657737bff6cacd46e080476a74bdf59d75d24a7ed5ec1ba3ad0a0d1a4","scope":"core","source_url":"https://arxiv.org/abs/2210.02663","title":"Towards Better Semantic Understanding of Mobile Interfaces","topics":["screen-understanding","grounding","data-generation"],"year":"2022"},{"environments":["web"],"id":"arxiv:2210.03347","kind":"paper","record_sha256":"d4a1298eecb532aad3063699eaaefb7c218193ec4f3e3e6b745038c4d20b824a","scope":"adjacent","source_url":"https://arxiv.org/abs/2210.03347","title":"Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding","topics":["screen-understanding","data-generation"],"year":"2022"},{"environments":["web"],"id":"arxiv:2210.03945","kind":"paper","record_sha256":"eb1f2187eb84c6c9f0bb374ff97bbf3638d4e36479456f9a4214f7ea6bfd2140","scope":"core","source_url":"https://arxiv.org/abs/2210.03945","title":"Understanding HTML with Large Language Models","topics":["screen-understanding","web-navigation","data-generation"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2211.07615","kind":"paper","record_sha256":"85ce1769f67364b4b6ebc23661ea8d62c02f29b83be3c3eb8e4ffc74d21ec926","scope":"core","source_url":"https://arxiv.org/abs/2211.07615","title":"UGIF: UI Grounded Instruction Following","topics":["mobile-navigation","grounding"],"year":"2022"},{"environments":["unspecified"],"id":"arxiv:2301.10165","kind":"paper","record_sha256":"be2df608a4d41ee8546b138cf88aad7ac048e7df27fb51ad3e7bc582bdf4a3d2","scope":"core","source_url":"https://arxiv.org/abs/2301.10165","title":"Lexi: Self-Supervised Learning of the UI Language","topics":["screen-understanding","grounding","data-generation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2301.13280","kind":"paper","record_sha256":"e2c9a1dee02509ea153bd92f6c39220cba7564dd3ae9b4f705469cfff4955722","scope":"core","source_url":"https://arxiv.org/abs/2301.13280","title":"WebUI: A Dataset for Enhancing Visual UI Understanding with Web Semantics","topics":["screen-understanding","data-generation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2303.17491","kind":"paper","record_sha256":"45ceb43be3a1e44f022c5e1ac2700ddfdf4b9f80f0bcff4c9e966bb439ef511f","scope":"core","source_url":"https://arxiv.org/abs/2303.17491","title":"Language Models can Solve Computer Tasks","topics":["web-navigation","self-reflection","training-free"],"year":"2023"},{"environments":["android"],"id":"arxiv:2304.07061","kind":"paper","record_sha256":"591c8af5acde1c6424745ac72abd37188851c5446a9db68e0d4932e2cce460fe","scope":"core","source_url":"https://arxiv.org/abs/2304.07061","title":"DroidBot-GPT: GPT-powered UI Automation for Android","topics":["mobile-navigation","automation"],"year":"2023"},{"environments":["android"],"id":"arxiv:2305.08144","kind":"paper","record_sha256":"0e45f0f2314fb83e8db17a8efb9690243bf57b06736f36d44c0d80754127bab1","scope":"core","source_url":"https://arxiv.org/abs/2305.08144","title":"Mobile-Env: Building Qualified Evaluation Benchmarks for LLM-GUI Interaction","topics":["evaluation","sandbox"],"year":"2023"},{"environments":["web"],"id":"arxiv:2305.11854","kind":"paper","record_sha256":"9a9fac749e919cd5de4744c1a6c0ee19f1c2716cbf4244aca93dacf8eeba8c31","scope":"core","source_url":"https://arxiv.org/abs/2305.11854","title":"Multimodal Web Navigation with Instruction-Finetuned Foundation Models","topics":["web-navigation","imitation-learning","action-prediction"],"year":"2023"},{"environments":["web"],"id":"arxiv:2305.14257","kind":"paper","record_sha256":"19c9ba1dc59513d49e7a1c2b27bdbc02a496ffe15979895776a28bf485fdc27d","scope":"core","source_url":"https://arxiv.org/abs/2305.14257","title":"Hierarchical Prompting Assists Large Language Model on Web Navigation","topics":["web-navigation","planning","efficiency"],"year":"2023"},{"environments":["web"],"id":"arxiv:2305.16653","kind":"paper","record_sha256":"407830e9f8f5790c8336ef18391d969d8ee5d24a68b4aaab88f29d161ab4c10a","scope":"adjacent","source_url":"https://arxiv.org/abs/2305.16653","title":"AdaPlanner: Adaptive Planning from Feedback with Language Models","topics":["planning","self-reflection","evaluation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2306.00245","kind":"paper","record_sha256":"2eccd714c167cb140e4e0b7b1424ef5fd40621a65d194d148806f699e38d8fe8","scope":"core","source_url":"https://arxiv.org/abs/2306.00245","title":"From Pixels to UI Actions: Learning to Follow Instructions via Graphical User Interfaces","topics":["grounding","action-prediction","web-navigation"],"year":"2023"},{"environments":["unspecified"],"id":"arxiv:2306.01242","kind":"paper","record_sha256":"f911296ef79fa1261f79cc0ffb4c8f67f7d6906eea30c6f21e858e73b357eac9","scope":"adjacent","source_url":"https://arxiv.org/abs/2306.01242","title":"Responsible Task Automation: Empowering Large Language Models as Responsible Task Automators","topics":["safety-security","evaluation","automation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2306.06070","kind":"paper","record_sha256":"43b14354201afd688aff283e54b1af2423e2617b086722b9f7fe88fe9aafe818","scope":"core","source_url":"https://arxiv.org/abs/2306.06070","title":"Mind2Web: Towards a Generalist Agent for the Web","topics":["data-generation","web-navigation","evaluation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2306.07863","kind":"paper","record_sha256":"30a8bc47209acf1e39eaf4219a733b6a4b367db3dda7fb8941737c9804d24957","scope":"core","source_url":"https://arxiv.org/abs/2306.07863","title":"Synapse: Trajectory-as-Exemplar Prompting with Memory for Computer Control","topics":["memory","web-navigation","planning","efficiency"],"year":"2023"},{"environments":["android"],"id":"arxiv:2307.10088","kind":"paper","record_sha256":"66db4203b60d301044b2c4f2806fa51eccf0d72df07f0fc05fc861982142a5c8","scope":"core","source_url":"https://arxiv.org/abs/2307.10088","title":"Android in the Wild: A Large-Scale Dataset for Android Device Control","topics":["data-generation","mobile-navigation","evaluation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2307.12856","kind":"paper","record_sha256":"d2a697261031996a1720f07f18b59336a99c1c0a046b263c68650e457da64bc2","scope":"core","source_url":"https://arxiv.org/abs/2307.12856","title":"A Real-World WebAgent with Planning, Long Context Understanding, and Program Synthesis","topics":["web-navigation","planning"],"year":"2023"},{"environments":["web"],"id":"arxiv:2307.13854","kind":"paper","record_sha256":"5862033a9c69b1d3c40b9a402162e4e24e5bae2dd5e434bc9838d62af71d02a6","scope":"core","source_url":"https://arxiv.org/abs/2307.13854","title":"WebArena: A Realistic Web Environment for Building Autonomous Agents","topics":["evaluation","web-navigation","long-horizon"],"year":"2023"},{"environments":["cross-platform"],"id":"arxiv:2308.03688","kind":"paper","record_sha256":"499212b502fdf6a886893b287d9c0e1da99f5b456bf6213fbb25e48705449122","scope":"adjacent","source_url":"https://arxiv.org/abs/2308.03688","title":"AgentBench: Evaluating LLMs as Agents","topics":["evaluation","planning"],"year":"2023"},{"environments":["android"],"id":"arxiv:2308.15272","kind":"paper","record_sha256":"aa27b20131ff745c2607f7688938f4be3fd52320f94ce01fd2d302e230e81f00","scope":"core","source_url":"https://arxiv.org/abs/2308.15272","title":"AutoDroid: LLM-powered Task Automation in Android","topics":["mobile-navigation","memory"],"year":"2023"},{"environments":["web"],"id":"arxiv:2309.08172","kind":"paper","record_sha256":"e01e4864f5b8c8b999d11d7f0b31af15f5d944c2670df207c67d6b992df887c4","scope":"core","source_url":"https://arxiv.org/abs/2309.08172","title":"LASER: LLM Agent with State-Space Exploration for Web Navigation","topics":["planning","robustness","web-navigation"],"year":"2023"},{"environments":["android"],"id":"arxiv:2309.11436","kind":"paper","record_sha256":"3fae95ddcda0faddd23732d70500e0556ffd7410affdb87932a39b27507de462","scope":"core","source_url":"https://arxiv.org/abs/2309.11436","title":"You Only Look at Screens: Multimodal Chain-of-Action Agents","topics":["mobile-navigation","planning"],"year":"2023"},{"environments":["web"],"id":"arxiv:2310.03720","kind":"paper","record_sha256":"bbd40ab88181ff085462df07b7ba85010f6f52ef23473e205ce8c026d399bcf8","scope":"core","source_url":"https://arxiv.org/abs/2310.03720","title":"SteP: Stacked LLM Policies for Web Actions","topics":["planning","web-navigation","workflow"],"year":"2023"},{"environments":["cross-platform"],"id":"arxiv:2310.04716","kind":"paper","record_sha256":"993c1bc96b90b70102582996ae41769fab221ba3e5344a65b7cce027eabf0517","scope":"core","source_url":"https://arxiv.org/abs/2310.04716","title":"Reinforced UI Instruction Grounding: Towards a Generic UI Task Automation API","topics":["grounding","reinforcement-learning"],"year":"2023"},{"environments":["unspecified"],"id":"arxiv:2310.04869","kind":"paper","record_sha256":"df591995ed0504f875381a5e32957afccbd531816677a2debfd0b8a5ed3cd28a","scope":"core","source_url":"https://arxiv.org/abs/2310.04869","title":"ILuvUI: Instruction-tuned LangUage-Vision modeling of UIs from Machine Conversations","topics":["data-generation","screen-understanding","grounding","planning"],"year":"2023"},{"environments":["web"],"id":"arxiv:2310.05126","kind":"paper","record_sha256":"3c83ad93378e92d171235a1f52b0d3520ff027d1c0dc1adedcb9b4e999cd162d","scope":"adjacent","source_url":"https://arxiv.org/abs/2310.05126","title":"UReader: Universal OCR-free Visually-situated Language Understanding with Multimodal Large Language Model","topics":["screen-understanding","efficiency"],"year":"2023"},{"environments":["web"],"id":"arxiv:2310.08740","kind":"paper","record_sha256":"fc77f466525ea3ee5dceb46c6f2300f8dcefb2b080c71f834f0adc026110c506","scope":"core","source_url":"https://arxiv.org/abs/2310.08740","title":"A Zero-Shot Language Agent for Computer Control with Structured Reflection","topics":["self-reflection","planning","web-navigation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2310.10634","kind":"paper","record_sha256":"3a552a6a14049838cfb0b5736cac2821bbd89b26f9b99c02f87ced96d6ee6966","scope":"adjacent","source_url":"https://arxiv.org/abs/2310.10634","title":"OpenAgents: An Open Platform for Language Agents in the Wild","topics":["web-navigation","tool-integration"],"year":"2023"},{"environments":["unspecified"],"id":"arxiv:2310.11441","kind":"paper","record_sha256":"5e274c197a4657f5e29cfc7a6e79265dffe9f672d2a6d812025befd2cb2873ec","scope":"adjacent","source_url":"https://arxiv.org/abs/2310.11441","title":"Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V","topics":["grounding"],"year":"2023"},{"environments":["unspecified"],"id":"arxiv:2310.14495","kind":"paper","record_sha256":"9ee351c19580b630b1ea3ec87496048f222b5a9e103d6f6b050dde7f28d16239","scope":"adjacent","source_url":"https://arxiv.org/abs/2310.14495","title":"InstructExcel: A Benchmark for Natural Language Instruction in Excel","topics":["evaluation","workflow","automation"],"year":"2023"},{"environments":["web"],"id":"arxiv:2310.16042","kind":"paper","record_sha256":"a012f0942caa07dce5913fb60b70193681def8dae854f96022cb0ea94f14e44b","scope":"core","source_url":"https://arxiv.org/abs/2310.16042","title":"WebWISE: Web Interface Control and Sequential Exploration with Large Language Models","topics":["web-navigation","tool-integration","planning"],"year":"2023"},{"environments":["web"],"id":"arxiv:2310.18331","kind":"paper","record_sha256":"7c187f30f9074ef02391801a739da9f9f8ee3ac49512eb244b1014377f0a3454","scope":"core","source_url":"https://arxiv.org/abs/2310.18331","title":"AllTogether: Investigating the Efficacy of Spliced Prompt for Web Navigation using Large Language Models","topics":["web-navigation","interaction-history","evaluation"],"year":"2023"},{"environments":["unspecified"],"id":"arxiv:2311.01767","kind":"paper","record_sha256":"4bdc5c6bb9ba54e7e2a4296f07e91f9abeb53be0cd82bd5e8916add0dcc89617","scope":"adjacent","source_url":"https://arxiv.org/abs/2311.01767","title":"PPTC Benchmark: Evaluating Large Language Models for PowerPoint Task Completion","topics":["evaluation","tool-integration"],"year":"2023"},{"environments":["ios","android"],"id":"arxiv:2311.07562","kind":"paper","record_sha256":"84071df1f1c16064a4bf08d0cf66e068ed43d69e259978252d4b05ba2ea112c2","scope":"core","source_url":"https://arxiv.org/abs/2311.07562","title":"GPT-4V in Wonderland: Large Multimodal Models for Zero-Shot Smartphone GUI Navigation","topics":["mobile-navigation","grounding"],"year":"2023"},{"environments":["web"],"id":"arxiv:2311.12983","kind":"paper","record_sha256":"c21ba8558b86ba3b6cd791f95cb9955ee04fd97ddea34ab744d133d510249cb0","scope":"adjacent","source_url":"https://arxiv.org/abs/2311.12983","title":"GAIA: a benchmark for General AI Assistants","topics":["evaluation","tool-integration"],"year":"2023"},{"environments":["web"],"id":"arxiv:2311.18751","kind":"paper","record_sha256":"2e60df85fc9eb2545302b448de5246c6c1585d7315a8167c8467567524ff3994","scope":"core","source_url":"https://arxiv.org/abs/2311.18751","title":"Exposing Limitations of Language Model Agents in Sequential-Task Compositions on the Web","topics":["evaluation","long-horizon","robustness","web-navigation"],"year":"2023"},{"environments":["unspecified"],"id":"arxiv:2312.03003","kind":"paper","record_sha256":"191d3f00a4bdb9a4f818ca6cc5f8501505441ce0602b2591300662de84008dfa","scope":"core","source_url":"https://arxiv.org/abs/2312.03003","title":"Explore, Select, Derive, and Recall: Augmenting LLM with Human-like Memory for Mobile Task Automation","topics":["mobile-navigation","memory","planning"],"year":"2023"},{"environments":["unspecified"],"id":"arxiv:2312.06677","kind":"paper","record_sha256":"bf55a772fdd9799c44dcb458ca5596b24b05a1c25e071a9563434428339388fe","scope":"core","source_url":"https://arxiv.org/abs/2312.06677","title":"Intelligent Virtual Assistants with LLM-based Process Automation","topics":["mobile-navigation","planning","automation"],"year":"2023"},{"environments":["desktop","android","web"],"id":"arxiv:2312.08914","kind":"paper","record_sha256":"2009e01fd8045f322b18e887cb3db989207661c52fd68d8ec66d8cac4412c5d5","scope":"core","source_url":"https://arxiv.org/abs/2312.08914","title":"CogAgent: A Visual Language Model for GUI Agents","topics":["screen-understanding","grounding"],"year":"2023"},{"environments":["android"],"id":"arxiv:2312.11190","kind":"paper","record_sha256":"911d82ce64fb74b41533ab6fec0697d9b191c562dc79c2d033cef8d27c664660","scope":"core","source_url":"https://arxiv.org/abs/2312.11190","title":"VisionTasker: Mobile Task Automation Using Vision Based UI Understanding and LLM Task Planning","topics":["screen-understanding","planning","mobile-navigation"],"year":"2023"},{"environments":["desktop","windows"],"id":"arxiv:2312.13108","kind":"paper","record_sha256":"3f57eecbe68c1c25bb0500c5dc91bd737aa50dd797999264c5666dbd1f64e301","scope":"core","source_url":"https://arxiv.org/abs/2312.13108","title":"ASSISTGUI: Task-Oriented Desktop Graphical User Interface Automation","topics":["evaluation","planning"],"year":"2023"},{"environments":["android"],"id":"arxiv:2312.13771","kind":"paper","record_sha256":"cca8cb306e9d32edcd55161340ee4c58affcb3d7b593c0fac43cd8e32e5cfeaf","scope":"core","source_url":"https://arxiv.org/abs/2312.13771","title":"AppAgent: Multimodal Agents as Smartphone Users","topics":["mobile-navigation","memory"],"year":"2023"},{"environments":["web"],"id":"arxiv:2312.15820","kind":"paper","record_sha256":"eae0381019e6aab06ca79f5d4c3372632d4e98426aa7e38728a3931ad6ce8c50","scope":"core","source_url":"https://arxiv.org/abs/2312.15820","title":"WebVLN: Vision-and-Language Navigation on Websites","topics":["web-navigation","data-generation","evaluation"],"year":"2023"},{"environments":["android"],"id":"arxiv:2312.16062","kind":"paper","record_sha256":"632d93f7db1d671d201e6b33029a5d8efd2e85233a9903ab7cc6750a2922da8d","scope":"core","source_url":"https://arxiv.org/abs/2312.16062","title":"AutoTask: Executing Arbitrary Voice Commands by Exploring and Learning from Mobile GUI","topics":["mobile-navigation","memory"],"year":"2023"},{"environments":["web"],"id":"arxiv:2401.01614","kind":"paper","record_sha256":"0dea08f3373757c9f5da63a74af40c5039e7b6db6e94e98cf5e07aeb9cb29d6c","scope":"core","source_url":"https://arxiv.org/abs/2401.01614","title":"GPT-4V(ision) is a Generalist Web Agent, if Grounded","topics":["web-navigation","grounding"],"year":"2024"},{"environments":["android"],"id":"arxiv:2401.04124","kind":"paper","record_sha256":"4f830cfbffe73362d3f9b57e9c9f3ae7728a2fcfff627f464b31dad8c8b5aec7","scope":"core","source_url":"https://arxiv.org/abs/2401.04124","title":"MobileAgent: enhancing mobile control via human-machine interaction and SOP integration","topics":["mobile-navigation","workflow","human-in-the-loop"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2401.10935","kind":"paper","record_sha256":"dadda1bc72de0d6dfab8ca0cb6e0f8571cffafe8a194febc91ca401f34a0af14","scope":"core","source_url":"https://arxiv.org/abs/2401.10935","title":"SeeClick: Harnessing GUI Grounding for Advanced Visual GUI Agents","topics":["grounding","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2401.13178","kind":"paper","record_sha256":"406812b3f9a2ed3f77a5cafe313728d129ebc4cf9977d87a6b60e1d60fa3c0fc","scope":"adjacent","source_url":"https://arxiv.org/abs/2401.13178","title":"AgentBoard: An Analytical Evaluation Board of Multi-turn LLM Agents","topics":["evaluation","observability","web-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2401.13649","kind":"paper","record_sha256":"ed4c9faca0c3850113fe1cc8c02f90f1fec1a08fb9612b907fa83e7d11959229","scope":"core","source_url":"https://arxiv.org/abs/2401.13649","title":"VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks","topics":["evaluation","grounding","web-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2401.13919","kind":"paper","record_sha256":"b6baf557abc0b70347d216dcf55d1c3dcbed213050fe136e69a0ac976a49acf6","scope":"core","source_url":"https://arxiv.org/abs/2401.13919","title":"WebVoyager: Building an End-to-End Web Agent with Large Multimodal Models","topics":["web-navigation","evaluation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2401.14268","kind":"paper","record_sha256":"c340609077b68aaf514d978dd631eb2207571ed22f386fb189860539d280d11a","scope":"core","source_url":"https://arxiv.org/abs/2401.14268","title":"GPTVoiceTasker: Advancing Multi-step Mobile Task Efficiency Through Dynamic Interface Exploration and Learning","topics":["mobile-navigation","human-in-the-loop","memory"],"year":"2024"},{"environments":["android"],"id":"arxiv:2401.16158","kind":"paper","record_sha256":"f2f23b755d18f15961c9839925e88cde3365a36a188f530c06a2022ec5a2e164","scope":"core","source_url":"https://arxiv.org/abs/2401.16158","title":"Mobile-Agent: Autonomous Multi-Modal Mobile Device Agent with Visual Perception","topics":["mobile-navigation","planning"],"year":"2024"},{"environments":["web"],"id":"arxiv:2402.04476","kind":"paper","record_sha256":"ccd7e71918df1ab361f574be6e684632868766dcb482b9b4a22091c690251474","scope":"core","source_url":"https://arxiv.org/abs/2402.04476","title":"Dual-View Visual Contextualization for Web Navigation","topics":["grounding","screen-understanding","web-navigation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2402.04615","kind":"paper","record_sha256":"77a222acd1422a5fd17d88dde864934a8df736159bc759a24d407c872356410c","scope":"core","source_url":"https://arxiv.org/abs/2402.04615","title":"ScreenAI: A Vision-Language Model for UI and Infographics Understanding","topics":["screen-understanding","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2402.05930","kind":"paper","record_sha256":"47c15df8c6537dd8bc2b26ecf2c8b6c9d7673d52da0b4caa9b57fb1fc441a37e","scope":"core","source_url":"https://arxiv.org/abs/2402.05930","title":"WebLINX: Real-World Website Navigation with Multi-Turn Dialogue","topics":["data-generation","web-navigation","evaluation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2402.06596","kind":"paper","record_sha256":"140f01e564f9fb36d18e12ad8ce2e00e9ede0a55f93514515dba4c9b2dae7cbd","scope":"core","source_url":"https://arxiv.org/abs/2402.06596","title":"Understanding the Weakness of Large Language Model Agents within a Complex Android Environment","topics":["evaluation","cross-app"],"year":"2024"},{"environments":["desktop","web"],"id":"arxiv:2402.07456","kind":"paper","record_sha256":"e84dc93d1a89c0568d7a03222b04a02872a2d40203ba029dff4aaf09b162e0f3","scope":"core","source_url":"https://arxiv.org/abs/2402.07456","title":"OS-Copilot: Towards Generalist Computer Agents with Self-Improvement","topics":["tool-integration","planning"],"year":"2024"},{"environments":["windows","desktop"],"id":"arxiv:2402.07939","kind":"paper","record_sha256":"142a14ea6d603d198d5a3c340ad9b509bada708fdbbcaedad404f652e17c6d17","scope":"core","source_url":"https://arxiv.org/abs/2402.07939","title":"UFO: A UI-Focused Agent for Windows OS Interaction","topics":["cross-app","planning","grounding","automation"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2402.07945","kind":"paper","record_sha256":"875d5b23d111b850a160ad774ad457edaa9b2070b3b95c28aabe5f1cae0f1132","scope":"core","source_url":"https://arxiv.org/abs/2402.07945","title":"ScreenAgent: A Vision Language Model-driven Computer Control Agent","topics":["automation","planning"],"year":"2024"},{"environments":["android"],"id":"arxiv:2402.11941","kind":"paper","record_sha256":"5461999d0922b24499dec63d11872e9392207fbda2f1b9ad032807bfc15b0b27","scope":"core","source_url":"https://arxiv.org/abs/2402.11941","title":"CoCo-Agent: A Comprehensive Cognitive MLLM Agent for Smartphone GUI Automation","topics":["mobile-navigation","screen-understanding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2402.14073","kind":"paper","record_sha256":"b92fcb75da1e903ed5b21e7c6a0d960c15708a6015afd5868fb49bfbbc87bd8b","scope":"adjacent","source_url":"https://arxiv.org/abs/2402.14073","title":"Improving Language Understanding from Screenshots","topics":["screen-understanding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2402.15057","kind":"paper","record_sha256":"36b77d2d4642e029852d18ee42f7cafff4e380b026ceabdf26571e704b2d7c81","scope":"core","source_url":"https://arxiv.org/abs/2402.15057","title":"On the Multi-turn Instruction Following for Conversational Web Agents","topics":["evaluation","memory","self-reflection","web-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2402.16965","kind":"paper","record_sha256":"0b52a04cd80ddb3e64e1d53149134d43ab05564806098f59ca65d4149ff84852","scope":"core","source_url":"https://arxiv.org/abs/2402.16965","title":"WIPI: A New Web Threat for LLM-Driven Web Agents","topics":["safety-security","web-navigation","evaluation"],"year":"2024"},{"environments":["desktop","web"],"id":"arxiv:2402.17553","kind":"paper","record_sha256":"86615372e856e51871ba89ec7bf4bdc8f333688895a16c25df03549f5915f185","scope":"core","source_url":"https://arxiv.org/abs/2402.17553","title":"OmniACT: A Dataset and Benchmark for Enabling Multimodal Generalist Autonomous Agents for Desktop and Web","topics":["evaluation","grounding"],"year":"2024"},{"environments":["android"],"id":"arxiv:2403.02713","kind":"paper","record_sha256":"3eb453772aaad5aeae81d46263a4b86ceefcf009731fb0a60e6bd5dab7b8b008","scope":"core","source_url":"https://arxiv.org/abs/2403.02713","title":"Android in the Zoo: Chain-of-Action-Thought for GUI Agents","topics":["mobile-navigation","planning"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2403.03186","kind":"paper","record_sha256":"c4135abf01a07b4ec0644ff2cd92fd41dcce326d2cfb76a487059d1c5bdee0ee","scope":"core","source_url":"https://arxiv.org/abs/2403.03186","title":"Cradle: Empowering Foundation Agents Towards General Computer Control","topics":["planning","screen-understanding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2403.03788","kind":"paper","record_sha256":"faf54e92547372c937f425ca4f7f6ed084a97c3aaec0c03f3961f4d6899d3a03","scope":"adjacent","source_url":"https://arxiv.org/abs/2403.03788","title":"PPTC-R benchmark: Towards Evaluating the Robustness of Large Language Models for PowerPoint Task Completion","topics":["evaluation","robustness","workflow"],"year":"2024"},{"environments":["android","ios"],"id":"arxiv:2403.04984","kind":"paper","record_sha256":"df6d7a3ccde3955ea50fc7e5585774a1796279da00cf619c851d01023ca778df","scope":"core","source_url":"https://arxiv.org/abs/2403.04984","title":"UI Semantic Group Detection: Grouping UI Elements with Similar Semantics in Mobile Graphical User Interface","topics":["screen-understanding","grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2403.07718","kind":"paper","record_sha256":"a9ef9ee103623f137ca698722e7eb28dd2c4ecd79a6596249a235cd1c9cabd69","scope":"core","source_url":"https://arxiv.org/abs/2403.07718","title":"WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?","topics":["evaluation","workflow"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2403.10170","kind":"paper","record_sha256":"b3398bc7607b29bc3596943597f728a2180cdf7760b56c8ffbaf9872e5a29857","scope":"core","source_url":"https://arxiv.org/abs/2403.10170","title":"Computer User Interface Understanding. A New Dataset and a Learning Framework","topics":["screen-understanding","video","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2403.11905","kind":"paper","record_sha256":"06b1e6d8c02ae5d5c966f1ed196f8ae793256f8bb848ce425a598143c312278e","scope":"core","source_url":"https://arxiv.org/abs/2403.11905","title":"Tur[k]ingBench: A Challenge Benchmark for Web Agents","topics":["evaluation","web-navigation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2403.17918","kind":"paper","record_sha256":"3b7b30ed77cf6ee617443c532b85d24f9da3c69a67ea14edb1816a141e1ce2a0","scope":"core","source_url":"https://arxiv.org/abs/2403.17918","title":"AgentStudio: A Toolkit for Building General Virtual Agents","topics":["evaluation","sandbox","grounding","video"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2403.20329","kind":"paper","record_sha256":"2ce513de73fd7a80beade6e6b976a36919bb7d3fbb01aeb952ed0863f0385cae","scope":"adjacent","source_url":"https://arxiv.org/abs/2403.20329","title":"ReALM: Reference Resolution As Language Modeling","topics":["screen-understanding","grounding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2404.02475","kind":"paper","record_sha256":"335815dfc9ee3c171ea8a4079895266d7cad29fd877518badd20f0b3a0c36227","scope":"core","source_url":"https://arxiv.org/abs/2404.02475","title":"Prompt2Task: Automating UI Tasks on Smartphones from Textual Prompts","topics":["mobile-navigation","human-in-the-loop","workflow"],"year":"2024"},{"environments":["web"],"id":"arxiv:2404.03648","kind":"paper","record_sha256":"1f557c956db7d53c2ae44f36a162115d7f8cc2cd8afb10a817df7c0372bc4cec","scope":"core","source_url":"https://arxiv.org/abs/2404.03648","title":"AutoWebGLM: A Large Language Model-based Web Navigating Agent","topics":["web-navigation","reinforcement-learning","data-generation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2404.05719","kind":"paper","record_sha256":"73409a8516a653efe7a93f6a2205ff687ea734d9f588260c39b3adaabe0a2b52","scope":"core","source_url":"https://arxiv.org/abs/2404.05719","title":"Ferret-UI: Grounded Mobile UI Understanding with Multimodal LLMs","topics":["screen-understanding","grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2404.05902","kind":"paper","record_sha256":"2779ff006dc02211c4785a378e6dd58bfd0c0c4a4cc8c58a46c89dd05b3cb7dd","scope":"core","source_url":"https://arxiv.org/abs/2404.05902","title":"WILBUR: Adaptive In-Context Learning for Robust and Accurate Web Agents","topics":["web-navigation","memory","robustness","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2404.05955","kind":"paper","record_sha256":"94c89b9417a97b807fb22518f09ab33591aea5c0f0ccdaf752a9a6a41099cd4e","scope":"core","source_url":"https://arxiv.org/abs/2404.05955","title":"VisualWebBench: How Far Have Multimodal LLMs Evolved in Web Page Understanding and Grounding?","topics":["evaluation","screen-understanding","grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2404.06474","kind":"paper","record_sha256":"c4d03316767bdc8138969779410408c2815e09a292ece9278e7f4c40059bbe0b","scope":"core","source_url":"https://arxiv.org/abs/2404.06474","title":"Autonomous Evaluation and Refinement of Digital Agents","topics":["evaluation","reward-models"],"year":"2024"},{"environments":["desktop","linux","windows","macos"],"id":"arxiv:2404.07972","kind":"paper","record_sha256":"12d7508c604d863520bd2f7681aa00c3561ac4d9b286b563c2f7bb1aeeeada36","scope":"core","source_url":"https://arxiv.org/abs/2404.07972","title":"OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments","topics":["evaluation","cross-app"],"year":"2024"},{"environments":["android"],"id":"arxiv:2404.08755","kind":"paper","record_sha256":"97ce5ffd5e450e55198985c854965e192d9f2d7cf0442afbf8b5eabda82a7b60","scope":"core","source_url":"https://arxiv.org/abs/2404.08755","title":"Training a Vision Language Model as Smartphone Assistant","topics":["mobile-navigation","interaction-history","action-prediction"],"year":"2024"},{"environments":["android"],"id":"arxiv:2404.08860","kind":"paper","record_sha256":"5b606ad9dba97dbfb90e668798d485af0d3c32a8df9b19c2d608ff0e15a46112","scope":"core","source_url":"https://arxiv.org/abs/2404.08860","title":"Enhancing Mobile \"How-to\" Queries with Automated Search Results Verification and Reranking","topics":["mobile-navigation","evaluation","automation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2404.08948","kind":"paper","record_sha256":"e8ab436bb3c5846c12c6764d2e0a59c67f551b11c6d14414581bb7608c020514","scope":"core","source_url":"https://arxiv.org/abs/2404.08948","title":"Large Language Models for Mobile GUI Text Input Generation: An Empirical Study","topics":["automation","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2404.09992","kind":"paper","record_sha256":"adfe6dfb4bf1cceaa7451c50993cab54e4e72e8aa2500669915ba13cdcde0745","scope":"core","source_url":"https://arxiv.org/abs/2404.09992","title":"MMInA: Benchmarking Multihop Multimodal Internet Agents","topics":["evaluation","long-horizon","memory"],"year":"2024"},{"environments":["web"],"id":"arxiv:2404.10887","kind":"paper","record_sha256":"59d605be46d96c1cfea84e75d15702b29a9a809fbb30e641f39ad7dabe7a68d3","scope":"core","source_url":"https://arxiv.org/abs/2404.10887","title":"Grounded Language Agent for Product Search via Intelligent Web Interactions","topics":["web-navigation","reinforcement-learning","imitation-learning"],"year":"2024"},{"environments":["web"],"id":"arxiv:2404.16048","kind":"paper","record_sha256":"28efbfd412d41e3290f4048f9950ff48abe0ddc75008aad77863014ef357b15c","scope":"core","source_url":"https://arxiv.org/abs/2404.16048","title":"GUIDE: Graphical User Interface Data for Execution","topics":["data-generation","grounding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2404.16054","kind":"paper","record_sha256":"5b2797bc9371d8be404beb6626fa9d183673f6ef93722bb8e7ea05e1c7ae21fa","scope":"core","source_url":"https://arxiv.org/abs/2404.16054","title":"LlamaTouch: A Faithful and Scalable Testbed for Mobile UI Task Automation","topics":["evaluation","mobile-navigation","state-tracking"],"year":"2024"},{"environments":["android"],"id":"arxiv:2404.16660","kind":"paper","record_sha256":"95f894a4d56b53c35a17a43bd369df177f9ec57576084d6ef47cb15f120484aa","scope":"core","source_url":"https://arxiv.org/abs/2404.16660","title":"Benchmarking Mobile Device Control Agents across Diverse Configurations","topics":["evaluation","robustness"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2404.18074","kind":"paper","record_sha256":"75f0192c480e3c6ea49f61577ac7c32458feafbf765ecc852b48f4a1729d4922","scope":"core","source_url":"https://arxiv.org/abs/2404.18074","title":"MMAC-Copilot: Multi-modal Agent Collaboration Operating Copilot","topics":["automation","evaluation","planning"],"year":"2024"},{"environments":["web"],"id":"arxiv:2405.00516","kind":"paper","record_sha256":"5b23d7b8b6e1b4353e5bdf34f9187a0ffd172e0bbc45eca75f37855add6331f2","scope":"core","source_url":"https://arxiv.org/abs/2405.00516","title":"Navigating WebAI: Training Agents to Complete Web Tasks with Large Language Models and Reinforcement Learning","topics":["web-navigation","reinforcement-learning","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2405.04497","kind":"paper","record_sha256":"63b18dd33e8a3ceb98e7c8089482ea64a022c7457c591815587684222968ed20","scope":"core","source_url":"https://arxiv.org/abs/2405.04497","title":"Unveiling Disparities in Web Task Handling Between Human and Web Agent","topics":["evaluation","human-in-the-loop","planning"],"year":"2024"},{"environments":["android"],"id":"arxiv:2405.07090","kind":"paper","record_sha256":"85cc216db741141e1ec182ed01f973014e25b479c583cce89a95aa467edbab57","scope":"core","source_url":"https://arxiv.org/abs/2405.07090","title":"MUD: Towards a Large-Scale and Noise-Filtered UI Dataset for Modern Style UI Modeling","topics":["data-generation","screen-understanding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2405.11120","kind":"paper","record_sha256":"e39bfc4be3adfb5b869e48d4e5aa16f13f49509d796010ca9b784c6b8070752e","scope":"core","source_url":"https://arxiv.org/abs/2405.11120","title":"Latent State Estimation Helps UI Agents to Reason","topics":["state-tracking","planning","evaluation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2405.12842","kind":"paper","record_sha256":"6e53475c326e717a51ad3a5fef9c10c49ffcb9ea1f4ebf007635682246c613c1","scope":"core","source_url":"https://arxiv.org/abs/2405.12842","title":"SmartFlow: Robotic Process Automation using LLMs","topics":["automation","workflow","screen-understanding"],"year":"2024"},{"environments":["android"],"id":"arxiv:2405.14573","kind":"paper","record_sha256":"6ba321909c19105bbb3c84a186a75a22e560d557451c30cac092abe4e3382ebc","scope":"core","source_url":"https://arxiv.org/abs/2405.14573","title":"AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents","topics":["evaluation","mobile-navigation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2405.15341","kind":"paper","record_sha256":"d83d967672846a37a13734a472bf0316b4dd03fecfd74ae317b5796cc242c99f","scope":"core","source_url":"https://arxiv.org/abs/2405.15341","title":"V-Zen: Efficient GUI Understanding and Precise Grounding With A Novel Multimodal LLM","topics":["grounding","action-prediction","screen-understanding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2405.20309","kind":"paper","record_sha256":"ded4bf1dbd18f389fe1a5988f94d42055a5c2284b79d4b6f4d200ff2f6bb6509","scope":"core","source_url":"https://arxiv.org/abs/2405.20309","title":"Large Language Models Can Self-Improve At Web Agent Tasks","topics":["data-generation","web-navigation","evaluation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2406.01014","kind":"paper","record_sha256":"cde495fdc311b839b5fd5675197bebdf91bdfb8a5d7ec79c122fbb70cda97c1f","scope":"core","source_url":"https://arxiv.org/abs/2406.01014","title":"Mobile-Agent-v2: Mobile Device Operation Assistant with Effective Navigation via Multi-Agent Collaboration","topics":["planning","memory","mobile-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2406.01623","kind":"paper","record_sha256":"8f55bd14827ee05f15451eeb4d54dd4d9f453c4fe837b576ad4471606917ecf9","scope":"core","source_url":"https://arxiv.org/abs/2406.01623","title":"WebSuite: Systematically Evaluating Why Web Agents Fail","topics":["evaluation","observability","web-navigation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2406.03679","kind":"paper","record_sha256":"1aa23496314c084e9725f24f70858c5bcc0a0ac38f0b02aa5c27d6da524d6c3e","scope":"core","source_url":"https://arxiv.org/abs/2406.03679","title":"On the Effects of Data Scale on UI Control Agents","topics":["data-generation","mobile-navigation","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2406.06947","kind":"paper","record_sha256":"81e29a3b395e458f5d236f58aab8b78d944015604831e3ee1a974fbcd0f2e282","scope":"core","source_url":"https://arxiv.org/abs/2406.06947","title":"CAAP: Context-Aware Action Planning Prompting to Solve Computer Tasks with Front-End UI Only","topics":["planning","screen-understanding","web-navigation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2406.07822","kind":"paper","record_sha256":"ae93bf1c40aa90494d370cf47b29c813a666ded8a86b50343be55db850b58845","scope":"core","source_url":"https://arxiv.org/abs/2406.07822","title":"Tell Me What's Next: Textual Foresight for Generic UI Representations","topics":["screen-understanding","world-models","data-generation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2406.08184","kind":"paper","record_sha256":"9c4554631fa51ee99cb113c279e16a1d9baff8906bcd07c6366ce52aefee71d9","scope":"core","source_url":"https://arxiv.org/abs/2406.08184","title":"MobileAgentBench: An Efficient and User-Friendly Benchmark for Mobile LLM Agents","topics":["evaluation","mobile-navigation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2406.08340","kind":"paper","record_sha256":"34d03c2243e1cddea9982b1aeb4a9cab37100323980cf428bbc7ba2f20290bf6","scope":"core","source_url":"https://arxiv.org/abs/2406.08340","title":"Practical, Automated Scenario-based Mobile App Testing","topics":["automation","cross-app","screen-understanding","evaluation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2406.08451","kind":"paper","record_sha256":"48c88d6b05a4665bedd53a1f208b98790389601a4ddd87d4fbab69fae18bb493","scope":"core","source_url":"https://arxiv.org/abs/2406.08451","title":"GUIOdyssey: A Comprehensive Dataset for Cross-App GUI Navigation on Mobile Devices","topics":["cross-app","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2406.09187","kind":"paper","record_sha256":"cef1230fcac8463f88ca918b93b01f463ff7a5c68722583efe67aacd1cb11b33","scope":"adjacent","source_url":"https://arxiv.org/abs/2406.09187","title":"GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoning","topics":["safety-security","evaluation","tool-integration"],"year":"2024"},{"environments":["desktop","web"],"id":"arxiv:2406.10227","kind":"paper","record_sha256":"f7b5c55082df7ee5542f378c6fa42b7c382322bf38dd285695994d89dad74608","scope":"core","source_url":"https://arxiv.org/abs/2406.10227","title":"VideoGUI: A Benchmark for GUI Automation from Instructional Videos","topics":["evaluation","video","planning"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2406.10819","kind":"paper","record_sha256":"f5eac3e7dfb44279a6a441e6a74be7754da5e66772dd7536a2cb646ca9a336d6","scope":"core","source_url":"https://arxiv.org/abs/2406.10819","title":"GUI-World: A Video Benchmark and Dataset for Multimodal GUI-oriented Understanding","topics":["evaluation","video","screen-understanding"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2406.11317","kind":"paper","record_sha256":"235c93f399bdf30451041e4294a85342f1d380c1d7c216a11058a25fd39d4102","scope":"core","source_url":"https://arxiv.org/abs/2406.11317","title":"GUICourse: From General Vision Language Models to Versatile GUI Agents","topics":["grounding","screen-understanding"],"year":"2024"},{"environments":["android"],"id":"arxiv:2406.11896","kind":"paper","record_sha256":"bf358cc4588b92d0cc8582c9b2a1fcc7f1869083da1ad882dd5cf8a24cbcb0af","scope":"core","source_url":"https://arxiv.org/abs/2406.11896","title":"DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning","topics":["reinforcement-learning","mobile-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2406.12373","kind":"paper","record_sha256":"d2b80d2079fba88f46d5604e92ebbf2b9b50ea5172cd56fd7e44f7aa72575806","scope":"core","source_url":"https://arxiv.org/abs/2406.12373","title":"WebCanvas: Benchmarking Web Agents in Online Environments","topics":["evaluation","web-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2406.12814","kind":"paper","record_sha256":"8435a688d793d3f873830f97713182b6529132d08380edf05bfec5bebb21cac4","scope":"core","source_url":"https://arxiv.org/abs/2406.12814","title":"Dissecting Adversarial Robustness of Multimodal LM Agents","topics":["safety-security","robustness","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2406.13264","kind":"paper","record_sha256":"db03d09d8e060f3c4f26195fac19c176a61a528102c51716f6e48e713a3a77d4","scope":"core","source_url":"https://arxiv.org/abs/2406.13264","title":"WONDERBREAD: A Benchmark for Evaluating Multimodal Foundation Models on Business Process Management Tasks","topics":["evaluation","workflow","video"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2406.13719","kind":"paper","record_sha256":"3516d6ab3cb6397e341b83d58493fb3236bb21d80dddf16cc916aeb2ba4fae48","scope":"core","source_url":"https://arxiv.org/abs/2406.13719","title":"GUI Action Narrator: Where and When Did That Action Take Place?","topics":["video","screen-understanding"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2406.14056","kind":"paper","record_sha256":"6f39fd128da6dc28f0b27111d78d73ab6e9b854b3e55383726da6836e4d21420","scope":"core","source_url":"https://arxiv.org/abs/2406.14056","title":"VGA: Vision GUI Assistant -- Minimizing Hallucinations through Image-Centric Fine-Tuning","topics":["screen-understanding","grounding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2406.14250","kind":"paper","record_sha256":"b2849614c57db9b916b667b78df69a622415f6a827400457b094ac5eacd90fe0","scope":"core","source_url":"https://arxiv.org/abs/2406.14250","title":"E-ANT: A Large-Scale Dataset for Efficient Automatic GUI NavigaTion","topics":["mobile-navigation","evaluation"],"year":"2024"},{"environments":["android","web","cross-platform"],"id":"arxiv:2406.14314","kind":"paper","record_sha256":"f5ac421ac57b335bde39428c5ec20af99d3aa139ed97b50d7e1aad1552fab25d","scope":"core","source_url":"https://arxiv.org/abs/2406.14314","title":"Identifying User Goals from UI Trajectories","topics":["evaluation","interaction-history","screen-understanding"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2406.19263","kind":"paper","record_sha256":"123bea303eb3b13601e548ff5e78fa499e3cc6b14c594b24f59f88db98991b41","scope":"core","source_url":"https://arxiv.org/abs/2406.19263","title":"Read Anywhere Pointed: Layout-aware GUI Screen Reading with Tree-of-Lens Grounding","topics":["screen-understanding","grounding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.00993","kind":"paper","record_sha256":"d18f4d19ca38fb3bc56522a869d5328f781677d7a3965337b8dc14b2334768dd","scope":"core","source_url":"https://arxiv.org/abs/2407.00993","title":"Mobile-Bench: An Evaluation Benchmark for LLM-based Mobile Agents","topics":["evaluation","cross-app","tool-integration"],"year":"2024"},{"environments":["web"],"id":"arxiv:2407.01476","kind":"paper","record_sha256":"3227fc889f15e9d2843cb55ab7787d4fb2be86295ce689a61b3b29140c36e313","scope":"core","source_url":"https://arxiv.org/abs/2407.01476","title":"Tree Search for Language Model Agents","topics":["planning","web-navigation","evaluation"],"year":"2024"},{"environments":["desktop","cross-platform"],"id":"arxiv:2407.01511","kind":"paper","record_sha256":"bdc3477e9d4967fec5b31ae5f885d3602d5653b78c38db9208ccc18858628e1a","scope":"core","source_url":"https://arxiv.org/abs/2407.01511","title":"CRAB: Cross-environment Agent Benchmark for Multimodal Language Model Agents","topics":["evaluation","cross-app"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2407.01558","kind":"paper","record_sha256":"8d59619b57d6c76c66e2e100dfee2c34a03177db4f1903785d001ef62dc803c3","scope":"core","source_url":"https://arxiv.org/abs/2407.01558","title":"Visual Grounding Methods for Efficient Interaction with Desktop Graphical User Interfaces","topics":["grounding","evaluation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.03037","kind":"paper","record_sha256":"5d3869a41fe177cb0eb43ec9a0937aa43385cf05d6bff73906e05632345816ed","scope":"core","source_url":"https://arxiv.org/abs/2407.03037","title":"Seeing is Believing: Vision-driven Non-crash Functional Bug Detection for Mobile Apps","topics":["automation","evaluation","screen-understanding","state-tracking"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.03913","kind":"paper","record_sha256":"4fa00ca134ec115041aaceeb6e509003782845cc7e2254f87609c126dc54851f","scope":"core","source_url":"https://arxiv.org/abs/2407.03913","title":"MobileExperts: A Dynamic Tool-Enabled Agent Team in Mobile Devices","topics":["mobile-navigation","planning","tool-integration"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.04346","kind":"paper","record_sha256":"d081fe11fcb11a459e2b0a3ccf3adf88b0bc7e185d8ef0a2fdfc5113e8f48cd3","scope":"core","source_url":"https://arxiv.org/abs/2407.04346","title":"MobileFlow: A Multimodal LLM For Mobile GUI Agent","topics":["mobile-navigation","screen-understanding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2407.05291","kind":"paper","record_sha256":"304ed495adbbcce7e4c290947447371c562e67e940529998edcfcfb40cda3ae5","scope":"core","source_url":"https://arxiv.org/abs/2407.05291","title":"WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks","topics":["evaluation","planning","workflow"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.08850","kind":"paper","record_sha256":"0e91e8ae96bcd98ad69cf08b3b21c75bf56e84e9260f015182e7615b43ff392f","scope":"adjacent","source_url":"https://arxiv.org/abs/2407.08850","title":"UICrit: Enhancing Automated Design Evaluation with a UICritique Dataset","topics":["screen-understanding","evaluation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.09018","kind":"paper","record_sha256":"7a727c76703fff83f2e148dfa105f476bd0834b199ef096112b01ecab92f5241","scope":"core","source_url":"https://arxiv.org/abs/2407.09018","title":"AUITestAgent: Automatic Requirements Oriented GUI Function Testing","topics":["automation","evaluation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.09295","kind":"paper","record_sha256":"9c025651f292ad69efaee299ee8b55a2bcc283de5768fdac150d71fb2eb6b8ff","scope":"core","source_url":"https://arxiv.org/abs/2407.09295","title":"Systematic Categorization, Construction and Evaluation of New Attacks against Multi-modal Mobile GUI Agents","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["desktop","web"],"id":"arxiv:2407.10956","kind":"paper","record_sha256":"0a3aec0ae9ff8da0b580158fb8b348a47bad85cef36cd3fa835d7624e99d286a","scope":"core","source_url":"https://arxiv.org/abs/2407.10956","title":"Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?","topics":["evaluation","science","workflow"],"year":"2024"},{"environments":["web"],"id":"arxiv:2407.13032","kind":"paper","record_sha256":"ad5f870525949724c4b408ae823a588d51d07721d69176140873bb74c4c0769b","scope":"core","source_url":"https://arxiv.org/abs/2407.13032","title":"Agent-E: From Autonomous Web Navigation to Foundational Design Principles in Agentic Systems","topics":["web-navigation","planning","efficiency","state-tracking"],"year":"2024"},{"environments":["web"],"id":"arxiv:2407.15673","kind":"paper","record_sha256":"568f683760a55720aa9e7aecb645f6f8cbd9601df5139ca3ffc176929c44b5b8","scope":"core","source_url":"https://arxiv.org/abs/2407.15673","title":"IDA: Breaking Barriers in No-code UI Automation Through Large Language Models and Human-Centric Design","topics":["human-in-the-loop","automation","workflow","industry"],"year":"2024"},{"environments":["web"],"id":"arxiv:2407.15711","kind":"paper","record_sha256":"9b77d37d05f8150ce5bcc26f1c13b9a90f75c6d93a656a13970aad172c4a8860","scope":"core","source_url":"https://arxiv.org/abs/2407.15711","title":"AssistantBench: Can Web Agents Solve Realistic and Time-Consuming Tasks?","topics":["evaluation","web-navigation","planning"],"year":"2024"},{"environments":["android"],"id":"arxiv:2407.17490","kind":"paper","record_sha256":"976fff8000c47ab673fb6db718abe9b8f557b7d82d02725812ab91dc9fbc5103","scope":"core","source_url":"https://arxiv.org/abs/2407.17490","title":"AMEX: Android Multi-annotation Expo Dataset for Mobile GUI Agents","topics":["grounding","screen-understanding","mobile-navigation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2407.18901","kind":"paper","record_sha256":"c5042f846c9ef8c12a74bc79ad9eecf4473d3652a91bede5c1e792ba0ca395ca","scope":"adjacent","source_url":"https://arxiv.org/abs/2407.18901","title":"AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents","topics":["evaluation","tool-integration","cross-app"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2407.19056","kind":"paper","record_sha256":"0b4c0bf2d4e6c4e1bf0abd45dba882fe866fc80cd730a5f5347e634372c289f3","scope":"core","source_url":"https://arxiv.org/abs/2407.19056","title":"OfficeBench: Benchmarking Language Agents across Multiple Applications for Office Automation","topics":["evaluation","cross-app","long-horizon"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2408.00203","kind":"paper","record_sha256":"64aa9cba28aeb6f42dbb03e144c8a56ea243e336bab88da9079b4d5fdfd0a918","scope":"core","source_url":"https://arxiv.org/abs/2408.00203","title":"OmniParser for Pure Vision Based GUI Agent","topics":["grounding","screen-understanding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2408.02544","kind":"paper","record_sha256":"69252f04ce71ded8db04848516ec641e3ba5e629aa93f2de46a41514e494e1bc","scope":"core","source_url":"https://arxiv.org/abs/2408.02544","title":"Caution for the Environment: Multimodal LLM Agents are Susceptible to Environmental Distractions","topics":["safety-security","robustness","evaluation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2408.06327","kind":"paper","record_sha256":"fc058b923fe3eb7575b45594c3c43effdd1f94e21b7fe146d1cf2397ce9e8837","scope":"adjacent","source_url":"https://arxiv.org/abs/2408.06327","title":"VisualAgentBench: Towards Large Multimodal Models as Visual Foundation Agents","topics":["evaluation","data-generation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2408.11824","kind":"paper","record_sha256":"d82db1d9c502fafc6896ee2b91a293f7a5a3614b4073dab4818cd9c27bc73b85","scope":"core","source_url":"https://arxiv.org/abs/2408.11824","title":"AppAgent v2: Advanced Agent for Flexible Mobile Interactions","topics":["mobile-navigation","memory"],"year":"2024"},{"environments":["web"],"id":"arxiv:2408.15978","kind":"paper","record_sha256":"80fba9f63cab578fe9344e651230cabf29ede07eace7cc57f14f000cf45a11cf","scope":"core","source_url":"https://arxiv.org/abs/2408.15978","title":"WebPilot: A Versatile and Autonomous Multi-Agent System for Web Task Execution with Strategic Exploration","topics":["planning","web-navigation","long-horizon"],"year":"2024"},{"environments":["web"],"id":"arxiv:2409.01927","kind":"paper","record_sha256":"ad2370e44e0aedce4990057d376daed368abab9dc11248f7de04781497e1e849","scope":"core","source_url":"https://arxiv.org/abs/2409.01927","title":"From Grounding to Planning: Benchmarking Bottlenecks in Web Agents","topics":["evaluation","grounding","planning"],"year":"2024"},{"environments":["web"],"id":"arxiv:2409.07429","kind":"paper","record_sha256":"dcc49da571ad4fd512ddba6bad77d62abb96c45911e301e29a37fd7e6ec5613f","scope":"core","source_url":"https://arxiv.org/abs/2409.07429","title":"Agent Workflow Memory","topics":["memory","workflow","web-navigation"],"year":"2024"},{"environments":["desktop","windows"],"id":"arxiv:2409.08264","kind":"paper","record_sha256":"aa1e1adf1dbabcb0603c3389c7d98eb60d67d4f23f6322833df02522692e7530","scope":"core","source_url":"https://arxiv.org/abs/2409.08264","title":"Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale","topics":["evaluation","cross-app"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2409.09354","kind":"paper","record_sha256":"891b11724575828f507f11238754d7ac2c9acb9300637bce11e42c2a27415ea3","scope":"core","source_url":"https://arxiv.org/abs/2409.09354","title":"PeriGuru: A Peripheral Robotic Mobile App Operation Assistant based on GUI Image Understanding and Prompting with LLM","topics":["mobile-navigation","screen-understanding","automation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2409.10741","kind":"paper","record_sha256":"cc41475e1b8708af1e220c71526694b80e452c866bb9e703c1cae0c99d213878","scope":"core","source_url":"https://arxiv.org/abs/2409.10741","title":"NaviQAte: Functionality-Guided Web Application Navigation","topics":["web-navigation","automation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2409.10811","kind":"paper","record_sha256":"42f303be4aafb81914577a044358636ee33df43bfc86b47f25b723e39fe2b147","scope":"core","source_url":"https://arxiv.org/abs/2409.10811","title":"Grounded GUI Understanding for Vision-Based Spatial Intelligent Agent: Exemplified by Extended Reality Apps","topics":["screen-understanding","grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2409.11295","kind":"paper","record_sha256":"43dc45786af29cfb507009dc1d3c85aba384916ce1db930526753877f5d0b3d5","scope":"core","source_url":"https://arxiv.org/abs/2409.11295","title":"EIA: Environmental Injection Attack on Generalist Web Agents for Privacy Leakage","topics":["safety-security","web-navigation","evaluation"],"year":"2024"},{"environments":["web","desktop"],"id":"arxiv:2409.12089","kind":"paper","record_sha256":"30dd2c60eeb806b73276dc7ec5c0cd3a8402be53bf706bac810e942df750c9f1","scope":"core","source_url":"https://arxiv.org/abs/2409.12089","title":"The Impact of Element Ordering on LM Agent Performance","topics":["screen-understanding","grounding","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2409.13711","kind":"paper","record_sha256":"bb6d34b3dd156190fa2f0e8cfaf1b9814c1cd8c06e8fd881e8dced45e804da42","scope":"core","source_url":"https://arxiv.org/abs/2409.13711","title":"WebQuest: A Benchmark for Multimodal QA on Web Page Sequences","topics":["evaluation","screen-understanding"],"year":"2024"},{"environments":["android"],"id":"arxiv:2409.14337","kind":"paper","record_sha256":"5547b423f8fdb99840470a05fc50005226289d0c04703dc929e50ae1440d037f","scope":"core","source_url":"https://arxiv.org/abs/2409.14337","title":"MobileViews: A Million-scale and Diverse Mobile GUI Dataset","topics":["screen-understanding","data-generation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2409.14818","kind":"paper","record_sha256":"18608dc186407dc8486b49ed089e0cc81b5890a15281df895241f9661462fd61","scope":"core","source_url":"https://arxiv.org/abs/2409.14818","title":"MobileVLM: A Vision-Language Model for Better Intra- and Inter-UI Understanding","topics":["screen-understanding","data-generation","action-prediction"],"year":"2024"},{"environments":["web"],"id":"arxiv:2409.15441","kind":"paper","record_sha256":"2c4b165d8a76e1463fcb041b5a5d1cb6bb8e4b72c64e8082e314d1b3d15fef17","scope":"core","source_url":"https://arxiv.org/abs/2409.15441","title":"Steward: Natural Language Web Automation","topics":["web-navigation","automation","efficiency"],"year":"2024"},{"environments":["web"],"id":"arxiv:2409.15637","kind":"paper","record_sha256":"ddebc42ea938156e0f2094dfc5849efca1c9b825f86c1610c311315c649acfe2","scope":"core","source_url":"https://arxiv.org/abs/2409.15637","title":"Synatra: Turning Indirect Knowledge into Direct Demonstrations for Digital Agents at Scale","topics":["data-generation","imitation-learning","web-navigation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2409.17140","kind":"paper","record_sha256":"9704f7565cb9e4c9749809ec0f630c8734bea3f90d9c1b20e6e0b40740b38b3e","scope":"core","source_url":"https://arxiv.org/abs/2409.17140","title":"AXIS: Efficient Human-Agent-Computer Interaction with API-First LLM-Based Agents","topics":["tool-integration","automation","efficiency"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2409.18060","kind":"paper","record_sha256":"e6499e3e82a7301ca6d0eee011c45e2ae9ef1d14cfcb029f650061d1f9299013","scope":"adjacent","source_url":"https://arxiv.org/abs/2409.18060","title":"Inferring Alt-text For UI Icons With Large Language Models During App Development","topics":["screen-understanding","ui-design","evaluation"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2410.00079","kind":"paper","record_sha256":"2ae6acb7f71320829bf141d472f7b3ebeef198698507673a6532023499bb1036","scope":"adjacent","source_url":"https://arxiv.org/abs/2410.00079","title":"Interactive Speculative Planning: Enhance Agent Efficiency through Co-design of System and User Interface","topics":["planning","efficiency","human-in-the-loop"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2410.00201","kind":"paper","record_sha256":"6440538471d649132f80b4a32f33cd746e35b0b2c8667c88869105ffc7523a1c","scope":"core","source_url":"https://arxiv.org/abs/2410.00201","title":"DreamStruct: Understanding Slides and User Interfaces via Synthetic Data Generation","topics":["data-generation","screen-understanding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2410.00467","kind":"paper","record_sha256":"93b3ddd9b3622133cdf49cb2b59f1c5ef8989fc818e31beebea52ee73de682c0","scope":"core","source_url":"https://arxiv.org/abs/2410.00467","title":"Dynamic Planning for LLM-based Graphical User Interface Automation","topics":["planning","interaction-history","mobile-navigation","robustness"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.00689","kind":"paper","record_sha256":"0cc3c9ba7b3549ba7e0515f70e173bde82fc4645f1d85f0b87a06d451d9d8bac","scope":"core","source_url":"https://arxiv.org/abs/2410.00689","title":"Multimodal Auto Validation For Self-Refinement in Web Agents","topics":["evaluation","robustness","web-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.02810","kind":"paper","record_sha256":"4981ed2934c190e8d6e854d7ecf8c8d9961169d0ee19bdc4bb2430d568c8209b","scope":"adjacent","source_url":"https://arxiv.org/abs/2410.02810","title":"StateAct: Enhancing LLM Base Agents via Self-prompting and State-tracking","topics":["planning","memory"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.02907","kind":"paper","record_sha256":"379720df46453f88cdd7d322fc261210a388affd5782c44e8a5dc511a58686f1","scope":"core","source_url":"https://arxiv.org/abs/2410.02907","title":"NNetNav: Unsupervised Learning of Browser Agents Through Environment Interaction in the Wild","topics":["data-generation","web-navigation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2410.05243","kind":"paper","record_sha256":"eaa07790cff437e918a2b541e83c747edecd6eb865feb37e317fcf9b84bc8778","scope":"core","source_url":"https://arxiv.org/abs/2410.05243","title":"Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents","topics":["grounding","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.06703","kind":"paper","record_sha256":"33e44ecc5b86b2b2534dda161ab3f88a45f4e7fb155caa1fd0fccaabbe1313a3","scope":"core","source_url":"https://arxiv.org/abs/2410.06703","title":"ST-WebAgentBench: A Benchmark for Evaluating Safety and Trustworthiness in Web Agents","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2410.08164","kind":"paper","record_sha256":"0a1b3e7664b06237fa094dde85515ea3c2ede368f1e6b4605f477f923bfdde25","scope":"core","source_url":"https://arxiv.org/abs/2410.08164","title":"Agent S: An Open Agentic Framework that Uses Computers Like a Human","topics":["planning","memory"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2410.09006","kind":"paper","record_sha256":"fdee754a6d63d3fc41e81f26a1769c53a1f195500e74baf3f7cc085f9f276394","scope":"core","source_url":"https://arxiv.org/abs/2410.09006","title":"From Interaction to Impact: Towards Safer AI Agents Through Understanding and Evaluating Mobile UI Operation Impacts","topics":["safety-security","mobile-navigation","evaluation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2410.11871","kind":"paper","record_sha256":"8875afaf1d24fa5680cf5f4a804acb3530eee336ca8d2a7bee18dc602b0e6ef4","scope":"core","source_url":"https://arxiv.org/abs/2410.11871","title":"TinyClick: Single-Turn Agent for Empowering GUI Automation","topics":["grounding","efficiency"],"year":"2024"},{"environments":["android"],"id":"arxiv:2410.11872","kind":"paper","record_sha256":"a427a7aa3800366ffa34eb5377208ebfb8fee794201eee0a648d91e3c582c06f","scope":"core","source_url":"https://arxiv.org/abs/2410.11872","title":"ClickAgent: Enhancing UI Location Capabilities of Autonomous Agents","topics":["mobile-navigation","grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.13232","kind":"paper","record_sha256":"14d357878e8f5ebe870cb9a73a79ccabb21e5df1b234436a16c76ea47862e0ec","scope":"core","source_url":"https://arxiv.org/abs/2410.13232","title":"Web Agents with World Models: Learning and Leveraging Environment Dynamics in Web Navigation","topics":["world-models","planning"],"year":"2024"},{"environments":["android"],"id":"arxiv:2410.13757","kind":"paper","record_sha256":"aee833a27ac08bb23ee529ef1de27e3e68c3a0db609517acc0298d57dac4820d","scope":"core","source_url":"https://arxiv.org/abs/2410.13757","title":"MobA: Multifaceted Memory-Enhanced Adaptive Planning for Efficient Mobile Task Automation","topics":["memory","planning","mobile-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.13824","kind":"paper","record_sha256":"44e2afc613865bd89cae59a91125781d682919eaf8ef58fe537b2e1d0cb89f24","scope":"core","source_url":"https://arxiv.org/abs/2410.13824","title":"Harnessing Webpage UIs for Text-Rich Visual Understanding","topics":["data-generation","screen-understanding","grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.13825","kind":"paper","record_sha256":"b9ee354525191733f5e5dd105a5d39d5d10da1ffda24df62c95df3b8e1c9b0cf","scope":"core","source_url":"https://arxiv.org/abs/2410.13825","title":"AgentOccam: A Simple Yet Strong Baseline for LLM-Based Web Agents","topics":["web-navigation","planning"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.13886","kind":"paper","record_sha256":"5da4b11ea07b22c33ffc941e7dc2ab206826f006d8951876f7ae8d023b8d67f0","scope":"core","source_url":"https://arxiv.org/abs/2410.13886","title":"Refusal-Trained LLMs Are Easily Jailbroken As Browser Agents","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2410.14803","kind":"paper","record_sha256":"5ca3e2553a4972c41b5256e2d8247a0943a53c8ae8c0f3bceec9a2d4e9ccce32","scope":"core","source_url":"https://arxiv.org/abs/2410.14803","title":"DistRL: An Asynchronous Distributed Reinforcement Learning Framework for On-Device Control Agents","topics":["reinforcement-learning","efficiency"],"year":"2024"},{"environments":["android"],"id":"arxiv:2410.15164","kind":"paper","record_sha256":"f836c4fe29212ed65c4399f626d923325166c5856baecc675a36d139c35ebbbc","scope":"core","source_url":"https://arxiv.org/abs/2410.15164","title":"SPA-Bench: A Comprehensive Benchmark for SmartPhone Agent Evaluation","topics":["evaluation","efficiency","mobile-navigation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.16464","kind":"paper","record_sha256":"4197fa96f0a1608779275676647961370037abf0a5f4ff9b0828836559009cd1","scope":"core","source_url":"https://arxiv.org/abs/2410.16464","title":"Beyond Browsing: API-Based Web Agents","topics":["web-navigation","tool-integration","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.17401","kind":"paper","record_sha256":"aacf89f06854c2299892fd0e1ca02b411679bf9263859f3327fe3a82ba37401e","scope":"core","source_url":"https://arxiv.org/abs/2410.17401","title":"AdvAgent: Controllable Blackbox Red-teaming on Web Agents","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2410.17520","kind":"paper","record_sha256":"14763fb7fba561f55c01a871b93e19261071a7f385e83e3763463d303c92b22f","scope":"core","source_url":"https://arxiv.org/abs/2410.17520","title":"MobileSafetyBench: Evaluating Safety of Autonomous Agents in Mobile Device Control","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2410.17883","kind":"paper","record_sha256":"440e005214bbea896f79f3cbcf5baa238efe3856dcac2c81074798ad5a99b700","scope":"core","source_url":"https://arxiv.org/abs/2410.17883","title":"Lightweight Neural App Control","topics":["mobile-navigation","action-prediction","efficiency","interaction-history"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2410.18603","kind":"paper","record_sha256":"77b0e4a374d79b66db66993335d2f810014581cedf84babec298da4288f557b5","scope":"core","source_url":"https://arxiv.org/abs/2410.18603","title":"AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant","topics":["automation","tool-integration"],"year":"2024"},{"environments":["desktop","cross-platform"],"id":"arxiv:2410.18963","kind":"paper","record_sha256":"d88593febf3f4a70c118210569eae3e030808e1d606e981a2e77ee0dc0e37633","scope":"core","source_url":"https://arxiv.org/abs/2410.18963","title":"OSCAR: Operating System Control via State-Aware Reasoning and Re-Planning","topics":["planning","state-tracking","tool-integration","robustness"],"year":"2024"},{"environments":["ios","android","web","cross-platform"],"id":"arxiv:2410.18967","kind":"paper","record_sha256":"bbb0d6b687fa799b9ef217b44ffb58908e45becf167b62480f5677f8ac299184","scope":"core","source_url":"https://arxiv.org/abs/2410.18967","title":"Ferret-UI 2: Mastering Universal User Interface Understanding Across Platforms","topics":["screen-understanding","grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.19100","kind":"paper","record_sha256":"896949c261f8d763928d893ca5aafcac0aa072bbb8e5e6ac793b026b87e1a8b1","scope":"core","source_url":"https://arxiv.org/abs/2410.19100","title":"VideoWebArena: Evaluating Long Context Multimodal Agents with Video Understanding Web Tasks","topics":["evaluation","video","memory"],"year":"2024"},{"environments":["web","cross-platform"],"id":"arxiv:2410.19461","kind":"paper","record_sha256":"31b98dc94a6c3117cf6293d66df0cc2c730d3f2402269666e7130698546efe65","scope":"core","source_url":"https://arxiv.org/abs/2410.19461","title":"EDGE: Enhanced Grounded GUI Understanding with Enriched Multi-Granularity Synthetic Data","topics":["data-generation","grounding","screen-understanding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.19609","kind":"paper","record_sha256":"ecbfa8d37daec52fe25ee7adf541f800e64dc6f76e31db77a690fc028bd9f9b2","scope":"core","source_url":"https://arxiv.org/abs/2410.19609","title":"OpenWebVoyager: Building Multimodal Web Agents via Iterative Real-World Exploration, Feedback and Optimization","topics":["web-navigation","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.22552","kind":"paper","record_sha256":"27e956bbe6218acedda678d9bde9c87bed65f76a150b9de8f249ba0640b05751","scope":"core","source_url":"https://arxiv.org/abs/2410.22552","title":"Auto-Intent: Automated Intent Discovery and Self-Exploration for Large Language Model Web Agents","topics":["planning","web-navigation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2410.23218","kind":"paper","record_sha256":"f2ffee6790016f0b3e7d09539033a10c35e8e10c2012ef3b068d77cd441a98d9","scope":"core","source_url":"https://arxiv.org/abs/2410.23218","title":"OS-ATLAS: A Foundation Action Model for Generalist GUI Agents","topics":["grounding","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.23252","kind":"paper","record_sha256":"ed1651ece6a5e7aa9d640012f2239961538e53f396cbc5000ea4721e82be855d","scope":"core","source_url":"https://arxiv.org/abs/2410.23252","title":"Evaluating Cultural and Social Awareness of LLM Web Agents","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2410.23555","kind":"paper","record_sha256":"ec9d9b5bb9b69d11250ccc535480d2e82e814b9010c6088ac8bef9fa2677e6d1","scope":"core","source_url":"https://arxiv.org/abs/2410.23555","title":"From Context to Action: Analysis of the Impact of State Representation and Context on the Generalization of Multi-Turn Web Navigation Agents","topics":["web-navigation","interaction-history","state-tracking"],"year":"2024"},{"environments":["android"],"id":"arxiv:2410.24024","kind":"paper","record_sha256":"42d30ac7f024cbaa15f5641bfa25fe31515fbcb35df1f0f188126b64d5b55f9e","scope":"core","source_url":"https://arxiv.org/abs/2410.24024","title":"AndroidLab: Training and Systematic Benchmarking of Android Autonomous Agents","topics":["evaluation","data-generation","mobile-navigation"],"year":"2024"},{"environments":["web","android"],"id":"arxiv:2411.00820","kind":"paper","record_sha256":"5054cd73e303de653428a1e66171f0cf7b226d1293071abca5e0dd387214deaa","scope":"core","source_url":"https://arxiv.org/abs/2411.00820","title":"AutoGLM: Autonomous Foundation Agents for GUIs","topics":["automation","reinforcement-learning","grounding"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2411.02006","kind":"survey","record_sha256":"f9da592ace7b5fd2c30366884adb4dd0a061367313b87b2bf7dbcf49cfefeaa6","scope":"core","source_url":"https://arxiv.org/abs/2411.02006","title":"Foundations and Recent Trends in Multimodal Mobile Agents: A Survey","topics":["survey","mobile-navigation","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2411.02337","kind":"paper","record_sha256":"df0cc8c863ef0c51c2d1814613d8fcb1c950431edde305f05b42ce1a2053f590","scope":"core","source_url":"https://arxiv.org/abs/2411.02337","title":"WebRL: Training LLM Web Agents via Self-Evolving Online Curriculum Reinforcement Learning","topics":["reinforcement-learning","web-navigation"],"year":"2024"},{"environments":["desktop","web"],"id":"arxiv:2411.02391","kind":"paper","record_sha256":"7c0809b194c874dc2f451c32ddceeba158415988a31fe8a26ba560f05f0127fd","scope":"core","source_url":"https://arxiv.org/abs/2411.02391","title":"Attacking Vision-Language Computer Agents via Pop-ups","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2411.03817","kind":"paper","record_sha256":"c73ab5d919ed6657ea803943e6ac44de3bff676338a7912604e6601b9f27d408","scope":"adjacent","source_url":"https://arxiv.org/abs/2411.03817","title":"From Novice to Expert: LLM Agent Policy Optimization via Step-wise Reinforcement Learning","topics":["reinforcement-learning","reward-models","self-reflection"],"year":"2024"},{"environments":["web","desktop"],"id":"arxiv:2411.04468","kind":"paper","record_sha256":"03b70f9c55745ca2f10eea5e08e11caa7336483e81ea1989e11fbd118e827029","scope":"adjacent","source_url":"https://arxiv.org/abs/2411.04468","title":"Magentic-One: A Generalist Multi-Agent System for Solving Complex Tasks","topics":["planning","tool-integration"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2411.04890","kind":"survey","record_sha256":"3abab2c1f4f949baa649115df7df7c99a9e254636e022e58402b12b6c4fcbe70","scope":"core","source_url":"https://arxiv.org/abs/2411.04890","title":"GUI Agents with Foundation Models: A Comprehensive Survey","topics":["survey"],"year":"2024"},{"environments":["web"],"id":"arxiv:2411.06559","kind":"paper","record_sha256":"fa538f8370c7704b24b74f8c61afc66c39a2c6c064dde24f04cd2d3e2f987707","scope":"core","source_url":"https://arxiv.org/abs/2411.06559","title":"Is Your LLM Secretly a World Model of the Internet? Model-Based Planning for Web Agents","topics":["world-models","planning"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2411.10323","kind":"paper","record_sha256":"8b4a43c0e44014c3d11ed247c8b8effc3c2aaa5045ed85b43b3eeb16888993ec","scope":"core","source_url":"https://arxiv.org/abs/2411.10323","title":"The Dawn of GUI Agent: A Preliminary Case Study with Claude 3.5 Computer Use","topics":["evaluation","automation","planning"],"year":"2024"},{"environments":["web","desktop","android","windows"],"id":"arxiv:2411.10943","kind":"survey","record_sha256":"c77fe98a748c471cc6e4699d79e96a8575f6de9d8766a2045cb7c037bff93995","scope":"core","source_url":"https://arxiv.org/abs/2411.10943","title":"Generalist Virtual Agents: A Survey on Autonomous Agents Across Digital Platforms","topics":["survey","web-navigation","mobile-navigation","evaluation","planning"],"year":"2024"},{"environments":["web"],"id":"arxiv:2411.13451","kind":"paper","record_sha256":"7c1f8dd2ecc3683782c95f0f32c7693348d4d607fd3c54021f629ff5c6a08fb2","scope":"core","source_url":"https://arxiv.org/abs/2411.13451","title":"AdaptAgent: Adapting Multimodal Web Agents with Few-Shot Learning from Human Demonstrations","topics":["web-navigation","data-generation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2411.13591","kind":"paper","record_sha256":"febc288eb9c4bcf6c94292f300ebb2af7beca5cc0cf94f0d41bb434bef77aaaf","scope":"core","source_url":"https://arxiv.org/abs/2411.13591","title":"Improved GUI Grounding via Iterative Narrowing","topics":["grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2411.15004","kind":"paper","record_sha256":"dbb0911d8bd17a90d68580884007710a2921780ae366bb5cfba5586f6cdc7d4f","scope":"core","source_url":"https://arxiv.org/abs/2411.15004","title":"ScribeAgent: Towards Specialized Web Agents Using Production-Scale Workflow Data","topics":["workflow","web-navigation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2411.17465","kind":"paper","record_sha256":"81614c38ce061c38229d2857cd6353146174135e9a5008e9e34b92b758539da0","scope":"core","source_url":"https://arxiv.org/abs/2411.17465","title":"ShowUI: One Vision-Language-Action Model for GUI Visual Agent","topics":["grounding","efficiency"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2411.18279","kind":"survey","record_sha256":"c60e34bc3e12b89b6b80e797d18f7f4dd293f7c476ef82307e19a0ff7964f8a3","scope":"core","source_url":"https://arxiv.org/abs/2411.18279","title":"Large Language Model-Brained GUI Agents: A Survey","topics":["survey"],"year":"2024"},{"environments":["android"],"id":"arxiv:2412.00402","kind":"paper","record_sha256":"5b799123653e44b27a76db5dfb69e88331471e0bfb0453305c002260f67c41e0","scope":"adjacent","source_url":"https://arxiv.org/abs/2412.00402","title":"DroidCall: A Dataset for LLM-powered Android Intent Invocation","topics":["tool-integration","data-generation"],"year":"2024"},{"environments":["cross-platform","web","desktop"],"id":"arxiv:2412.01268","kind":"paper","record_sha256":"e9dbd754561910311264dcc9f9591e6809176c71561ab3d88315c863135a50c7","scope":"core","source_url":"https://arxiv.org/abs/2412.01268","title":"Ponder & Press: Advancing Visual GUI Agent towards General Computer Control","topics":["grounding","planning"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2412.04454","kind":"paper","record_sha256":"425bdc1097cd70dd1641a725fd8a771df5d7cbfc81b43b564abd7085c53dec61","scope":"core","source_url":"https://arxiv.org/abs/2412.04454","title":"Aguvis: Unified Pure Vision Agents for Autonomous GUI Interaction","topics":["grounding","planning","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2412.05467","kind":"paper","record_sha256":"8478c4d89b40df6573028f6ca61b124da1ff2a144ad15ad6d7ae0b803aefe3aa","scope":"core","source_url":"https://arxiv.org/abs/2412.05467","title":"The BrowserGym Ecosystem for Web Agent Research","topics":["evaluation","sandbox"],"year":"2024"},{"environments":["web","android","cross-platform"],"id":"arxiv:2412.09362","kind":"paper","record_sha256":"7c2ae8bc5d9d195bd4a13e9ab067496bfa23cbcdc582efe5b3bec5c524953fea","scope":"core","source_url":"https://arxiv.org/abs/2412.09362","title":"Falcon-UI: Understanding GUI Before Following User Instructions","topics":["screen-understanding","data-generation"],"year":"2024"},{"environments":["web"],"id":"arxiv:2412.09605","kind":"paper","record_sha256":"060066c48d6c937b9a6fc340464ccd9337a33063e192e015543e9bc83309cdf6","scope":"core","source_url":"https://arxiv.org/abs/2412.09605","title":"AgentTrek: Agent Trajectory Synthesis via Guiding Replay with Web Tutorials","topics":["data-generation","web-navigation"],"year":"2024"},{"environments":["desktop","windows"],"id":"arxiv:2412.10047","kind":"paper","record_sha256":"1d99dbf868c804211f22af0811bb2dc09e503af8ed3660498643836d4508b9a6","scope":"core","source_url":"https://arxiv.org/abs/2412.10047","title":"Large Action Models: From Inception to Implementation","topics":["automation","data-generation","evaluation"],"year":"2024"},{"environments":["web","desktop"],"id":"arxiv:2412.10342","kind":"paper","record_sha256":"73bcc165d51233fd494c6741fe8fc7e551adee3f5e710eba507801f11f8e9d82","scope":"core","source_url":"https://arxiv.org/abs/2412.10342","title":"Iris: Breaking GUI Complexity with Adaptive Focus and Self-Refining","topics":["grounding","screen-understanding","efficiency"],"year":"2024"},{"environments":["web"],"id":"arxiv:2412.10742","kind":"paper","record_sha256":"053ba97a4fe5536fd6ea4edf942ae3715a5d012b1fae73e9875d2f54f093fdc6","scope":"core","source_url":"https://arxiv.org/abs/2412.10742","title":"WEPO: Web Element Preference Optimization for LLM-based Web Navigation","topics":["grounding","web-navigation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2412.10840","kind":"paper","record_sha256":"620f7e78528565892436d1629dee62917406ff782d7adc2e0614930dd3fb5887","scope":"core","source_url":"https://arxiv.org/abs/2412.10840","title":"Attention-driven GUI Grounding: Leveraging Pretrained Multimodal Large Language Models without Fine-Tuning","topics":["grounding"],"year":"2024"},{"environments":["web"],"id":"arxiv:2412.13194","kind":"paper","record_sha256":"f3c4446a9a6e3e47fb740db19e63548d9662cdf0b2bc99e82e8c80fd1e6bee5c","scope":"core","source_url":"https://arxiv.org/abs/2412.13194","title":"Proposer-Agent-Evaluator(PAE): Autonomous Skill Discovery For Foundation Model Internet Agents","topics":["data-generation","reinforcement-learning","web-navigation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2412.13501","kind":"survey","record_sha256":"332bae01dcdecbb7052e412d72780e4d62e958403f4e50745d248c83a55b79ee","scope":"core","source_url":"https://arxiv.org/abs/2412.13501","title":"GUI Agents: A Survey","topics":["survey"],"year":"2024"},{"environments":["web","desktop"],"id":"arxiv:2412.14161","kind":"paper","record_sha256":"4844c0dd69e1f56de5086d9fb9a5c22d1ed5aa79718d960da2cc69c36aec05d8","scope":"core","source_url":"https://arxiv.org/abs/2412.14161","title":"TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks","topics":["evaluation","long-horizon","workflow"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2412.14470","kind":"paper","record_sha256":"bdd78f958c12283f5bf1c6e4c79bc7ed2fef28fddc9aa2f04dcafa132a994b89","scope":"adjacent","source_url":"https://arxiv.org/abs/2412.14470","title":"Agent-SafetyBench: Evaluating the Safety of LLM Agents","topics":["safety-security","evaluation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2412.16256","kind":"paper","record_sha256":"a3a5bdcf8359ffdad7508eddc938a5050e41a7987cc3c60bf19aed435e59e8f8","scope":"core","source_url":"https://arxiv.org/abs/2412.16256","title":"Aria-UI: Visual Grounding for GUI Instructions","topics":["grounding","memory"],"year":"2024"},{"environments":["desktop"],"id":"arxiv:2412.17589","kind":"paper","record_sha256":"34fbff9913f07ea2cb32758a4337d7722a179360d6b48e4b53cf499ec42f1afe","scope":"core","source_url":"https://arxiv.org/abs/2412.17589","title":"PC Agent: While You Sleep, AI Works -- A Cognitive Journey into Digital World","topics":["recording","data-generation","planning"],"year":"2024"},{"environments":["android"],"id":"arxiv:2412.18116","kind":"paper","record_sha256":"2be9b459db2bbb516879d70bf401c10096a5d28f730b6cc2cbb463e1115c48b4","scope":"core","source_url":"https://arxiv.org/abs/2412.18116","title":"AutoDroid-V2: Boosting SLM-based GUI Agents via Code Generation","topics":["mobile-navigation","tool-integration"],"year":"2024"},{"environments":["unspecified"],"id":"arxiv:2412.18426","kind":"paper","record_sha256":"402ec559e726a69d2438fb6f366dd721c571b9aa442c247c23557d88e5fb3385","scope":"core","source_url":"https://arxiv.org/abs/2412.18426","title":"GUI Testing Arena: A Unified Benchmark for Advancing Autonomous GUI Testing Agent","topics":["evaluation","automation","mobile-navigation"],"year":"2024"},{"environments":["cross-platform"],"id":"arxiv:2412.19723","kind":"paper","record_sha256":"0bcbb519c9e404ca70a3e7e296555b9f685295bef63edb78cf20277ce4cf40a5","scope":"core","source_url":"https://arxiv.org/abs/2412.19723","title":"OS-Genesis: Automating GUI Agent Trajectory Construction via Reverse Task Synthesis","topics":["data-generation"],"year":"2024"},{"environments":["android"],"id":"arxiv:2501.01149","kind":"paper","record_sha256":"ac65245285b829efeda0e9c7a2fe972bd54e648905f0173fb03329378fcdf597","scope":"core","source_url":"https://arxiv.org/abs/2501.01149","title":"A3: Android Agent Arena for Mobile GUI Agents with Essential-State Procedural Evaluation","topics":["evaluation","mobile-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2501.04575","kind":"paper","record_sha256":"f9c0287f2e1db02775c4d8e254a136510bf1ba63719ec1bd6f679127fe0eedd6","scope":"core","source_url":"https://arxiv.org/abs/2501.04575","title":"InfiGUIAgent: A Multimodal Generalist GUI Agent with Native Reasoning and Reflection","topics":["grounding","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2501.07572","kind":"paper","record_sha256":"fe277d71eca31d337757ab075d1fc871edb2cc47860cbeb4bc52acc936d22d4a","scope":"core","source_url":"https://arxiv.org/abs/2501.07572","title":"WebWalker: Benchmarking LLMs in Web Traversal","topics":["evaluation","web-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2501.11425","kind":"paper","record_sha256":"4a59d244108a2b47c80260778000143e7bd6478a509b872a63eca1a71479b43f","scope":"adjacent","source_url":"https://arxiv.org/abs/2501.11425","title":"Agent-R: Training Language Model Agents to Reflect via Iterative Self-Training","topics":["data-generation","planning","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2501.11733","kind":"paper","record_sha256":"75315d2a976a1701478588334312518f4fbbe3eda028299cabce29cc1fc389d6","scope":"core","source_url":"https://arxiv.org/abs/2501.11733","title":"Mobile-Agent-E: Self-Evolving Mobile Assistant for Complex Tasks","topics":["mobile-navigation","memory","long-horizon"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2501.12326","kind":"paper","record_sha256":"1210398f0cae14ede5984319d7e3a6191abdc25c9aabfd33a77986404a08de07","scope":"core","source_url":"https://arxiv.org/abs/2501.12326","title":"UI-TARS: Pioneering Automated GUI Interaction with Native Agents","topics":["grounding","planning","data-generation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2501.13896","kind":"paper","record_sha256":"f17ca5023d0aef5470ca3f74d0ffd63d9d6c8f055bf28f779417f966d354ab3c","scope":"core","source_url":"https://arxiv.org/abs/2501.13896","title":"GUI-Bee: Align GUI Action Grounding to Novel Environments via Autonomous Exploration","topics":["grounding","data-generation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2501.16150","kind":"survey","record_sha256":"47f4f60743e58a3b652194d8b8fa67afbb725ae4e6ebae4db8fa3176cd6f3d32","scope":"core","source_url":"https://arxiv.org/abs/2501.16150","title":"A Comprehensive Survey of Agents for Computer Use: Foundations, Challenges, and Future Directions","topics":["survey"],"year":"2025"},{"environments":["web"],"id":"arxiv:2501.16609","kind":"paper","record_sha256":"0aeaa020233a142296002926ededdde08a8428d1719caebd72a180c72d5556a6","scope":"core","source_url":"https://arxiv.org/abs/2501.16609","title":"CowPilot: A Framework for Autonomous and Human-Agent Collaborative Web Navigation","topics":["web-navigation","human-in-the-loop","evaluation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2502.01977","kind":"paper","record_sha256":"1669da65656f17179f2e744491f23ead3901bacb846c2b445b875eb6cc390052","scope":"core","source_url":"https://arxiv.org/abs/2502.01977","title":"AutoGUI: Scaling GUI Grounding with Automatic Functionality Annotations from LLMs","topics":["grounding","data-generation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2502.02955","kind":"paper","record_sha256":"96c87e089cf709a94d4b1bc486bb895cdf8d37fe93e17528a405cf2e4c313b45","scope":"core","source_url":"https://arxiv.org/abs/2502.02955","title":"ReachAgent: Enhancing Mobile Agent via Page Reaching and Operation","topics":["mobile-navigation","planning","data-generation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2502.02982","kind":"paper","record_sha256":"b7810c30981422b5197ebca75142dbe32e0a7d3a9a93706ceba48b245644d2d0","scope":"core","source_url":"https://arxiv.org/abs/2502.02982","title":"MobileA3gent: Training Mobile GUI Agents Using Decentralized Self-Sourced Data from Diverse Users","topics":["mobile-navigation","data-generation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2502.06395","kind":"paper","record_sha256":"f8dbff0f7341467b216edccffcc41f18956767cc3040e919dc9cdcc112b22d3d","scope":"core","source_url":"https://arxiv.org/abs/2502.06395","title":"AppVLM: A Lightweight Vision Language Model for Online App Control","topics":["mobile-navigation","imitation-learning","data-generation","efficiency"],"year":"2025"},{"environments":["web"],"id":"arxiv:2502.07942","kind":"paper","record_sha256":"8f272e28a30265bb32d0e794a34b77deefce01f04b5852684635864e53b13207","scope":"core","source_url":"https://arxiv.org/abs/2502.07942","title":"Symbiotic Cooperation for Web Agents: Harnessing Complementary Strengths of Large and Small LLMs","topics":["data-generation","web-navigation","efficiency"],"year":"2025"},{"environments":["web","cross-platform"],"id":"arxiv:2502.07949","kind":"paper","record_sha256":"7cd25a2e8b12d6cf15533b9d3c5a50426f307ddf37c9722d403bf549aa64c343","scope":"core","source_url":"https://arxiv.org/abs/2502.07949","title":"Advancing Autonomous VLM Agents via Variational Subgoal-Conditioned Reinforcement Learning","topics":["reinforcement-learning","planning"],"year":"2025"},{"environments":["desktop","web"],"id":"arxiv:2502.08047","kind":"paper","record_sha256":"ee819c71fdc89eca5d54eaf8ccbbeae6245064f4a223d65f6865d9a235db74a8","scope":"core","source_url":"https://arxiv.org/abs/2502.08047","title":"WorldGUI: An Interactive Benchmark for Desktop GUI Automation from Any Starting Point","topics":["evaluation","robustness","planning"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2502.08226","kind":"paper","record_sha256":"ce9da8ae0c4ad9e949231880cc65df5fed064fa0070b181e7b047eb6c149f601","scope":"core","source_url":"https://arxiv.org/abs/2502.08226","title":"TRISHUL: Towards Region Identification and Screen Hierarchy Understanding for Large VLM based GUI Agents","topics":["grounding","screen-understanding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2502.11357","kind":"paper","record_sha256":"19ff24dca7a38f27055f62ab4db75d601905014b4364da643855f16f8f3af82c","scope":"core","source_url":"https://arxiv.org/abs/2502.11357","title":"Explorer: Scaling Exploration-driven Web Trajectory Synthesis for Multimodal Web Agents","topics":["data-generation","web-navigation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2502.13053","kind":"paper","record_sha256":"0e4cd3e7f64d1d6e59dc7ec6880a2b8b02423b0fd24e8f0edcf9aa1ae614d453","scope":"core","source_url":"https://arxiv.org/abs/2502.13053","title":"Evaluating the Robustness of Multimodal Agents Against Active Environmental Injection Attacks","topics":["safety-security","robustness","evaluation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2502.14282","kind":"paper","record_sha256":"8b6303558cc05421efba4779a138b9a251b3bd0d037442ab4f35c7d1d590ebbb","scope":"core","source_url":"https://arxiv.org/abs/2502.14282","title":"PC-Agent: A Hierarchical Multi-Agent Collaboration Framework for Complex Task Automation on PC","topics":["planning","cross-app"],"year":"2025"},{"environments":["android"],"id":"arxiv:2502.15760","kind":"paper","record_sha256":"be71d6c3ecc167b0d01bef57bec445659ebe8285fd35ebf6f35749d6e41d9c52","scope":"core","source_url":"https://arxiv.org/abs/2502.15760","title":"Digi-Q: Learning Q-Value Functions for Training Device-Control Agents","topics":["reinforcement-learning","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2502.17110","kind":"paper","record_sha256":"bbf1b06ffae37600e6934cafea369d6052acdcc26ea02b4f58d6f2c290f96faf","scope":"core","source_url":"https://arxiv.org/abs/2502.17110","title":"Mobile-Agent-V: A Video-Guided Approach for Effortless and Efficient Operational Knowledge Injection in Mobile Automation","topics":["mobile-navigation","video","memory"],"year":"2025"},{"environments":["web"],"id":"arxiv:2502.17903","kind":"paper","record_sha256":"e2b5f118989553ba7bf3d7735bc6803cc224ae0cda66ef36bbaa812bc6d382af","scope":"core","source_url":"https://arxiv.org/abs/2502.17903","title":"Towards Sustainable Web Agents: A Plea for Transparency and Dedicated Metrics for Energy Consumption","topics":["evaluation","efficiency"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2502.18525","kind":"paper","record_sha256":"96903e55d127010856fdfaa271fbc7a74685c9662aa45a44ef3c5a013e7e372b","scope":"core","source_url":"https://arxiv.org/abs/2502.18525","title":"Programming with Pixels: Can Computer-Use Agents do Software Engineering?","topics":["evaluation","tool-integration","automation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2502.18906","kind":"paper","record_sha256":"59c9ee1adbfdda24551cd397b65c54e1c6bffe32c147208725453827d404264a","scope":"core","source_url":"https://arxiv.org/abs/2502.18906","title":"VEM: Environment-Free Exploration for Training GUI Agent with Value Environment Model","topics":["reinforcement-learning","world-models"],"year":"2025"},{"environments":["web"],"id":"arxiv:2502.20383","kind":"paper","record_sha256":"1066da7e9849c99a7c876262348951a1b4aa0ab95f75525876e8397815c42bc3","scope":"core","source_url":"https://arxiv.org/abs/2502.20383","title":"Why Are Web AI Agents More Vulnerable Than Standalone LLMs? A Security Analysis","topics":["safety-security","evaluation","web-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.00401","kind":"paper","record_sha256":"5ebb108f107860fffbe73189eb66dd0a04693993ff816a7e3ec62e855b62345b","scope":"core","source_url":"https://arxiv.org/abs/2503.00401","title":"Smoothing Grounding and Reasoning for MLLM-Powered GUI Agents with Query-Oriented Pivot Tasks","topics":["grounding","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.01861","kind":"paper","record_sha256":"80b6415f022fd85d643ee2650c6aa9bb7ccaf6a123252e417efccfa77ba2b6c2","scope":"core","source_url":"https://arxiv.org/abs/2503.01861","title":"Towards Enterprise-Ready Computer Using Generalist Agent","topics":["automation","evaluation","workflow"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.02197","kind":"paper","record_sha256":"460063a1595768af7f1c8f311e2b1d5bcf9ff0b61292d572cd52719c3ecb6d04","scope":"adjacent","source_url":"https://arxiv.org/abs/2503.02197","title":"ATLaS: Agent Tuning via Learning Critical Steps","topics":["data-generation","planning","efficiency"],"year":"2025"},{"environments":["android"],"id":"arxiv:2503.02268","kind":"paper","record_sha256":"800880517e522df440bb37ecbb951c417322b5722fb5516ad090837e06f4e9a2","scope":"core","source_url":"https://arxiv.org/abs/2503.02268","title":"AppAgentX: Evolving GUI Agents as Proficient Smartphone Users","topics":["memory","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2503.02403","kind":"paper","record_sha256":"ff05f6fc25947a7087db8c08b1ff1dec74e141bcaadb23d934eadd1115648362","scope":"core","source_url":"https://arxiv.org/abs/2503.02403","title":"AutoEval: A Practical Framework for Autonomous Evaluation of Mobile Agents","topics":["evaluation","mobile-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.02682","kind":"paper","record_sha256":"63703ae4781d1e59e331994bd5536b6e7e5e2438f5133546c2eac8ffb44b1fba","scope":"adjacent","source_url":"https://arxiv.org/abs/2503.02682","title":"MPO: Boosting LLM Agents with Meta Plan Optimization","topics":["planning","memory"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.02950","kind":"paper","record_sha256":"9e93cbae059137d18830c9cc720bb1a5ddf4a5d2c8507142301a29359db90089","scope":"core","source_url":"https://arxiv.org/abs/2503.02950","title":"LiteWebAgent: The Open-Source Suite for VLM-Based Web-Agent Applications","topics":["web-navigation","tool-integration","planning"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.03196","kind":"paper","record_sha256":"25df5f51f64f4787e10e9cc00191a04842569a67ed9e2c4beee13ef6d1202515","scope":"core","source_url":"https://arxiv.org/abs/2503.03196","title":"SpiritSight Agent: Advanced GUI Agent with One Look","topics":["grounding","screen-understanding","data-generation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2503.03743","kind":"paper","record_sha256":"cc5b6c0897510198778f6d092ca5fa500943cddff31f12840634f4fc8ef5ec54","scope":"core","source_url":"https://arxiv.org/abs/2503.03743","title":"CHOP: Mobile Operating Assistant with Constrained High-frequency Optimized Subtask Planning","topics":["mobile-navigation","planning"],"year":"2025"},{"environments":["desktop","windows"],"id":"arxiv:2503.04730","kind":"paper","record_sha256":"71308d808831a201fce8c487e21a74ddbe10e1cb35d865c8563044db6560dd47","scope":"core","source_url":"https://arxiv.org/abs/2503.04730","title":"WinClick: GUI Grounding with Multimodal Large Language Models","topics":["grounding","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.04957","kind":"paper","record_sha256":"823a9e37264bce388ef53ae8cf4eb6995a13291e6ff7fcab9d312bb6bdb2cc54","scope":"core","source_url":"https://arxiv.org/abs/2503.04957","title":"SafeArena: Evaluating the Safety of Autonomous Web Agents","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2503.05143","kind":"paper","record_sha256":"ca8c80ac5afddad11c6e45e13dcdedbedfa11c75055c136815e76d6064e9778f","scope":"core","source_url":"https://arxiv.org/abs/2503.05143","title":"FedMABench: Benchmarking Mobile Agents on Decentralized Heterogeneous User Data","topics":["evaluation","mobile-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.06470","kind":"paper","record_sha256":"0f8083553591b42f985f6b89e62cd2f5301c32b17bfcced3c492db8a9f4068f4","scope":"core","source_url":"https://arxiv.org/abs/2503.06470","title":"Think Twice, Click Once: Enhancing GUI Grounding via Fast and Slow Systems","topics":["grounding","efficiency","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.07919","kind":"paper","record_sha256":"c1e36cb0c653498bdf5d1b5422e498aa8a396a345446b5170add78e76282a5e0","scope":"core","source_url":"https://arxiv.org/abs/2503.07919","title":"BEARCUBS: A benchmark for computer-using web agents","topics":["evaluation","web-navigation","video"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2503.09241","kind":"paper","record_sha256":"b92b3c8fbb2b2836eae7f03cdccfd8c8087d275b6f73662cdae37fe09e18b2a7","scope":"core","source_url":"https://arxiv.org/abs/2503.09241","title":"In-Context Defense in Computer Agents: An Empirical Study","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["desktop","windows"],"id":"arxiv:2503.09263","kind":"paper","record_sha256":"bad51141f2f622f1df1de0db6cfe4010a511edd27cae23318bb4e9dbd1c8d49f","scope":"core","source_url":"https://arxiv.org/abs/2503.09263","title":"COLA: A Scalable Multi-Agent Framework For Windows UI Task Automation","topics":["planning","memory","human-in-the-loop"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.09572","kind":"paper","record_sha256":"01522df78e9b02c99651ce0e523e5e7ffe62da4ebf46a2eae9a221c333a0b7f3","scope":"core","source_url":"https://arxiv.org/abs/2503.09572","title":"Plan-and-Act: Improving Planning of Agents for Long-Horizon Tasks","topics":["planning","long-horizon","data-generation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.09780","kind":"paper","record_sha256":"2c75f3230f6d389dc5a768ceacb9a64014de5fc6e30f65716bc15e37b784d249","scope":"core","source_url":"https://arxiv.org/abs/2503.09780","title":"AgentDAM: Privacy Leakage Evaluation for Autonomous Web Agents","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.11069","kind":"paper","record_sha256":"6abe7534353aeb4b4342c7e349f49e6546af654eaad383684eb03e8f506e21a1","scope":"core","source_url":"https://arxiv.org/abs/2503.11069","title":"API Agents vs. GUI Agents: Divergence and Convergence","topics":["tool-integration","survey","automation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2503.11170","kind":"paper","record_sha256":"d9a7083e71389d29997ea863a800ca9d95a3040b3ba347b1201121bbe1a721e7","scope":"core","source_url":"https://arxiv.org/abs/2503.11170","title":"DeskVision: Large Scale Desktop Region Captioning for Advanced GUI Agents","topics":["data-generation","grounding","screen-understanding"],"year":"2025"},{"environments":["desktop","windows"],"id":"arxiv:2503.12532","kind":"paper","record_sha256":"3d38364a09a88bf35ad7326a57fcb966cf7bbf7af0c5b3560a5dcd0b6ebb3e78","scope":"core","source_url":"https://arxiv.org/abs/2503.12532","title":"STEVE: A Step Verification Pipeline for Computer-use Agent Training","topics":["data-generation","evaluation","reinforcement-learning"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.14021","kind":"paper","record_sha256":"75bafc3f2b339d92744a01e4fd88280ec4673a95ca184da5320c30527bc053f4","scope":"core","source_url":"https://arxiv.org/abs/2503.14021","title":"MP-GUI: Modality Perception with MLLMs for GUI Understanding","topics":["screen-understanding","grounding","data-generation"],"year":"2025"},{"environments":["web","cross-platform"],"id":"arxiv:2503.15515","kind":"paper","record_sha256":"43ed9d0adf255c5a4985e12fdea7fe61cda808d2870f31bc5a0c7a4d3abd9934","scope":"core","source_url":"https://arxiv.org/abs/2503.15515","title":"Towards Computer-Using Personal Agents","topics":["safety-security","memory","human-in-the-loop"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2503.15661","kind":"paper","record_sha256":"e71d5d84fe59d9ea7193065b87eecae3726e8c00daa59ad5b1b797081dee6b82","scope":"core","source_url":"https://arxiv.org/abs/2503.15661","title":"UI-Vision: A Desktop-centric GUI Benchmark for Visual Perception and Interaction","topics":["grounding","evaluation","screen-understanding"],"year":"2025"},{"environments":["android"],"id":"arxiv:2503.15937","kind":"paper","record_sha256":"5e5277369b9dc34a76603af3447580a46567915b8c669adf9a126a08557fab84","scope":"core","source_url":"https://arxiv.org/abs/2503.15937","title":"Advancing Mobile GUI Agents: A Verifier-Driven Approach to Practical Deployment","topics":["mobile-navigation","efficiency","planning"],"year":"2025"},{"environments":["android"],"id":"arxiv:2503.16465","kind":"paper","record_sha256":"5356a6152cb6e330437eeeed54fd9b45f4df912a541590998bbba04afb494009","scope":"core","source_url":"https://arxiv.org/abs/2503.16465","title":"OS-Kairos: Adaptive Interaction for MLLM-Powered GUI Agents","topics":["human-in-the-loop","safety-security","mobile-navigation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2503.16788","kind":"paper","record_sha256":"b50902c319c77f6d82848a38eb98542e8529594dd05ce6173a0384bcda658845","scope":"core","source_url":"https://arxiv.org/abs/2503.16788","title":"Does Chain-of-Thought Reasoning Help Mobile GUI Agent? An Empirical Study","topics":["evaluation","mobile-navigation","planning"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.17709","kind":"paper","record_sha256":"b8361678893e17e6a31e887790d014b0c32976fa9367c6315cbd81b478635e62","scope":"core","source_url":"https://arxiv.org/abs/2503.17709","title":"GUI-Xplore: Empowering Generalizable GUI Agents with One Exploration","topics":["data-generation","planning","video"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2503.18492","kind":"paper","record_sha256":"7a2eac9f74845ea3da4edb40f64a791a03fd7705d0a79c8e9090a819fb1d14bf","scope":"core","source_url":"https://arxiv.org/abs/2503.18492","title":"VeriSafe Agent: Safeguarding Mobile GUI Agent via Logic-based Action Verification","topics":["safety-security","mobile-navigation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2503.19537","kind":"paper","record_sha256":"32557f9cd750ab1aeb686d350e8c96441753a85a5a46c60d565c68be8d2f0b60","scope":"core","source_url":"https://arxiv.org/abs/2503.19537","title":"Agent-Initiated Interaction in Phone UI Automation","topics":["human-in-the-loop","mobile-navigation","evaluation"],"year":"2025"},{"environments":["desktop","web"],"id":"arxiv:2503.20279","kind":"paper","record_sha256":"1b022d3ff449c5ecd8a235c800fd9ccb25db016231b08668ae5cb0148858c7a7","scope":"core","source_url":"https://arxiv.org/abs/2503.20279","title":"sudo rm -rf agentic_security","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.20978","kind":"paper","record_sha256":"00b1f114e39d6e7425b3e22d564fc25fc7d72f08fec4e921bfe0af310add8357","scope":"core","source_url":"https://arxiv.org/abs/2503.20978","title":"ScreenLLM: Stateful Screen Schema for Efficient Action Understanding and Prediction","topics":["screen-understanding","memory"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.21620","kind":"paper","record_sha256":"d85bccbd9442b8c957fbf2c4b65df52d9f5ec95de4a07951ed55824b497b37e8","scope":"core","source_url":"https://arxiv.org/abs/2503.21620","title":"UI-R1: Enhancing Efficient Action Prediction of GUI Agents by Reinforcement Learning","topics":["reinforcement-learning","grounding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.22738","kind":"paper","record_sha256":"6fa6a1f1e4c728e90514e6210a30ed6804ff080caf2535be9ca1cf7e04c883a1","scope":"core","source_url":"https://arxiv.org/abs/2503.22738","title":"ShieldAgent: Shielding Agents via Verifiable Safety Policy Reasoning","topics":["safety-security"],"year":"2025"},{"environments":["web"],"id":"arxiv:2503.23350","kind":"survey","record_sha256":"7c17451d2a680c52689b6960ddfd8927615ba86a8b36bb4009c40f4c60735b13","scope":"core","source_url":"https://arxiv.org/abs/2503.23350","title":"A Survey of WebAgents: Towards Next-Generation AI Agents for Web Automation with Large Foundation Models","topics":["survey","web-navigation","safety-security"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2503.23434","kind":"survey","record_sha256":"2ad9215c8a39a9c088cfc9d255d4e2a3695d5b1812f22e55cc81893a65191600","scope":"core","source_url":"https://arxiv.org/abs/2503.23434","title":"Towards Trustworthy GUI Agents: A Survey","topics":["survey","safety-security"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2503.24180","kind":"paper","record_sha256":"ea3e352bdd9ab924deeee3a4e35b2588b710389ee74e04955d25bd433d6e53b6","scope":"core","source_url":"https://arxiv.org/abs/2503.24180","title":"Navi-plus: Managing Ambiguous GUI Navigation Tasks with Follow-up Questions","topics":["human-in-the-loop","evaluation","planning"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2504.00906","kind":"paper","record_sha256":"58827491a5499cbedf58aa1f058aa7db9e171d37f10c78fbc46553439597f240","scope":"core","source_url":"https://arxiv.org/abs/2504.00906","title":"Agent S2: A Compositional Generalist-Specialist Framework for Computer Use Agents","topics":["grounding","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.01382","kind":"paper","record_sha256":"861940f4b25a4547e10508d8f2a7b0a3b58e2ac62b50d85d80409705345d57b1","scope":"core","source_url":"https://arxiv.org/abs/2504.01382","title":"An Illusion of Progress? Assessing the Current State of Web Agents","topics":["evaluation","web-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.01495","kind":"paper","record_sha256":"4346c71da8314c41ed01cc819374072ef492f2c7aa09d5bfe270fe4f16b97b61","scope":"core","source_url":"https://arxiv.org/abs/2504.01495","title":"Are Autonomous Web Agents Good Testers?","topics":["automation","evaluation","web-navigation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2504.02357","kind":"paper","record_sha256":"827fe8c81335c93e9fc6da88fd0b4ddc73a88485a44761ee886fb1b8b94bad37","scope":"core","source_url":"https://arxiv.org/abs/2504.02357","title":"ReuseDroid: A VLM-empowered Android UI Test Migrator Boosted by Active Feedback","topics":["automation","evaluation","cross-app"],"year":"2025"},{"environments":["web","desktop","cross-platform"],"id":"arxiv:2504.04716","kind":"paper","record_sha256":"dd2906c7dac54d780aecb4502c2887d7706ece69970442b5d1f35a4a58818f0b","scope":"core","source_url":"https://arxiv.org/abs/2504.04716","title":"On the Robustness of GUI Grounding Models Against Image Attacks","topics":["grounding","robustness","safety-security","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.07079","kind":"paper","record_sha256":"e991ba4becc579ef033a92f3fc857ffa243337eeb429d850269b3a495b562724","scope":"core","source_url":"https://arxiv.org/abs/2504.07079","title":"SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skills","topics":["web-navigation","tool-integration","planning"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2504.07981","kind":"paper","record_sha256":"513393931f48b096a9448e0bc09584b79bc2b4bf0bbdb6cd421baee3fc8d38d0","scope":"core","source_url":"https://arxiv.org/abs/2504.07981","title":"ScreenSpot-Pro: GUI Grounding for Professional High-Resolution Computer Use","topics":["grounding","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.08942","kind":"paper","record_sha256":"b80f457cfcac301287ec496363a8773d44269609ab1908d301d1ea6eff06ff2b","scope":"core","source_url":"https://arxiv.org/abs/2504.08942","title":"AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories","topics":["evaluation","web-navigation"],"year":"2025"},{"environments":["android","web"],"id":"arxiv:2504.09352","kind":"paper","record_sha256":"7d2904c34ceb6bf2f9051f43998a98bd99f6ac587af42ea70b652aab6b059ef4","scope":"core","source_url":"https://arxiv.org/abs/2504.09352","title":"Explorer: Robust Collection of Interactable GUI Elements","topics":["data-generation","recording","grounding","planning"],"year":"2025"},{"environments":["web","android"],"id":"arxiv:2504.10127","kind":"paper","record_sha256":"07ddea79984076adadabc3f52aa5344951819d252230cd6379d35e6118e91703","scope":"core","source_url":"https://arxiv.org/abs/2504.10127","title":"Breaking the Data Barrier -- Building GUI Agents Through Task Generalization","topics":["data-generation","planning","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.10445","kind":"paper","record_sha256":"4a430f0fc9a84ee0b5ccc22b2930ede1356aab333b31149073d5e3d960339289","scope":"core","source_url":"https://arxiv.org/abs/2504.10445","title":"RealWebAssist: A Benchmark for Long-Horizon Web Assistance with Real-World Users","topics":["evaluation","long-horizon","human-in-the-loop","grounding"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2504.10458","kind":"paper","record_sha256":"9f6c2f66a1ba13cb9358fd604ef4dce55eff827feea7a48128e3f4983a369e79","scope":"core","source_url":"https://arxiv.org/abs/2504.10458","title":"GUI-R1 : A Generalist R1-Style Vision-Language Action Model For GUI Agents","topics":["reinforcement-learning","grounding"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2504.11257","kind":"paper","record_sha256":"2a21253c87ee5807238158bef551098600e88d332222c7782734193def869d05","scope":"core","source_url":"https://arxiv.org/abs/2504.11257","title":"UI-E2I-Synth: Advancing GUI Grounding with Large-Scale Instruction Synthesis","topics":["data-generation","grounding","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.11281","kind":"paper","record_sha256":"4fd1c274efe4c6e9f02407ffdd27d27c80e06688dce7f941bfcc9cacfdbe4131","scope":"core","source_url":"https://arxiv.org/abs/2504.11281","title":"The Obvious Invisible Threat: LLM-Powered GUI Agents' Vulnerability to Fine-Print Injections","topics":["safety-security","human-in-the-loop","evaluation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2504.11675","kind":"paper","record_sha256":"164d63cc2577f38360d871323559f2df244011ef87ef132a266aade683bc3305","scope":"core","source_url":"https://arxiv.org/abs/2504.11675","title":"VLM-Fuzz: Vision Language Model Assisted Recursive Depth-first Search Exploration for Effective UI Testing of Android Apps","topics":["automation","evaluation","mobile-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.11788","kind":"paper","record_sha256":"9152ec78924bef7ca2545a7afe50f749601ab7b22035198eabab190297352e11","scope":"core","source_url":"https://arxiv.org/abs/2504.11788","title":"WebRollback: Enhancing Web Agents with Explicit Rollback Mechanisms","topics":["web-navigation","planning","robustness"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2504.12679","kind":"paper","record_sha256":"cdd2b29f6667d3be9dae9de56a7619c40cfa8100f323013cba795a0873eae333","scope":"core","source_url":"https://arxiv.org/abs/2504.12679","title":"TongUI: Internet-Scale Trajectories from Multimodal Web Tutorials for Generalized GUI Agents","topics":["data-generation","grounding","video"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2504.13805","kind":"paper","record_sha256":"2be7ce5ec7c9a6ed5ec01f682b5808f1c29ee3103c87747e5cc3b2505c2b3564","scope":"core","source_url":"https://arxiv.org/abs/2504.13805","title":"LearnAct: Few-Shot Mobile GUI Agent with a Unified Demonstration Benchmark","topics":["mobile-navigation","memory","human-in-the-loop","evaluation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2504.13865","kind":"survey","record_sha256":"a5978df66e1a12122bca472c396dc6a3f6db52cbc9c9339313ed738d3602e5d0","scope":"core","source_url":"https://arxiv.org/abs/2504.13865","title":"A Survey on (M)LLM-Based GUI Agents","topics":["survey"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2504.13936","kind":"paper","record_sha256":"4ed6c96e523a523a1d3d5f103615e82eb3c8d559d12677c2ceab81760f9ed1f1","scope":"core","source_url":"https://arxiv.org/abs/2504.13936","title":"ViMo: A Generative Visual GUI World Model for App Agents","topics":["world-models","planning","mobile-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2504.14239","kind":"paper","record_sha256":"0053e3870756d3f56619eeaa9cbd359d1fb065738ce7a9c7f9578ae32c7d910b","scope":"core","source_url":"https://arxiv.org/abs/2504.14239","title":"InfiGUI-R1: Advancing Multimodal GUI Agents from Reactive Actors to Deliberative Reasoners","topics":["planning","reinforcement-learning"],"year":"2025"},{"environments":["desktop","windows"],"id":"arxiv:2504.14603","kind":"paper","record_sha256":"dcaf9d4302222d4eb4e6742916f98dd776d69dcf4bdd22e6c936ef4c85f2a974","scope":"core","source_url":"https://arxiv.org/abs/2504.14603","title":"UFO2: The Desktop AgentOS","topics":["automation","cross-app","tool-integration"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2504.16073","kind":"paper","record_sha256":"b43fcba14e5604dc2c6ccc34bbf3afb3245ce3e549405ab65a27907459dff3f8","scope":"core","source_url":"https://arxiv.org/abs/2504.16073","title":"Guiding VLM Agents with Process Rewards at Inference Time for GUI Navigation","topics":["reward-models","self-reflection","grounding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.16419","kind":"paper","record_sha256":"147a376e074f8b6c96fa5b142509f390c1bde77c2e35a918e00fb461ca43f794","scope":"core","source_url":"https://arxiv.org/abs/2504.16419","title":"PixelWeb: The First Web GUI Dataset with Pixel-Wise Labels","topics":["data-generation","grounding","screen-understanding"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2504.17934","kind":"paper","record_sha256":"15e0e1829b1618c358bb98f74c0c4de3a09ce36a94fabc54490ee0748727f733","scope":"core","source_url":"https://arxiv.org/abs/2504.17934","title":"Toward a Human-Centered Evaluation Framework for Trustworthy LLM-Powered GUI Agents","topics":["evaluation","safety-security","human-in-the-loop"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.18575","kind":"paper","record_sha256":"54f488b38d11873eee6ff9a19ae6dea91ce72fd5c5b645508366960d12073428","scope":"core","source_url":"https://arxiv.org/abs/2504.18575","title":"WASP: Benchmarking Web Agent Security Against Prompt Injection Attacks","topics":["safety-security","evaluation","web-navigation"],"year":"2025"},{"environments":["android","ios"],"id":"arxiv:2504.19838","kind":"survey","record_sha256":"8493b73746a84ffac7edc2ecd0fed49f0f5188bb7f17c8929ab8429064673cd6","scope":"core","source_url":"https://arxiv.org/abs/2504.19838","title":"LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects","topics":["survey"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2504.20464","kind":"survey","record_sha256":"63ee18ff335163a3a5472b403564910e98b125b93cb017094baad0d8c1256196","scope":"core","source_url":"https://arxiv.org/abs/2504.20464","title":"A Survey on GUI Agents with Foundation Models Enhanced by Reinforcement Learning","topics":["survey","reinforcement-learning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2504.21024","kind":"paper","record_sha256":"3aa2f0d83080ded691056ac26c03cdc8fa21b6da6897f3b87617967399a0799e","scope":"core","source_url":"https://arxiv.org/abs/2504.21024","title":"WebEvolver: Enhancing Web Agent Self-Improvement with Coevolving World Model","topics":["world-models","data-generation","web-navigation","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.00416","kind":"paper","record_sha256":"4e8f0e887d727105b69cef3b7b0de10bf93ffaadb1fdb80d4fac838624a4088e","scope":"core","source_url":"https://arxiv.org/abs/2505.00416","title":"ScaleTrack: Scaling and back-tracking Automated GUI Agents","topics":["grounding","planning","data-generation","interaction-history"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.00684","kind":"paper","record_sha256":"fe562f421fe7d886b5621eaf41ebf85574d6d5ef1404a7ab413a0b26d6dcb1a4","scope":"core","source_url":"https://arxiv.org/abs/2505.00684","title":"Visual Test-time Scaling for GUI Agent Grounding","topics":["grounding","screen-understanding","web-navigation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2505.03570","kind":"paper","record_sha256":"8844947ec98f0c792e008cc0b8e6e3d96ab8352daa1c923f7fe9d3f01ae7b0ab","scope":"core","source_url":"https://arxiv.org/abs/2505.03570","title":"OSUniverse: Benchmark for Multimodal GUI-navigation AI Agents","topics":["evaluation","cross-app","long-horizon"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.05026","kind":"paper","record_sha256":"4bab6f32a89be93d6fc280d8a8dcbb6e4438273a09f3216f9e2329916236601f","scope":"adjacent","source_url":"https://arxiv.org/abs/2505.05026","title":"Do MLLMs Capture How Interfaces Guide User Behavior? A Benchmark for Multimodal UI/UX Design Understanding","topics":["screen-understanding","ui-design","evaluation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2505.05440","kind":"paper","record_sha256":"df62d26b488027554eb843dc58db948d0e4aed1a9b6def9e6e7feda9c5e56436","scope":"core","source_url":"https://arxiv.org/abs/2505.05440","title":"EcoAgent: An Efficient Device-Cloud Collaborative Multi-Agent Framework for Mobile Automation","topics":["mobile-navigation","planning","efficiency"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.09875","kind":"paper","record_sha256":"41627931777c8bd9d277e9f9fa8e07cceff688c8ecdae3246c3a5905d307ecdb","scope":"core","source_url":"https://arxiv.org/abs/2505.09875","title":"Characterizing Unintended Consequences in Human-GUI Agent Collaboration for Web Browsing","topics":["human-in-the-loop","evaluation","safety-security"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.10593","kind":"paper","record_sha256":"b550e9044198ee4ac747f3252f4344178e626be4e2eb80e8106c6eac6d90cf50","scope":"core","source_url":"https://arxiv.org/abs/2505.10593","title":"LLM-Explorer: Towards Efficient and Affordable LLM-based Exploration for Mobile Apps","topics":["mobile-navigation","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.10831","kind":"paper","record_sha256":"dc3b71fe0984dc6147a5ad60384f8198ded6c921d579d6df36a5fcba28ad2aa8","scope":"core","source_url":"https://arxiv.org/abs/2505.10831","title":"Creating General User Models from Computer Use","topics":["memory","human-in-the-loop","cross-app"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2505.10924","kind":"survey","record_sha256":"8a6a39ad469a7785aaeef404f78c2cdf3072778ec7517f72c6d5b74bb8b22b6b","scope":"core","source_url":"https://arxiv.org/abs/2505.10924","title":"A Survey on the Safety and Security Threats of Computer-Using Agents: JARVIS or Ultron?","topics":["survey","safety-security"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.11891","kind":"paper","record_sha256":"1ce28465b7b0cc1eca2ae30ab5b0b194b5347582fc88e1b689d1f30a8fc99c36","scope":"core","source_url":"https://arxiv.org/abs/2505.11891","title":"Mobile-Bench-v2: A More Realistic and Comprehensive Benchmark for VLM-based Mobile Agents","topics":["evaluation","mobile-navigation","robustness","human-in-the-loop"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.12299","kind":"paper","record_sha256":"85a5b6fccdf157cbe91850bd83e0170dea4b2672e927da54ed88c9946236476c","scope":"core","source_url":"https://arxiv.org/abs/2505.12299","title":"MobileIPL: Enhancing Mobile Agents Thinking Process via Iterative Preference Learning","topics":["mobile-navigation","reinforcement-learning","planning","data-generation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.12370","kind":"paper","record_sha256":"cf2a9ce5205ea940ea1399f1a9651a957504032f70ead68926b95f1901dbc903","scope":"core","source_url":"https://arxiv.org/abs/2505.12370","title":"Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learning","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["android"],"id":"arxiv:2505.12493","kind":"paper","record_sha256":"4a65de1fee627687f195253e9cd4d1be843bd91d4078bf351796a82a07ac2385","scope":"core","source_url":"https://arxiv.org/abs/2505.12493","title":"GUI-Shift: Enhancing VLM-Based GUI Agents through Self-supervised Reinforcement Learning","topics":["reinforcement-learning","data-generation","grounding","mobile-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2505.12632","kind":"paper","record_sha256":"5b7f2dca02ca5b987ba14f234c923a858d6fa4c7529d0c7e4a02b1acb0066fc9","scope":"core","source_url":"https://arxiv.org/abs/2505.12632","title":"Scalable Video-to-Dataset Generation for Cross-Platform Mobile Agents","topics":["video","data-generation","mobile-navigation"],"year":"2025"},{"environments":["desktop","web","cross-platform"],"id":"arxiv:2505.12842","kind":"paper","record_sha256":"bb9363f1dea5d4f608faf5a46e50a0d28c279eb0266a9f9dd2845f96e85cfc26","scope":"core","source_url":"https://arxiv.org/abs/2505.12842","title":"GEM: Gaussian Embedding Modeling for Out-of-Distribution Detection in GUI Agents","topics":["robustness","safety-security","evaluation","efficiency"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2505.13227","kind":"paper","record_sha256":"9afabb0ca3d3abf2f42accbfc98f253aa0e63e275061164f464d00cb75d87dad","scope":"core","source_url":"https://arxiv.org/abs/2505.13227","title":"Scaling Computer-Use Grounding via User Interface Decomposition and Synthesis","topics":["grounding","data-generation","evaluation"],"year":"2025"},{"environments":["windows","desktop"],"id":"arxiv:2505.13909","kind":"paper","record_sha256":"49b35095ecc5d2f8dd91ca9582075e3f0e0d993a24a8a8dbaaca960bd15ff18b","scope":"core","source_url":"https://arxiv.org/abs/2505.13909","title":"Efficient Agent Training for Computer Use","topics":["data-generation","evaluation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2505.14141","kind":"paper","record_sha256":"7a309dacadbc0949d49963879091dd70996691caec04a762eaa1bd6cd53e4dc6","scope":"core","source_url":"https://arxiv.org/abs/2505.14141","title":"Building a Stable Planner: An Extended Finite State Machine Based Planning Module for Mobile GUI Agent","topics":["planning","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.14289","kind":"paper","record_sha256":"d8080ad2dde2681fcf41aac3651e4b852d140e47e839b2c0fe2bcb11258542a0","scope":"core","source_url":"https://arxiv.org/abs/2505.14289","title":"EVA: Evolving Semantic Adversaries for Red-Teaming GUI Agents Against Environmental Injection Attacks","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.14418","kind":"paper","record_sha256":"3244437c25e3ec7deeb1037424a8171e2e6d69934411d749a7f947a37e79d878","scope":"core","source_url":"https://arxiv.org/abs/2505.14418","title":"Hidden Ghost Hand: Unveiling Backdoor Vulnerabilities in MLLM-Powered Mobile GUI Agents","topics":["safety-security","mobile-navigation","interaction-history"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.15259","kind":"paper","record_sha256":"d1d1e99adb6931d0a944dfe688ea7ada77f190817f34b34c263f6913824c421a","scope":"core","source_url":"https://arxiv.org/abs/2505.15259","title":"ReGUIDE: Data Efficient GUI Grounding via Spatial Reasoning and Search","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.15277","kind":"paper","record_sha256":"15a9845aa3eba000d93c895bdc1a1f6f8944ee43b6c4278170a2a2ae8810dd37","scope":"core","source_url":"https://arxiv.org/abs/2505.15277","title":"Web-Shepherd: Advancing PRMs for Reinforcing Web Agents","topics":["reinforcement-learning","evaluation","data-generation","web-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2505.15810","kind":"paper","record_sha256":"f0e0a722bcacb919f1b866b74da0e6ffcd1f57617be3749cbcbcdb80e979aff3","scope":"core","source_url":"https://arxiv.org/abs/2505.15810","title":"GUI-G1: Understanding R1-Zero-Like Training for Visual Grounding in GUI Agents","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2505.16282","kind":"paper","record_sha256":"1054e9d0e49ec587d4030afc8c64043918a3d36c8b6afadc3a967fef7f5f6b22","scope":"core","source_url":"https://arxiv.org/abs/2505.16282","title":"ARPO:End-to-End Policy Optimization for GUI Agents with Experience Replay","topics":["reinforcement-learning","long-horizon"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.16421","kind":"paper","record_sha256":"579ee06847958d6da95b0d4a667af2e96cc4e4d290b8e3880e4220060a652271","scope":"core","source_url":"https://arxiv.org/abs/2505.16421","title":"WebAgent-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning","topics":["reinforcement-learning","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.16422","kind":"paper","record_sha256":"7bf528ba727d79509967fcf48046b53504d2ea7fae1aca2427db8c8a862e350f","scope":"core","source_url":"https://arxiv.org/abs/2505.16422","title":"Unlocking Smarter Device Control: Foresighted Planning with a World Model-Driven Code Execution Approach","topics":["world-models","planning","mobile-navigation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2505.16827","kind":"paper","record_sha256":"ae22d9e5e956a8a299bfbe3536d545528ae64ec05e5a04456df1730fd64fff9e","scope":"core","source_url":"https://arxiv.org/abs/2505.16827","title":"GUI-explorer: Autonomous Exploration and Mining of Transition-aware Knowledge for GUI Agent","topics":["mobile-navigation","data-generation","memory","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.18121","kind":"paper","record_sha256":"7af9f8ffbe33c03af66353f220ca0fe76ee3b42ff45b9c7560e56176361c77b6","scope":"core","source_url":"https://arxiv.org/abs/2505.18121","title":"ProgRM: Build Better GUI Agents with Progress Rewards","topics":["reinforcement-learning","data-generation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2505.18829","kind":"paper","record_sha256":"d99b7d3459c59f8d10ed46162aed76c2937e9bbd398f862f24c12335d3b439e6","scope":"core","source_url":"https://arxiv.org/abs/2505.18829","title":"LiteCUA: Computer as MCP Server for Computer-Use Agent on AIOS","topics":["tool-integration","automation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.19095","kind":"paper","record_sha256":"f8a4a5b34d23ded0564ab5f8f6d850fa9f8da9a030c50138915f2efaa307c6eb","scope":"core","source_url":"https://arxiv.org/abs/2505.19095","title":"ScreenExplorer: Training a Vision-Language Model for Diverse Exploration in Open GUI World","topics":["reinforcement-learning","world-models","data-generation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2505.19897","kind":"paper","record_sha256":"489cb27e2a78bbb3dfeaf2d587b1161907c9ec779ae2f5f2bf503636a96a43f1","scope":"core","source_url":"https://arxiv.org/abs/2505.19897","title":"ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows","topics":["evaluation","science","workflow"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.20013","kind":"paper","record_sha256":"3357c9c74f03b297adbfa1f51578208538bc611cfc3cd32a7c2eef81ee871b02","scope":"core","source_url":"https://arxiv.org/abs/2505.20013","title":"WebCoT: Enhancing Web Agent Reasoning by Reconstructing Chain-of-Thought in Reflection, Branching, and Rollback","topics":["data-generation","planning","self-reflection","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.20660","kind":"paper","record_sha256":"05ed7fcea32d7233dd5acb838047d775ab3f18e9e564d2b40f743b9fedb0484b","scope":"core","source_url":"https://arxiv.org/abs/2505.20660","title":"BacktrackAgent: Enhancing GUI Agent with Error Detection and Backtracking Mechanism","topics":["robustness","planning","data-generation","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.21279","kind":"paper","record_sha256":"028930646ce0359b958b46945a5a41da6da25560f826e7c39adadebc9ed192be","scope":"core","source_url":"https://arxiv.org/abs/2505.21279","title":"XBOUND: Exploring Capability Boundaries of Device-Control Agents at the State Level","topics":["evaluation","grounding"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2505.21496","kind":"paper","record_sha256":"ebc5c49a1b80c5aeb007b0365af8670f7f903e177db24e298cae32290b331434","scope":"core","source_url":"https://arxiv.org/abs/2505.21496","title":"UI-Genie: A Self-Improving Approach for Iteratively Boosting MLLM-based Mobile GUI Agents","topics":["data-generation","reinforcement-learning","mobile-navigation","interaction-history"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.21499","kind":"paper","record_sha256":"c6f90b9bd7c4ff9745cc7e1146f6b5ffa0ac8acdc638bc0387834e1502b470d6","scope":"core","source_url":"https://arxiv.org/abs/2505.21499","title":"AdInject: Real-World Black-Box Attacks on Web Agents via Advertising Delivery","topics":["safety-security","web-navigation","evaluation"],"year":"2025"},{"environments":["web","desktop"],"id":"arxiv:2505.21936","kind":"paper","record_sha256":"16021cfeb7dbd621979891c9bd105a99df04b538a87bfff470abd68f32f254b4","scope":"core","source_url":"https://arxiv.org/abs/2505.21936","title":"RedTeamCUA: Realistic Adversarial Testing of Computer-Use Agents in Hybrid Web-OS Environments","topics":["safety-security","evaluation","sandbox"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2505.21964","kind":"paper","record_sha256":"7d1142f416fa34a3b962ce73268560d7d9d4765d60bbb1dd76f6b77d7be5743b","scope":"core","source_url":"https://arxiv.org/abs/2505.21964","title":"UI-Evol: Automatic Knowledge Evolving for Computer Use Agents","topics":["memory","self-reflection","robustness"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.22942","kind":"paper","record_sha256":"4c6cea6f63fca79dcf57c94b94f7f6560893f5a7028fe8fc4f2d3ac16158ee92","scope":"core","source_url":"https://arxiv.org/abs/2505.22942","title":"WorkForceAgent-R1: Incentivizing Reasoning Capability in LLM-based Web Agents via Reinforcement Learning","topics":["reinforcement-learning","workflow"],"year":"2025"},{"environments":["android"],"id":"arxiv:2505.23596","kind":"paper","record_sha256":"63f80aad44c44b2c4da5c95a72b8da4f00e49bd3364b1608326970399d96c60c","scope":"core","source_url":"https://arxiv.org/abs/2505.23596","title":"Agent-SAMA: State-Aware Mobile Assistant","topics":["planning","memory","mobile-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2505.23762","kind":"paper","record_sha256":"21b970772e111e9ea74824a39614f651dcb22f0c29a021d63e4c339033e884be","scope":"core","source_url":"https://arxiv.org/abs/2505.23762","title":"ZeroGUI: Automating Online GUI Learning at Zero Human Cost","topics":["reinforcement-learning","data-generation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2505.24838","kind":"paper","record_sha256":"db09f83f45be63c4ad68a3ee35b5998cee42a9016086f381e53342cf6e133b59","scope":"core","source_url":"https://arxiv.org/abs/2505.24838","title":"VideoCAD: A Dataset and Model for Learning Long-Horizon 3D CAD UI Interactions from Video","topics":["video","data-generation","science","long-horizon"],"year":"2025"},{"environments":["web"],"id":"arxiv:2505.24878","kind":"paper","record_sha256":"c42f79eac3e2cafdecae61f00ee8d8daefe080d723b88660819c61fadfceb441","scope":"core","source_url":"https://arxiv.org/abs/2505.24878","title":"Open CaptchaWorld: A Comprehensive Web-based Platform for Testing and Benchmarking Multimodal LLM Agents","topics":["evaluation","web-navigation","screen-understanding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.00618","kind":"paper","record_sha256":"313f7855c5ca9b20d6de8caa2c077a4819418380d7005371147801f6a15d535c","scope":"core","source_url":"https://arxiv.org/abs/2506.00618","title":"RiOSWorld: Benchmarking the Risk of Multimodal Computer-Use Agents","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2506.01391","kind":"paper","record_sha256":"87b956b3a31e410e687fd0329171648576009eaef38b92df8380e3d59cf2b1e4","scope":"core","source_url":"https://arxiv.org/abs/2506.01391","title":"AgentCPM-GUI: Building Mobile-Use Agents with Reinforcement Fine-Tuning","topics":["reinforcement-learning","mobile-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.02456","kind":"paper","record_sha256":"d56778f5d2d46aa15721b3790a5367543cf7d8314ad8c21d7dd167c2eca55d43","scope":"core","source_url":"https://arxiv.org/abs/2506.02456","title":"VPI-Bench: Visual Prompt Injection Attacks for Computer-Use Agents","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2506.02539","kind":"paper","record_sha256":"46279538c725dccf3a4364ed19b8fb1756d03f23989927ea1b6ffeda9b24708f","scope":"core","source_url":"https://arxiv.org/abs/2506.02539","title":"VerificAgent: Domain-Specific Memory Verification for Scalable Oversight of Aligned Computer-Use Agents","topics":["memory","human-in-the-loop","safety-security"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.02839","kind":"paper","record_sha256":"aaa6096bdef795b2e574737924890ee70f197d68dbde6df2f0e948c05ed75619","scope":"core","source_url":"https://arxiv.org/abs/2506.02839","title":"DeepShop: A Benchmark for Deep Research Shopping Agents","topics":["evaluation","web-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.02865","kind":"paper","record_sha256":"c53eac2f066e81e9bb9044ed36f8163b824f77aa256acbb2a5aa4e1966b6f562","scope":"core","source_url":"https://arxiv.org/abs/2506.02865","title":"Surfer-H Meets Holo1: Cost-Efficient Web Agent Powered by Open Weights","topics":["web-navigation","grounding","data-generation","evaluation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2506.03095","kind":"paper","record_sha256":"7ee0c358fe3219fb8959df5d6e40a723a932e2d35f919d148b6a7e9b8790baf2","scope":"core","source_url":"https://arxiv.org/abs/2506.03095","title":"DPO Learning with LLMs-Judge Signal for Computer Use Agents","topics":["data-generation","evaluation","efficiency"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2506.03143","kind":"paper","record_sha256":"57b7634d9027251287097211a7d45dc1cc365d996b759c449b515c2f0270272c","scope":"core","source_url":"https://arxiv.org/abs/2506.03143","title":"GUI-Actor: Coordinate-Free Visual Grounding for GUI Agents","topics":["grounding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.03533","kind":"paper","record_sha256":"71c5c6b5dcf30027f709226746ad1361b0805f238e8c09864f9089f0b79f3d0a","scope":"core","source_url":"https://arxiv.org/abs/2506.03533","title":"Go-Browse: Training Web Agents with Structured Exploration","topics":["data-generation","web-navigation"],"year":"2025"},{"environments":["macos","desktop"],"id":"arxiv:2506.04135","kind":"paper","record_sha256":"c7073edbb095fedb66e63561aec498ec7c1df3376107e8cad0541320c659058c","scope":"core","source_url":"https://arxiv.org/abs/2506.04135","title":"macOSWorld: A Multilingual Interactive Benchmark for GUI Agents","topics":["evaluation","workflow","safety-security"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2506.04614","kind":"paper","record_sha256":"c0ae939a02d5eb2286070b21cd18585743546f062189e0c622f99103c466476c","scope":"core","source_url":"https://arxiv.org/abs/2506.04614","title":"Look Before You Leap: A GUI-Critic-R1 Model for Pre-Operative Error Diagnosis in GUI Automation","topics":["planning","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.05079","kind":"paper","record_sha256":"8759026049e9b6cdf6124f92bcdb6d8d6d7dc6a6b468b633774da3038c0cbe49","scope":"core","source_url":"https://arxiv.org/abs/2506.05079","title":"Scenario-Guided LLM-based Mobile App GUI Testing","topics":["automation","mobile-navigation","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.07217","kind":"paper","record_sha256":"4fa8db62fd2683ad8fb37580c1b7da93d2bb96413f879c59c0f1963ebaa9a77c","scope":"core","source_url":"https://arxiv.org/abs/2506.07217","title":"BIMgent: Towards Autonomous Building Modeling via Computer-use Agents","topics":["industry","planning","workflow","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.07385","kind":"paper","record_sha256":"099c09556e6ed5adf74cdf3c2fd1453e6e7bb1956eec37bdac3773fa6b163600","scope":"core","source_url":"https://arxiv.org/abs/2506.07385","title":"GUIPilot: A Consistency-based Mobile GUI Testing Approach for Detecting Application-specific Bugs","topics":["evaluation","mobile-navigation","robustness"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2506.07672","kind":"paper","record_sha256":"f0091353d4911e325da66191ee2af7f280671e0c948a81ba4236a23bb2bb0d1d","scope":"core","source_url":"https://arxiv.org/abs/2506.07672","title":"MCPWorld: A Unified Benchmarking Testbed for API, GUI, and Hybrid Computer Use Agents","topics":["evaluation","tool-integration","sandbox"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.07976","kind":"paper","record_sha256":"3198b48c9eeb240d2021aead9c238c49e2ee16f266d553bf901b022f160dfc4d","scope":"core","source_url":"https://arxiv.org/abs/2506.07976","title":"Thinking vs. Doing: Agents that Reason by Scaling Test-Time Interaction","topics":["reinforcement-learning","web-navigation","planning","long-horizon"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.08012","kind":"paper","record_sha256":"fa4484f3d74322c98c7895aa9cb486c29d1bd44b7e49c15e8d1a6fffb66269e4","scope":"core","source_url":"https://arxiv.org/abs/2506.08012","title":"GUI-Reflection: Empowering Multimodal GUI Models with Self-Reflection Behavior","topics":["robustness","data-generation","evaluation","mobile-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.08136","kind":"paper","record_sha256":"fd79794a447a6379032cfb03dcc6f25141524f14dd223bf9af96bcab13331076","scope":"core","source_url":"https://arxiv.org/abs/2506.08136","title":"EconWebArena: Benchmarking Autonomous Agents on Economic Tasks in Realistic Web Environments","topics":["evaluation","web-navigation","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.08972","kind":"paper","record_sha256":"1b1970245125087eda8d079ee98c5b86c335c916527265d3b61281fa0cf4f366","scope":"core","source_url":"https://arxiv.org/abs/2506.08972","title":"Atomic-to-Compositional Generalization for Mobile Agents with A New Benchmark and Scheduling System","topics":["evaluation","long-horizon","planning","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.09373","kind":"paper","record_sha256":"71ac154dd39b347f534ea28ce176cbaf257dd563110b0f66d9048732587824d5","scope":"core","source_url":"https://arxiv.org/abs/2506.09373","title":"LPO: Towards Accurate GUI Agent Interaction via Location Preference Optimization","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["android","web","cross-platform"],"id":"arxiv:2506.10387","kind":"paper","record_sha256":"2c33fbd0eb15469545ee51d0bf0dfcd9b1544de563a97933fb8a0b3aaffdcfd6","scope":"core","source_url":"https://arxiv.org/abs/2506.10387","title":"Mirage-1: Augmenting and Updating GUI Agent with Hierarchical Multimodal Skills","topics":["planning","memory","long-horizon","mobile-navigation","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.11127","kind":"paper","record_sha256":"94b764d8df598944a510d5ad9647544f4da2a3f95867c8af87ebdce85929e1d7","scope":"core","source_url":"https://arxiv.org/abs/2506.11127","title":"UITron-Speech: Towards Automated GUI Agents Based on Speech Instructions","topics":["grounding","action-prediction","data-generation"],"year":"2025"},{"environments":["desktop","web"],"id":"arxiv:2506.14205","kind":"paper","record_sha256":"cfb74bbe80378fcf5418272160ade3ec0ca3936292b28eda63294a62faef476d","scope":"core","source_url":"https://arxiv.org/abs/2506.14205","title":"AgentSynth: Scalable Task Generation for Generalist Computer-Use Agents","topics":["data-generation","evaluation","long-horizon"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2506.14477","kind":"paper","record_sha256":"c1a709329580857c120cd0b153ca4ce50a6139d906a7743408938d66ebe47b98","scope":"core","source_url":"https://arxiv.org/abs/2506.14477","title":"GUI-Robust: A Comprehensive Dataset for Testing GUI Agent Robustness in Real-World Anomalies","topics":["robustness","evaluation","data-generation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2506.14866","kind":"paper","record_sha256":"eb0c9c51f53899966b7e4dad2822122969ddee7ed63aedc01a8bba5678979d1a","scope":"core","source_url":"https://arxiv.org/abs/2506.14866","title":"OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.15425","kind":"paper","record_sha256":"ff7d0961f0a511e0122c49fd83ddcf055da62956d6213b1df03803577115a1e9","scope":"core","source_url":"https://arxiv.org/abs/2506.15425","title":"Understanding GUI Agent Localization Biases through Logit Sharpness","topics":["grounding","evaluation","robustness","training-free"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.15677","kind":"paper","record_sha256":"d8ca294202ba2fe5c671bc2b5cb222f5339f6199744585562d7554839a0bbbd5","scope":"core","source_url":"https://arxiv.org/abs/2506.15677","title":"Embodied Web Agents: Bridging Physical-Digital Realms for Integrated Agent Intelligence","topics":["evaluation","web-navigation","planning"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2506.16042","kind":"paper","record_sha256":"69f65313433ab315c30d59b3ef7ef9806bf976e2a543a8900d2ae0b1ae449cbc","scope":"core","source_url":"https://arxiv.org/abs/2506.16042","title":"OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agents","topics":["evaluation","efficiency","observability"],"year":"2025"},{"environments":["web"],"id":"arxiv:2506.17318","kind":"paper","record_sha256":"939d9723f402d56b9fb13153f53aa03e45e54c1697f02712d0b3e52a354576c8","scope":"core","source_url":"https://arxiv.org/abs/2506.17318","title":"Context manipulation attacks : Web agents are susceptible to corrupted memory","topics":["safety-security","memory","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.17913","kind":"paper","record_sha256":"3336a42b7223f7eb5cd31d483c6155fe4dfae5d3aec3880bc2c4835b605942f0","scope":"core","source_url":"https://arxiv.org/abs/2506.17913","title":"Learning, Reasoning, Refinement: A Framework for Kahneman's Dual-System Intelligence in GUI Agents","topics":["grounding","planning","reinforcement-learning","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2506.18158","kind":"paper","record_sha256":"b5a29f5580909366e15b52b5141f05ba2c9f588d0ddf8cd308460b9c850af2e1","scope":"core","source_url":"https://arxiv.org/abs/2506.18158","title":"Chain-of-Memory: Enhancing GUI Agents for Cross-Application Navigation","topics":["memory","state-tracking","cross-app","data-generation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2506.20332","kind":"paper","record_sha256":"73a4daef331d0999eb0be3b8ffd9e4cc46bbcb13473907fad6aeb6708ac94169","scope":"core","source_url":"https://arxiv.org/abs/2506.20332","title":"Mobile-R1: Towards Interactive Capability for VLM-Based Mobile Agent via Systematic Training","topics":["reinforcement-learning","mobile-navigation"],"year":"2025"},{"environments":["web","desktop","cross-platform"],"id":"arxiv:2506.23491","kind":"paper","record_sha256":"97906505000f76269896ca84036f70ec58fba6270f5f1078a7781c1c410c71e3","scope":"core","source_url":"https://arxiv.org/abs/2506.23491","title":"ZonUI-3B: A Lightweight Vision-Language Model for Cross-Resolution GUI Grounding","topics":["grounding","data-generation","efficiency"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2507.00008","kind":"paper","record_sha256":"8a188ed4cf1663da279124d1e9d687e3ccb97d3124bd7b39d13afabea06d7ef6","scope":"core","source_url":"https://arxiv.org/abs/2507.00008","title":"DiMo-GUI: Advancing Test-time Scaling in GUI Grounding via Modality-Aware Visual Reasoning","topics":["grounding","screen-understanding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.00210","kind":"paper","record_sha256":"a9c6a46a55d1c7a0cda7b810530dd9239c015ac6e1b30995cb675e0e182444e5","scope":"core","source_url":"https://arxiv.org/abs/2507.00210","title":"LineRetriever: Planning-Aware Observation Reduction for Web Agents","topics":["web-navigation","efficiency","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.00938","kind":"paper","record_sha256":"c7d420cd521766395283cf24366fbdb289196fc30d212b1a31e18c236fa5631d","scope":"core","source_url":"https://arxiv.org/abs/2507.00938","title":"WebArxiv: A Reproducible Benchmark for Evaluating Multimodal Web Agents on arXiv Tasks","topics":["evaluation","web-navigation","memory"],"year":"2025"},{"environments":["web","cross-platform"],"id":"arxiv:2507.03730","kind":"paper","record_sha256":"d6daec76c3db65bdbf33652acbe4465117af035a223e5c6fcf4db110a311a9bb","scope":"core","source_url":"https://arxiv.org/abs/2507.03730","title":"Less is More: Empowering GUI Agent with Context-Aware Simplification","topics":["efficiency","interaction-history","web-navigation","mobile-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.04103","kind":"paper","record_sha256":"7c596e89c72d2483192a44803cc560ba77c794021902a146429d853ececb386d","scope":"core","source_url":"https://arxiv.org/abs/2507.04103","title":"How to Train Your LLM Web Agent: A Statistical Diagnosis","topics":["imitation-learning","reinforcement-learning","efficiency","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2507.04227","kind":"paper","record_sha256":"89ef78a12c92a56ec377b544b9e600afce5a41d3cfdf911ab3adcb15ef22f5ba","scope":"core","source_url":"https://arxiv.org/abs/2507.04227","title":"Mobile GUI Agents under Real-world Threats: Are We There Yet?","topics":["mobile-navigation","safety-security","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.05445","kind":"paper","record_sha256":"6c5c01488edb4dd8e26e2fec6ac534cc36d888f81522c03e42bcbe3f2cef9dd2","scope":"core","source_url":"https://arxiv.org/abs/2507.05445","title":"A Systematization of Security Vulnerabilities in Computer Use Agents","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["web","cross-platform"],"id":"arxiv:2507.05673","kind":"paper","record_sha256":"1e395e03fb3519b8787106a3aad47695c46dcf1b80d90b6abcdfa46721607277","scope":"core","source_url":"https://arxiv.org/abs/2507.05673","title":"R-VLM: Region-Aware Vision Language Model for Precise GUI Grounding","topics":["grounding","screen-understanding","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2507.05720","kind":"paper","record_sha256":"eef8efaefbf47236cf0781ee3b77f94b0b139cf054b1eccea9d0ac3b92aed1df","scope":"core","source_url":"https://arxiv.org/abs/2507.05720","title":"MobileGUI-RL: Advancing Mobile GUI Agent through Reinforcement Learning in Online Environment","topics":["mobile-navigation","reinforcement-learning","data-generation","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2507.05791","kind":"paper","record_sha256":"96ddfa9591fd329724e448a6c516eb9a8fd893daa538cc03fd38825bdffe56eb","scope":"core","source_url":"https://arxiv.org/abs/2507.05791","title":"GTA1: GUI Test-time Scaling Agent","topics":["planning","grounding","reinforcement-learning"],"year":"2025"},{"environments":["web","desktop"],"id":"arxiv:2507.06134","kind":"paper","record_sha256":"ada9212477b4bfb6656aee266ec16e650a1edd4ddee73d01a0001ef082b13c77","scope":"core","source_url":"https://arxiv.org/abs/2507.06134","title":"OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety","topics":["safety-security","evaluation"],"year":"2025"},{"environments":["web","desktop"],"id":"arxiv:2507.06899","kind":"paper","record_sha256":"b94dddd7f7e9aa8db53c99e1fbb097ef11c96f47ffcc98278110f0e2439a1f66","scope":"core","source_url":"https://arxiv.org/abs/2507.06899","title":"VisualTrap: A Stealthy Backdoor Attack on GUI Agents via Visual Grounding Manipulation","topics":["safety-security","grounding","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2507.10610","kind":"paper","record_sha256":"102f20c763d06959d45698c543867210375fa3f963dd496a1e392a7e653edd72","scope":"core","source_url":"https://arxiv.org/abs/2507.10610","title":"LaSM: Layer-wise Scaling Mechanism for Defending Pop-up Attack on GUI Agents","topics":["safety-security","grounding","training-free"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.12844","kind":"paper","record_sha256":"d1783dffa23036332c03db2f3806e6c1beae580dd7da591fc8e4bd41685fe23e","scope":"core","source_url":"https://arxiv.org/abs/2507.12844","title":"Machine-Readable Ads: Accessibility and Trust Patterns for AI Web Agents interacting with Online Advertisements","topics":["evaluation","ui-design","safety-security"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.14293","kind":"paper","record_sha256":"795dbcbab60402290870641d259b2483dc3c85e1853492f451224d8ac4e9d86b","scope":"core","source_url":"https://arxiv.org/abs/2507.14293","title":"WebGuard: Building a Generalizable Guardrail for Web Agents","topics":["safety-security","reward-models","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.14799","kind":"paper","record_sha256":"27e31ae3012ea91218f333574aa78c5c4fc12c6adf4052c960926f256847faaf","scope":"core","source_url":"https://arxiv.org/abs/2507.14799","title":"Manipulating LLM Web Agents with Indirect Prompt Injection Attack via HTML Accessibility Tree","topics":["safety-security","web-navigation","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2507.15846","kind":"paper","record_sha256":"4a19568e984ee31904b28e28ef006df27304d93c741e0625f439d4badfbf99bd","scope":"core","source_url":"https://arxiv.org/abs/2507.15846","title":"GUI-G$^2$: Gaussian Reward Modeling for GUI Grounding","topics":["grounding","reinforcement-learning","reward-models"],"year":"2025"},{"environments":["android"],"id":"arxiv:2507.16853","kind":"paper","record_sha256":"d52c14ad82df1014f748e3e9c9b5ebb621d83b17398835a0bc309e8de3800fdf","scope":"core","source_url":"https://arxiv.org/abs/2507.16853","title":"MobileUse: A GUI Agent with Hierarchical Reflection for Autonomous Mobile Operation","topics":["mobile-navigation","self-reflection","robustness","long-horizon"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2507.19132","kind":"paper","record_sha256":"6e9588db30f95b266de663ce24004fb696fddc28c97ffa13b6c1ffda47c17706","scope":"core","source_url":"https://arxiv.org/abs/2507.19132","title":"OS-MAP: How Far Can Computer-Using Agents Go in Breadth and Depth?","topics":["evaluation","automation","planning"],"year":"2025"},{"environments":["web","desktop","windows","macos","linux","ios","android","cross-platform"],"id":"arxiv:2507.19478","kind":"paper","record_sha256":"bfd6e0880753889fe23a4312ad37c1fb91e9df98c5f68bea1ae51ba48e077fca","scope":"core","source_url":"https://arxiv.org/abs/2507.19478","title":"MMBench-GUI: Hierarchical Multi-Platform Evaluation Framework for GUI Agents","topics":["evaluation","grounding","efficiency","planning"],"year":"2025"},{"environments":["android"],"id":"arxiv:2507.21071","kind":"paper","record_sha256":"3a27cf139b678702be9244633cb4248ecead5ea8256cc33d4f2e6ce00b297590","scope":"core","source_url":"https://arxiv.org/abs/2507.21071","title":"FingerTip 20K: A Benchmark for Proactive and Personalized Mobile LLM Agents","topics":["evaluation","data-generation","mobile-navigation","interaction-history"],"year":"2025"},{"environments":["android"],"id":"arxiv:2507.21953","kind":"paper","record_sha256":"e2e3975f8a0a4ad19533711082a3e145bde1f8a58daf3d4ef144dbe43b97ceaf","scope":"core","source_url":"https://arxiv.org/abs/2507.21953","title":"MapAgent: Trajectory-Constructed Memory-Augmented Planning for Mobile Task Automation","topics":["memory","planning","mobile-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2507.22025","kind":"paper","record_sha256":"304e1ed85bc6b5559f33505cb4078ca9e43902bb8445a62b68c59994537b4e88","scope":"core","source_url":"https://arxiv.org/abs/2507.22025","title":"UI-AGILE: Advancing GUI Agents with Effective Reinforcement Learning and Precise Inference-Time Grounding","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2507.22358","kind":"paper","record_sha256":"c4ab9c5cc2996e4ad3d0573332334f58f9adb9bfb297fba1dbdfe5c2e9aa8896","scope":"core","source_url":"https://arxiv.org/abs/2507.22358","title":"Magentic-UI: Towards Human-in-the-loop Agentic Systems","topics":["human-in-the-loop","tool-integration","web-navigation","safety-security"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2507.23779","kind":"paper","record_sha256":"095f0c2be772111414437d627e5c0e59ea2c0ee7ff20f4f6820b06e51a99ba1f","scope":"core","source_url":"https://arxiv.org/abs/2507.23779","title":"Phi-Ground Tech Report: Advancing Perception in GUI Grounding","topics":["grounding","screen-understanding"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.01330","kind":"paper","record_sha256":"1faf0d0ad855c05805d643bdfd8b05df64ed1cb55556abcc46016eebd5efc223","scope":"core","source_url":"https://arxiv.org/abs/2508.01330","title":"NaturalGAIA: A Verifiable Benchmark and Hierarchical Framework for Long-Horizon GUI Tasks","topics":["evaluation","long-horizon","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2508.01858","kind":"paper","record_sha256":"b79df891e16a746a38c6f54231ef2e1768427db2f7ac5c19c57df6d29a7094ee","scope":"core","source_url":"https://arxiv.org/abs/2508.01858","title":"Web-CogReasoner: Towards Multimodal Knowledge-Induced Cognitive Reasoning for Web Agents","topics":["data-generation","planning","evaluation","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.02046","kind":"paper","record_sha256":"ac143a44374723bf188d194ec8c865800720b203ef36e1b9dd847d36fabdb846","scope":"core","source_url":"https://arxiv.org/abs/2508.02046","title":"NaviMaster: Learning a Unified Policy for GUI and Embodied Navigation Tasks","topics":["reinforcement-learning","data-generation","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.03700","kind":"paper","record_sha256":"4090ed182ea51b6d3a4319363c24e794728bcbf0713970ce85f6ffa34bcfcda2","scope":"core","source_url":"https://arxiv.org/abs/2508.03700","title":"MagicGUI: A Foundational Mobile GUI Agent with Scalable Data Pipeline and Reinforcement Fine-tuning","topics":["mobile-navigation","data-generation","reinforcement-learning"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2508.03923","kind":"paper","record_sha256":"8bf64a16e960663c8e57073625eadb34ee00f36ef574eab08699507a2e298008","scope":"core","source_url":"https://arxiv.org/abs/2508.03923","title":"CoAct-1: Computer-using Multi-Agent System with Coding Actions","topics":["tool-integration","automation","efficiency","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.04025","kind":"paper","record_sha256":"37944b3db67203a5d17f7548003d53fda622a1d109a9fe7cdb32017d2250ba91","scope":"core","source_url":"https://arxiv.org/abs/2508.04025","title":"Uncertainty-Aware GUI Agent: Adaptive Perception through Component Recommendation and Human-in-the-Loop Refinement","topics":["mobile-navigation","human-in-the-loop","screen-understanding","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.04037","kind":"paper","record_sha256":"90a4e794da5923c59bb8d593a8b3bc17301a86d90ba6451393f06eb9bf5250d3","scope":"core","source_url":"https://arxiv.org/abs/2508.04037","title":"Evolving in Tasks: Empowering the Multi-modality Large Language Model as the Computer Use Agent","topics":["data-generation","reinforcement-learning","grounding","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.04389","kind":"paper","record_sha256":"87a6febaef63f71fa0a1104e09b5a78e23490982ad728e89be155b933e16872d","scope":"core","source_url":"https://arxiv.org/abs/2508.04389","title":"GuirlVG: Incentivize GUI Visual Grounding via Empirical Exploration on Reinforcement Learning","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2508.04412","kind":"paper","record_sha256":"fa3a1ca859f3083aa04251900afec7d7d4ef989b9fa558df902b6b3bfdf3fdb0","scope":"core","source_url":"https://arxiv.org/abs/2508.04412","title":"Beyond Pixels: Exploring DOM Downsampling for LLM-Based Web Agents","topics":["web-navigation","efficiency","screen-understanding"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2508.04482","kind":"survey","record_sha256":"b907b7afc9db647792eae44efda64e25515bca9a7856bf6ff785144ab5a92164","scope":"core","source_url":"https://arxiv.org/abs/2508.04482","title":"OS Agents: A Survey on MLLM-based Agents for General Computing Devices Use","topics":["survey"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2508.04700","kind":"paper","record_sha256":"d68e97663906a6dfd8a5e216f96a9d8d58b227be41581d7ab4d08bf4a9aeb03b","scope":"core","source_url":"https://arxiv.org/abs/2508.04700","title":"SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience","topics":["reinforcement-learning","data-generation","world-models","imitation-learning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2508.05081","kind":"paper","record_sha256":"020ecc2e4b9332ac783904bc344815be50b74c8d1ebc48a2458f6ce304f915ca","scope":"core","source_url":"https://arxiv.org/abs/2508.05081","title":"Cognitive Duality for Adaptive Web Agents","topics":["planning","efficiency","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.05615","kind":"paper","record_sha256":"d30865b1a7325e558ffd848fc8bb233e61b0f0f3188a702cdd04e43828590c14","scope":"core","source_url":"https://arxiv.org/abs/2508.05615","title":"Test-Time Reinforcement Learning for GUI Grounding via Region Consistency","topics":["grounding","reinforcement-learning","training-free"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2508.05731","kind":"paper","record_sha256":"1024c881c5bcd42aed82f7164a83d0ff8aa7118d616e490f1733394aca5952d2","scope":"core","source_url":"https://arxiv.org/abs/2508.05731","title":"InfiGUI-G1: Advancing GUI Grounding with Adaptive Exploration Policy Optimization","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.09057","kind":"paper","record_sha256":"405e24eb5f164cc9bcaa351eed594e48d5390d851e8dddb0bdd8c2107cbd07c3","scope":"core","source_url":"https://arxiv.org/abs/2508.09057","title":"MVISU-Bench: Benchmarking Mobile Agents for Real-World Tasks by Multi-App, Vague, Interactive, Single-App and Unethical Instructions","topics":["evaluation","mobile-navigation","human-in-the-loop","safety-security"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2508.09123","kind":"paper","record_sha256":"0049b7910b34e349f5dab42c0151007a13f8fba2ac60ac54ee0354749dd5e484","scope":"core","source_url":"https://arxiv.org/abs/2508.09123","title":"OpenCUA: Open Foundations for Computer-Use Agents","topics":["data-generation","planning"],"year":"2025"},{"environments":["desktop","web","cross-platform"],"id":"arxiv:2508.09241","kind":"paper","record_sha256":"299dc1c98645b4727f02abab278bcb26392831814e7c899c831d270a0aaa1701","scope":"core","source_url":"https://arxiv.org/abs/2508.09241","title":"FineState-Bench: A Comprehensive Benchmark for Fine-Grained State Control in GUI Agents","topics":["evaluation","grounding","observability","screen-understanding"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2508.10833","kind":"paper","record_sha256":"4405fb0d971fe352c34657b22f1279fd80cd052dc7cf78f7cc1047b6c947087c","scope":"core","source_url":"https://arxiv.org/abs/2508.10833","title":"UI-Venus Technical Report: Building High-performance UI Agents with RFT","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["android"],"id":"arxiv:2508.11360","kind":"paper","record_sha256":"87242c1b590fdbe5f76695db9cddf32b9f281eecb264d5c6e9c3a09928ee5f6b","scope":"core","source_url":"https://arxiv.org/abs/2508.11360","title":"CRAFT-GUI: Curriculum-Reinforced Agent For GUI Tasks","topics":["reinforcement-learning","reward-models","mobile-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2508.13024","kind":"paper","record_sha256":"9e582f52239a63ffce28e8a75790d7975a7bc166f8aa8e928475ee0dcd37a7ea","scope":"core","source_url":"https://arxiv.org/abs/2508.13024","title":"WebMall -- A Multi-Shop Benchmark for Evaluating Web Agents","topics":["evaluation","web-navigation","sandbox"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.13634","kind":"paper","record_sha256":"a0f46a36dd64693bca1d553d7209115e378e2384631ec3119671aea094fed9c6","scope":"core","source_url":"https://arxiv.org/abs/2508.13634","title":"V2P: Visual Attention Calibration for GUI Grounding via Background Suppression and Center Peaking","topics":["grounding","screen-understanding"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2508.14040","kind":"paper","record_sha256":"50eea590972465317d7e0a5413b0f1ac3568b498a90a32596d7a1e0ea71bb52d","scope":"core","source_url":"https://arxiv.org/abs/2508.14040","title":"ComputerRL: Scaling End-to-End Online Reinforcement Learning for Computer Use Agents","topics":["reinforcement-learning","tool-integration"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.14104","kind":"paper","record_sha256":"eb0f0a37108dfe64d5799fd5573ef8028a9d378f3ce2b7b545bffda181f0cf0c","scope":"core","source_url":"https://arxiv.org/abs/2508.14104","title":"You Don't Know Until You Click:Automated GUI Testing for Production-Ready Software Evaluation","topics":["evaluation","automation","observability"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2508.15144","kind":"paper","record_sha256":"af0bca4b46a767e94fd430a26f43cd883bfbf6f9270673c8200830fb5b6b8912","scope":"core","source_url":"https://arxiv.org/abs/2508.15144","title":"Mobile-Agent-v3: Fundamental Agents for GUI Automation","topics":["data-generation","planning","mobile-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2508.15832","kind":"paper","record_sha256":"3709a4157fd2e08826ff63ac6027dc072f75696e271f925876edf23e4ed07cbe","scope":"core","source_url":"https://arxiv.org/abs/2508.15832","title":"A Functionality-Grounded Benchmark for Evaluating Web Agents in E-commerce Domains","topics":["evaluation","safety-security","data-generation","web-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2508.16987","kind":"paper","record_sha256":"d3931400be2ad81caecb45f029683415e4017c6bfb771cab0de8791517a68eae","scope":"core","source_url":"https://arxiv.org/abs/2508.16987","title":"WebSight: A Vision-First Architecture for Robust Web Agents","topics":["web-navigation","grounding","planning","memory"],"year":"2025"},{"environments":["android"],"id":"arxiv:2508.18689","kind":"paper","record_sha256":"e10f71d31e4f45d34ed77e21f6742b91320fbec45fac9017bed9db13929a26fe","scope":"core","source_url":"https://arxiv.org/abs/2508.18689","title":"AppAgent-Pro: A Proactive GUI Agent System for Multidomain Information Integration and User Assistance","topics":["planning","workflow"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2508.19679","kind":"paper","record_sha256":"6268a66efdf2b674d693a08cf786f54c5a673c149a8c30322764c69699097be8","scope":"core","source_url":"https://arxiv.org/abs/2508.19679","title":"InquireMobile: Teaching VLM-based Mobile Agent to Request Human Assistance via Reinforcement Fine-Tuning","topics":["human-in-the-loop","reinforcement-learning","evaluation","mobile-navigation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2508.20096","kind":"paper","record_sha256":"c13ce4a7e8be10ba6ccf9e94dab270bd0fa5f672f4494e75d81ac9578172c693","scope":"core","source_url":"https://arxiv.org/abs/2508.20096","title":"CODA: Coordinating the Cerebrum and Cerebellum for a Dual-Brain Computer Use Agent with Decoupled Reinforcement Learning","topics":["reinforcement-learning","planning","science","imitation-learning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2508.21456","kind":"paper","record_sha256":"bb8cc192f67a1625b4f025936a4b13aaafed54284dd69264fc6675eb8c99f8d9","scope":"core","source_url":"https://arxiv.org/abs/2508.21456","title":"Morae: Proactively Pausing UI Agents for User Choices","topics":["human-in-the-loop","web-navigation","evaluation"],"year":"2025"},{"environments":["desktop","cross-platform"],"id":"arxiv:2508.21767","kind":"paper","record_sha256":"855644c4afa032d1dc74ebbf95d990fbd6e3e7afa4e6b525fd90f76c02d12de0","scope":"core","source_url":"https://arxiv.org/abs/2508.21767","title":"UItron: Foundational GUI Agent with Advanced Perception and Planning","topics":["data-generation","reinforcement-learning","planning","grounding"],"year":"2025"},{"environments":["android","web","desktop","cross-platform"],"id":"arxiv:2509.00366","kind":"paper","record_sha256":"f27b78dd396aa808dfe540d1aef59a46e1cd0410c23e4cdc976ba81e96941e0b","scope":"core","source_url":"https://arxiv.org/abs/2509.00366","title":"KG-RAG: Enhancing GUI Agent Decision-Making via Knowledge Graph-Driven Retrieval-Augmented Generation","topics":["memory","planning","mobile-navigation","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.00531","kind":"paper","record_sha256":"c97f7da4c0a3463fe824e278f09761f4fc276515fd33f56aa332f2fcb20030ba","scope":"core","source_url":"https://arxiv.org/abs/2509.00531","title":"MobiAgent: A Systematic Framework for Customizable Mobile Agents","topics":["mobile-navigation","efficiency","data-generation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.01052","kind":"paper","record_sha256":"e76f4445da910a7d5e41cd270656c3510503f213f728b6664d8508ed94f0334b","scope":"adjacent","source_url":"https://arxiv.org/abs/2509.01052","title":"FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games","topics":["evaluation","long-horizon","memory","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2509.01619","kind":"paper","record_sha256":"def1b4457e22bd74123f8a1b00702f872196dd02f7ff29992d2f5f4f82711592","scope":"adjacent","source_url":"https://arxiv.org/abs/2509.01619","title":"Throttling Web Agents Using Reasoning Gates","topics":["safety-security","evaluation","tool-integration"],"year":"2025"},{"environments":["android"],"id":"arxiv:2509.01720","kind":"paper","record_sha256":"8914f1f0c5b3d74813392414de05d3ab83053159d1e05eef5293bbc2fadf0124","scope":"core","source_url":"https://arxiv.org/abs/2509.01720","title":"Succeed or Learn Slowly: Sample Efficient Off-Policy Reinforcement Learning for Mobile App Control","topics":["reinforcement-learning","mobile-navigation","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.02322","kind":"paper","record_sha256":"eccbe36a2c91c146c5e92ad308e48f5f106c7fc547e4d8973f1421894de41f29","scope":"core","source_url":"https://arxiv.org/abs/2509.02322","title":"OmniActor: A Generalist GUI and Embodied Agent for 2D&3D Worlds","topics":["action-prediction","data-generation","imitation-learning"],"year":"2025"},{"environments":["web","desktop","windows","android","cross-platform"],"id":"arxiv:2509.02544","kind":"paper","record_sha256":"6cca54f136affaa4bca32b5edf8d6d591ac9eb99fc9688c848b2b5d4c7f688f7","scope":"core","source_url":"https://arxiv.org/abs/2509.02544","title":"UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning","topics":["reinforcement-learning","data-generation","sandbox","tool-integration"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.03891","kind":"paper","record_sha256":"1fe82b01a5c731376b1dc478a570a324d1b679a0ff87500cf37c4a3066152a75","scope":"core","source_url":"https://arxiv.org/abs/2509.03891","title":"MobileRAG: Enhancing Mobile Agent with Retrieval-Augmented Generation","topics":["memory","mobile-navigation","evaluation","long-horizon"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.04243","kind":"paper","record_sha256":"9676bd46729af566de0e72d482d4e594b9c8e35e16686ee35f8a61d57a622d96","scope":"core","source_url":"https://arxiv.org/abs/2509.04243","title":"Learning Active Perception via Self-Evolving Preference Optimization for GUI Grounding","topics":["grounding","data-generation","screen-understanding"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.04908","kind":"paper","record_sha256":"3e59c5f6bc61a1dcf9a3932eb09ebeaca569bc43428f40741957a438057e2dc6","scope":"core","source_url":"https://arxiv.org/abs/2509.04908","title":"SparkUI-Parser: Enhancing GUI Perception with Robust Grounding and Parsing","topics":["grounding","screen-understanding","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.06477","kind":"paper","record_sha256":"ea45db7d0f759027fd97e3369fc198910723e3b09e389f0f11c6bed7b8b1c15c","scope":"core","source_url":"https://arxiv.org/abs/2509.06477","title":"MAS-Bench: A Unified Benchmark for Shortcut-Augmented Hybrid Mobile GUI Agents","topics":["evaluation","mobile-navigation","tool-integration","workflow"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.07553","kind":"paper","record_sha256":"5323e55bc00dfbc3a13cc075e975b8256b837066f3ddc67bf91515d5e4f3fe9c","scope":"core","source_url":"https://arxiv.org/abs/2509.07553","title":"VeriOS: Query-Driven Proactive Human-Agent-GUI Interaction for Trustworthy OS Agents","topics":["human-in-the-loop","safety-security","reinforcement-learning"],"year":"2025"},{"environments":["desktop","linux","web"],"id":"arxiv:2509.07764","kind":"paper","record_sha256":"b1d3e890ff7298930ea19b120325346c409d854ad78f7f38dcddd9f8246a97b5","scope":"core","source_url":"https://arxiv.org/abs/2509.07764","title":"AgentSentinel: An End-to-End and Real-Time Security Defense Framework for Computer-Use Agents","topics":["safety-security","observability","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.10723","kind":"paper","record_sha256":"75d51476bda734bff51538e26a70de72bf15632c33c637dc0db7c788d2f77776","scope":"core","source_url":"https://arxiv.org/abs/2509.10723","title":"Dark Patterns Meet GUI Agents: LLM Agent Susceptibility to Manipulative Interfaces and the Role of Human Oversight","topics":["human-in-the-loop","safety-security","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2509.11250","kind":"paper","record_sha256":"3816ec7bfe2fb97aae688cbb8ac8701b78f5d4a99805b3167210480c39a64794","scope":"core","source_url":"https://arxiv.org/abs/2509.11250","title":"Environmental Injection Attacks against GUI Agents in Realistic Dynamic Environments","topics":["safety-security","robustness","evaluation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2509.11543","kind":"paper","record_sha256":"715e9b5aa9f34959f0d43ce021e3d32b3d4fcd3dd3b916352ca45a7b136454a3","scope":"core","source_url":"https://arxiv.org/abs/2509.11543","title":"UI-S1: Advancing GUI Automation via Semi-online Reinforcement Learning","topics":["reinforcement-learning","evaluation","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.11548","kind":"paper","record_sha256":"2e6e1bc6897ad3aa4fdf0104f8b19b828e2f0ff3a453a7c69ca2e49eb247c5a9","scope":"core","source_url":"https://arxiv.org/abs/2509.11548","title":"How Auxiliary Reasoning Unleashes GUI Grounding in VLMs","topics":["grounding","training-free","screen-understanding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2509.11939","kind":"paper","record_sha256":"284f28ea28e44ee3c1a98f4466b9656db20831c328f07f3f5d4fd924e533272b","scope":"core","source_url":"https://arxiv.org/abs/2509.11939","title":"PrivWeb: Unobtrusive and Content-aware Privacy Protection For Web Agents","topics":["safety-security","human-in-the-loop","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.13615","kind":"paper","record_sha256":"374f49932f2b36a5c4f420d72bed1f580298d9f6654133c079d39b55def4d2bd","scope":"core","source_url":"https://arxiv.org/abs/2509.13615","title":"See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles","topics":["state-tracking","grounding","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.13704","kind":"paper","record_sha256":"fef7f1490d27fef2502d722fa136831846cb2d908a212d5d2d530c82053072a6","scope":"core","source_url":"https://arxiv.org/abs/2509.13704","title":"InfraMind: A Novel Exploration-based GUI Agentic Framework for Mission-critical Industrial Management","topics":["industry","planning","memory","safety-security"],"year":"2025"},{"environments":["web"],"id":"arxiv:2509.14172","kind":"paper","record_sha256":"19f9983d0b3b33c61709436c5a5499be343a6972a684c774aee0812985f1553f","scope":"core","source_url":"https://arxiv.org/abs/2509.14172","title":"TGPO: Tree-Guided Preference Optimization for Robust Web Agent Reinforcement Learning","topics":["reinforcement-learning","reward-models","web-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2509.14382","kind":"paper","record_sha256":"c29e2eeb1ed50432bdfae0cb0302c0c502f986b028b9f916051b77ddbc2f2ce8","scope":"core","source_url":"https://arxiv.org/abs/2509.14382","title":"Detecting Pipeline Failures through Fine-Grained Analysis of Web Agents","topics":["evaluation","observability","web-navigation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2509.15221","kind":"paper","record_sha256":"f59939b755decdfe86d34f2d49468ba9395282eb63346021094d846e07019d5f","scope":"core","source_url":"https://arxiv.org/abs/2509.15221","title":"ScaleCUA: Scaling Open-Source Computer Use Agents with Cross-Platform Data","topics":["data-generation","automation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.15532","kind":"paper","record_sha256":"16c423240a1b91bbec1009286aabb32c673fd87b7b1c8297db60208786a0a26a","scope":"core","source_url":"https://arxiv.org/abs/2509.15532","title":"GUI-ARP: Enhancing Grounding with Adaptive Region Perception for GUI Agents","topics":["grounding","reinforcement-learning","screen-understanding"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.15566","kind":"paper","record_sha256":"73cca12ef2d3eda9273be14e4f7ae4c1dd61cfcf30dd248272fb53913c66bb7a","scope":"core","source_url":"https://arxiv.org/abs/2509.15566","title":"BTL-UI: Blink-Think-Link Reasoning Model for GUI Agent","topics":["grounding","planning","data-generation","reinforcement-learning"],"year":"2025"},{"environments":["android"],"id":"arxiv:2509.15738","kind":"paper","record_sha256":"09fdcf1b934b3cb4fef3b56e0305ecd7cb0e1c9126e95655d4ab5760e248f6db","scope":"core","source_url":"https://arxiv.org/abs/2509.15738","title":"GUI-ReWalk: Massive Data Generation for GUI Agent via Stochastic Exploration and Intent-Aware Reasoning","topics":["data-generation","cross-app","long-horizon","planning"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2509.17328","kind":"paper","record_sha256":"72ed039e0762cdecc3006f6683b170a3b006591a650c951f0196f24e6fc4e171","scope":"core","source_url":"https://arxiv.org/abs/2509.17328","title":"UIPro: Unleashing Superior Interaction Capability For GUI Agents","topics":["data-generation","action-prediction","grounding","imitation-learning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.17917","kind":"paper","record_sha256":"f7ba1b2fb51bd6dc481656d791faa488d1afe9c6e6a3c5e01ec886a60641d81d","scope":"core","source_url":"https://arxiv.org/abs/2509.17917","title":"Orcust: Stepwise-Feedback Reinforcement Learning for GUI Agent","topics":["reinforcement-learning","reward-models","data-generation","sandbox"],"year":"2025"},{"environments":["android"],"id":"arxiv:2509.18119","kind":"paper","record_sha256":"3f02e90c5e3e7cf83efbd8deb39f553a4260dd17dedd44e5896aa47d99a700fe","scope":"core","source_url":"https://arxiv.org/abs/2509.18119","title":"MobileRL: Online Agentic Reinforcement Learning for Mobile GUI Agents","topics":["reinforcement-learning","mobile-navigation","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.20729","kind":"paper","record_sha256":"2ca2108d2fc57d956c263ffb04c48b7633adad915c240f63cdcf11a7b2e0957a","scope":"core","source_url":"https://arxiv.org/abs/2509.20729","title":"Robust, Observable, and Evolvable Agentic Systems Engineering: A Principled Framework Validated via the Fairy GUI Agent","topics":["observability","human-in-the-loop","memory","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.21552","kind":"paper","record_sha256":"96378a38a967ea5a0e00e4732c4e7d526aabea0d3b100473ef0bd5a7e755c8e3","scope":"core","source_url":"https://arxiv.org/abs/2509.21552","title":"Learning GUI Grounding with Spatial Reasoning from Visual Feedback","topics":["grounding","reinforcement-learning","action-prediction","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.21823","kind":"paper","record_sha256":"21576a8e309a7dc869a35905c5e1c7f10b0083e083e57502dd07f6fb7826f92c","scope":"core","source_url":"https://arxiv.org/abs/2509.21823","title":"ProRe: A Proactive Reward System for GUI Agents via Reasoner-Actor Collaboration","topics":["reward-models","evaluation","planning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2509.21982","kind":"paper","record_sha256":"c0893ed6d68edd260ccfa49f1e8b57d5dc158aa2a4469d2a52a6b934dce337fb","scope":"core","source_url":"https://arxiv.org/abs/2509.21982","title":"RISK: A Framework for GUI Agents in E-commerce Risk Management","topics":["industry","data-generation","evaluation","reinforcement-learning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.22256","kind":"paper","record_sha256":"8d7d40ce5e8cb207ac4dff564649c8e575318f1916d4652d606c05952186d0a7","scope":"core","source_url":"https://arxiv.org/abs/2509.22256","title":"Secure and Efficient Access Control for Computer-Use Agents via Context Space","topics":["safety-security","tool-integration"],"year":"2025"},{"environments":["android"],"id":"arxiv:2509.23263","kind":"paper","record_sha256":"106506e9e46508e15f49961bb5efbb074934863a82ffa61188abf7461e4c84c7","scope":"core","source_url":"https://arxiv.org/abs/2509.23263","title":"GUI-PRA: Process Reward Agent for GUI Tasks","topics":["reward-models","evaluation","screen-understanding"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2509.23866","kind":"paper","record_sha256":"03dc4abd1a838cb6128f8eac37cef46fc1a382a2752b0cb2a5c6885f75a6e961","scope":"core","source_url":"https://arxiv.org/abs/2509.23866","title":"Efficient Multi-turn RL for GUI Agents via Decoupled Training and Adaptive Data Curation","topics":["reinforcement-learning","efficiency","data-generation","sandbox"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.24133","kind":"paper","record_sha256":"95833e4f087b60e40190a418acc52b00aed4657c7a6d4d009a026ed0561934f3","scope":"core","source_url":"https://arxiv.org/abs/2509.24133","title":"Generalist Scanner Meets Specialist Locator: A Synergistic Coarse-to-Fine Framework for Robust GUI Grounding","topics":["grounding","planning","evaluation"],"year":"2025"},{"environments":["web","android"],"id":"arxiv:2509.24183","kind":"paper","record_sha256":"a1108f3ae77f8fdb1c6fd1fb23c465a1c351cd5af01b46101a4a5c689b32a849","scope":"core","source_url":"https://arxiv.org/abs/2509.24183","title":"Retrieval-augmented GUI Agents with Generative Guidelines","topics":["memory","planning","imitation-learning","web-navigation","mobile-navigation"],"year":"2025"},{"environments":["android","linux","desktop","cross-platform"],"id":"arxiv:2509.25047","kind":"paper","record_sha256":"12156b95310309da37444c0c297857ed1ac2824ff9eee8bb02b72b1fbd14b7b1","scope":"core","source_url":"https://arxiv.org/abs/2509.25047","title":"Scaling Synthetic Task Generation for Agents via Exploration","topics":["data-generation","reinforcement-learning","imitation-learning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2509.26506","kind":"paper","record_sha256":"989a29bce4a4450aeec3c47b2ad675645671dd691ac1b4815e33f452038594e2","scope":"core","source_url":"https://arxiv.org/abs/2509.26506","title":"SCUBA: Salesforce Computer Use Benchmark","topics":["evaluation","industry","workflow","sandbox"],"year":"2025"},{"environments":["android","web","desktop","cross-platform"],"id":"arxiv:2509.26539","kind":"paper","record_sha256":"4a316abcb1256d8bcb5f2dcd3611ea137761c604f8ee7f5061d6f09035543fca","scope":"core","source_url":"https://arxiv.org/abs/2509.26539","title":"Ferret-UI Lite: Lessons from Building Small On-Device GUI Agents","topics":["grounding","reinforcement-learning","tool-integration","efficiency"],"year":"2025"},{"environments":["web","cross-platform"],"id":"arxiv:2510.00413","kind":"paper","record_sha256":"c515712260afa789c862cb87570efe97280621a5e089caeef06dc77fdf5c9892","scope":"core","source_url":"https://arxiv.org/abs/2510.00413","title":"PAL-UI: Planning with Active Look-back for Vision-Based GUI Agents","topics":["memory","planning","interaction-history","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.00536","kind":"paper","record_sha256":"20efd24bf1016eb721a8ab56665f7024865fd17c53933d5d2f33be2b08af528c","scope":"core","source_url":"https://arxiv.org/abs/2510.00536","title":"GUI-KV: Efficient GUI Agents via KV Cache with Spatio-Temporal Awareness","topics":["efficiency","interaction-history","training-free","screen-understanding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.01354","kind":"paper","record_sha256":"b2e1f363140c5d4a3b98b3696546b1397cfca223ac3ecbef84d7ee898fa51630","scope":"core","source_url":"https://arxiv.org/abs/2510.01354","title":"WAInjectBench: Benchmarking Prompt Injection Detections for Web Agents","topics":["safety-security","evaluation","robustness"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.01524","kind":"paper","record_sha256":"d5b5a1c157499f6886f5f26d26236d3efcb67fbcc2eaedf3dbfa83a194519df0","scope":"core","source_url":"https://arxiv.org/abs/2510.01524","title":"WALT: Web Agents that Learn Tools","topics":["web-navigation","tool-integration","efficiency","automation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2510.01670","kind":"paper","record_sha256":"baf4d29d08a629cd8bd3ff0aaf4bea22d8d153f4b64d8fceb4054da06f5b2bce","scope":"core","source_url":"https://arxiv.org/abs/2510.01670","title":"Just Do It!? Computer-Use Agents Exhibit Blind Goal-Directedness","topics":["evaluation","safety-security","robustness"],"year":"2025"},{"environments":["desktop","windows","android"],"id":"arxiv:2510.02250","kind":"paper","record_sha256":"bdf20ea55b57f7679361ee59782ff409633242cd32f56654dd9fd937c4440c19","scope":"core","source_url":"https://arxiv.org/abs/2510.02250","title":"Scaling Agents for Computer Use","topics":["evaluation","reward-models","long-horizon"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.02418","kind":"paper","record_sha256":"7ebcf2f716ef75611837be9c0f2ba06fe581634dbd8cc1aa0af4270e7a370710","scope":"core","source_url":"https://arxiv.org/abs/2510.02418","title":"BrowserArena: Evaluating LLM Agents on Real-World Web Navigation Tasks","topics":["evaluation","web-navigation","human-in-the-loop"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.03230","kind":"paper","record_sha256":"0b4e644d85dfcb5d58c488fc51e5e07c0659401c23ca202338329f9bf7856d8c","scope":"core","source_url":"https://arxiv.org/abs/2510.03230","title":"Improving GUI Grounding with Explicit Position-to-Coordinate Mapping","topics":["grounding","screen-understanding","robustness"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.03285","kind":"paper","record_sha256":"b8cf745c30a8f1c461a785fc9741e3b0f0cdf4fc68637974b7f7d86c477a7002","scope":"core","source_url":"https://arxiv.org/abs/2510.03285","title":"WAREX: Web Agent Reliability Evaluation on Existing Benchmarks","topics":["evaluation","robustness","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.04039","kind":"paper","record_sha256":"9d1d0421ced06da2f8a41372d2f5d569992bda996eb8ef653c7d5a1ea9c97c6e","scope":"core","source_url":"https://arxiv.org/abs/2510.04039","title":"\\textsc{GUI-Spotlight}: Adaptive Iterative Focus Refinement for Enhanced GUI Visual Grounding","topics":["grounding","planning","tool-integration"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.04373","kind":"paper","record_sha256":"6eb137daaa7969664ecf3752a597a745549bfd35194e3e425c5016882ee74052","scope":"core","source_url":"https://arxiv.org/abs/2510.04373","title":"JEF-Hinter: Leveraging Offline Knowledge for Improving Web Agents Adaptation","topics":["memory","interaction-history","web-navigation"],"year":"2025"},{"environments":["desktop","windows"],"id":"arxiv:2510.04607","kind":"paper","record_sha256":"77e8b50af4afc925ab9c98ccce14171aa455c2d0d854d5374a146948e133c23b","scope":"core","source_url":"https://arxiv.org/abs/2510.04607","title":"From Imperative to Declarative: Towards LLM-friendly OS Interfaces for Boosted Computer-Use Agents","topics":["automation","tool-integration","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.07988","kind":"paper","record_sha256":"b54a4f19baeec627a4c515599edd03ba6d640cd508ddbc815b6ba6fde47fa8fa","scope":"core","source_url":"https://arxiv.org/abs/2510.07988","title":"ReInAgent: A Context-Aware GUI Agent Enabling Human-in-the-Loop Mobile Task Navigation","topics":["human-in-the-loop","memory","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.09038","kind":"paper","record_sha256":"843f527fae6f9d6a8df57ee0463b10e7e0086c92babd98bce148042780784758","scope":"core","source_url":"https://arxiv.org/abs/2510.09038","title":"Auto-scaling Continuous Memory for GUI Agent","topics":["memory","interaction-history","efficiency"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.09872","kind":"paper","record_sha256":"c711bfa8b681930af8fbd4ed6791138e7a8438cbdb1286cf0d188c7a9436dbd4","scope":"core","source_url":"https://arxiv.org/abs/2510.09872","title":"WARC-Bench: Web Archive Based Benchmark for GUI Subtask Executions","topics":["evaluation","web-navigation","sandbox","reinforcement-learning"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.10073","kind":"paper","record_sha256":"86c53e7dcbb0a8ab29a4573d9472ac6f73481ceb3c7230da31a172d77307d3cb","scope":"core","source_url":"https://arxiv.org/abs/2510.10073","title":"SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents","topics":["safety-security","evaluation","robustness"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.10666","kind":"paper","record_sha256":"9e23d4a16c7bb5554e518e5189b8e2239bcb3ab2bcbe50a1638b6d08d537fdbc","scope":"core","source_url":"https://arxiv.org/abs/2510.10666","title":"BrowserAgent: Building Web Agents with Human-Inspired Web Browsing Actions","topics":["web-navigation","imitation-learning","memory"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.11035","kind":"paper","record_sha256":"485894554bbb537aba39683c8d01bd59e644e07c965fc03c431d1aec2f3d4dd8","scope":"core","source_url":"https://arxiv.org/abs/2510.11035","title":"SusBench: An Online Benchmark for Evaluating Dark Pattern Susceptibility of Computer-Use Agents","topics":["ui-design","safety-security","evaluation"],"year":"2025"},{"environments":["desktop","web"],"id":"arxiv:2510.11892","kind":"paper","record_sha256":"6133db8184e65ce2ff63538215321dfe45a70359afb323c97f49055ef2ed6613","scope":"core","source_url":"https://arxiv.org/abs/2510.11892","title":"R-WoM: Retrieval-augmented World Model For Computer-use Agents","topics":["world-models","memory","planning","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.12200","kind":"paper","record_sha256":"8cf2403357efc7f15799a39571685454bc8b3d6f468a5bfa3ddaae6ba7394875","scope":"core","source_url":"https://arxiv.org/abs/2510.12200","title":"HackWorld: Evaluating Computer-Use Agents on Exploiting Web Application Vulnerabilities","topics":["safety-security","evaluation","web-navigation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.13543","kind":"paper","record_sha256":"2a7c155b1537eb78a574bfa6103b2e5132e6e78af82d8c4115f540c0111857db","scope":"core","source_url":"https://arxiv.org/abs/2510.13543","title":"In-Browser LLM-Guided Fuzzing for Real-Time Prompt Injection Testing in Agentic AI Browsers","topics":["safety-security","automation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2510.14388","kind":"paper","record_sha256":"3344fd12c3e6571278fc1332dca84c9b7eff17b559968f51b6252deb496ef18a","scope":"core","source_url":"https://arxiv.org/abs/2510.14388","title":"Hi-Agent: Hierarchical Vision-Language Agents for Mobile Device Control","topics":["planning","reinforcement-learning","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.14621","kind":"paper","record_sha256":"0f1aec7008db05dfd76151d84cdb66e5d1e62b9246d423532a1db71702650c84","scope":"core","source_url":"https://arxiv.org/abs/2510.14621","title":"ColorBench: Benchmarking Mobile Agents with Graph-Structured Framework for Complex Long-Horizon Tasks","topics":["evaluation","mobile-navigation","long-horizon","state-tracking"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.15455","kind":"paper","record_sha256":"df6aae799505a0c7364b9791505d7368607a5d58ad9ffd4cf4df9d542b08dac7","scope":"core","source_url":"https://arxiv.org/abs/2510.15455","title":"CORE: Reducing UI Exposure in Mobile Agents via Collaboration Between Cloud and Local LLMs","topics":["safety-security","mobile-navigation","efficiency"],"year":"2025"},{"environments":["macos","desktop"],"id":"arxiv:2510.16051","kind":"paper","record_sha256":"1fa5b6e7908f66bbe7eef92605d5f674db6acf232e10d043952513f14fa197c1","scope":"core","source_url":"https://arxiv.org/abs/2510.16051","title":"GUIrilla: A Scalable Framework for Automated Desktop UI Exploration","topics":["data-generation","screen-understanding","recording"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.16252","kind":"paper","record_sha256":"4cd1acbbf31f14fe8b41c942604ecef0dfb429bb32ded0704d1e0e14fe6d1a30","scope":"core","source_url":"https://arxiv.org/abs/2510.16252","title":"WEBSERV: A Full-Stack and RL-Ready Web Environment for Training Web Agents at Scale","topics":["sandbox","reinforcement-learning","efficiency","web-navigation"],"year":"2025"},{"environments":["desktop","windows"],"id":"arxiv:2510.17790","kind":"paper","record_sha256":"734bef5566aec1f7a34f25a66fae13a72d37af491d2ea2ee4d8ca840cdf8c217","scope":"core","source_url":"https://arxiv.org/abs/2510.17790","title":"UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Action","topics":["tool-integration","reinforcement-learning","data-generation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.18113","kind":"paper","record_sha256":"27d2cb7fa871c35593bb795ba005c7c05bdcd0c1114e2db49d59b87c3c10461a","scope":"core","source_url":"https://arxiv.org/abs/2510.18113","title":"Investigating the Impact of Dark Patterns on LLM-Based Web Agents","topics":["safety-security","ui-design","evaluation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2510.18488","kind":"paper","record_sha256":"4bfca229ede1e97a38c66c77ba01ad1b1741cb3f06c426204ecd8f6c7aba3f54","scope":"core","source_url":"https://arxiv.org/abs/2510.18488","title":"AndroidControl-Curated: Revealing the True Potential of GUI Agents through Benchmark Purification","topics":["evaluation","data-generation","grounding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2510.19205","kind":"paper","record_sha256":"bed516cea6840c98271ef1749dc606c480c7ac8dd0c69d5d197762e828a79588","scope":"core","source_url":"https://arxiv.org/abs/2510.19205","title":"WebGraphEval: Multi-Turn Trajectory Evaluation for Web Agents using Graph Representation","topics":["evaluation","observability","interaction-history"],"year":"2025"},{"environments":["desktop","web"],"id":"arxiv:2510.19488","kind":"paper","record_sha256":"9caf00c76df46f3b40e228e9cbcd1b4394b0d1e8b47a5153de01a4c41ba15b74","scope":"core","source_url":"https://arxiv.org/abs/2510.19488","title":"VideoAgentTrek: Computer Use Pretraining from Unlabeled Videos","topics":["data-generation","video"],"year":"2025"},{"environments":["web","desktop","android"],"id":"arxiv:2510.19949","kind":"paper","record_sha256":"7bcf331db74af2fe2590c7b9f4917038aac689443e326c8ab4464972486fbb0e","scope":"core","source_url":"https://arxiv.org/abs/2510.19949","title":"Surfer 2: The Next Generation of Cross-Platform Computer Use Agents","topics":["planning","memory","long-horizon","robustness"],"year":"2025"},{"environments":["cross-platform","android"],"id":"arxiv:2510.20286","kind":"paper","record_sha256":"91dd6804687fa365cdd27b2a6f4b904708acf5b4be037d35e21b25293c46f1e2","scope":"core","source_url":"https://arxiv.org/abs/2510.20286","title":"UI-Ins: Enhancing GUI Grounding with Multi-Perspective Instruction-as-Reasoning","topics":["grounding","reinforcement-learning"],"year":"2025"},{"environments":["android"],"id":"arxiv:2510.22009","kind":"paper","record_sha256":"5dfed4b534cb0d10e84002fe8036ef42b050e59613502b775a0734e538a3456d","scope":"core","source_url":"https://arxiv.org/abs/2510.22009","title":"OpenPhone: Mobile Agentic Foundation Models","topics":["reinforcement-learning","memory","efficiency","mobile-navigation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2510.24168","kind":"paper","record_sha256":"31c309e39a663ec7a77b1accca3d5e46b1ff17491b3f1267ab88050a8fb56302","scope":"core","source_url":"https://arxiv.org/abs/2510.24168","title":"MGA: Memory-Driven GUI Agent for Observation-Centric Interaction","topics":["memory","state-tracking","long-horizon"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.24411","kind":"paper","record_sha256":"92f89e2b07cc71772293baa4b65808dc6bc2d05f9cb29b52b98e3d5c92e8e884","scope":"core","source_url":"https://arxiv.org/abs/2510.24411","title":"OS-Sentinel: Towards Safety-Enhanced Mobile GUI Agents via Hybrid Validation in Realistic Workflows","topics":["safety-security","evaluation","mobile-navigation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2510.24563","kind":"paper","record_sha256":"751513dbec35b4ab271afdc2db349ebdcf060796cabc3c740ca0d5d342e43667","scope":"core","source_url":"https://arxiv.org/abs/2510.24563","title":"OSWorld-MCP: Benchmarking MCP Tool Invocation In Computer-Use Agents","topics":["evaluation","tool-integration","planning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2510.27266","kind":"paper","record_sha256":"12045bdad3fd452dc892303a8885d359f0b0ef22453e3ec80a0d20f309e230fd","scope":"core","source_url":"https://arxiv.org/abs/2510.27266","title":"Enhancing Trustworthy GUI Grounding via Self-Critiqued Reinforcement Learning","topics":["grounding","reinforcement-learning","robustness","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.00810","kind":"paper","record_sha256":"2d495da3b9908359468d42a89c8498e4d772a7fde58282530488a014249df3aa","scope":"core","source_url":"https://arxiv.org/abs/2511.00810","title":"GUI-AIMA: Aligning Intrinsic Multimodal Attention with a Context Anchor for GUI Grounding","topics":["grounding","screen-understanding","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.04137","kind":"paper","record_sha256":"18c707513c0ed630b878631ece0536cdc88d7ea7bfda8021b70675526cf8767e","scope":"core","source_url":"https://arxiv.org/abs/2511.04137","title":"Learning from Online Videos at Inference Time for Computer-Use Agents","topics":["video","memory","planning","data-generation"],"year":"2025"},{"environments":["desktop","windows"],"id":"arxiv:2511.04307","kind":"paper","record_sha256":"59d68483c46b1b274cdc395ca6fa2ed14ae20b9d1515cbbebc9d303d3da2578c","scope":"core","source_url":"https://arxiv.org/abs/2511.04307","title":"GUI-360$^\\circ$: A Comprehensive Dataset and Benchmark for Computer-Using Agents","topics":["data-generation","grounding","evaluation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2511.06101","kind":"paper","record_sha256":"00d524387232da417c785fbadc5aef54907ae649d444a1529b96e7948e2c9276","scope":"core","source_url":"https://arxiv.org/abs/2511.06101","title":"SynthAgent: Adapting Web Agents with Synthetic Supervision","topics":["data-generation","imitation-learning","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.06417","kind":"paper","record_sha256":"f32200f1cadc263c3cb353f982c5213be453ea5a9b890440f616de3a57b4f1e4","scope":"core","source_url":"https://arxiv.org/abs/2511.06417","title":"AUTO-Explorer: Automated Data Collection for GUI Agent","topics":["data-generation","grounding","evaluation"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2511.07332","kind":"paper","record_sha256":"5fcca76c183311c62ba5bc00c13fdbfa74bebec2f7e30bb6f67149356944521c","scope":"core","source_url":"https://arxiv.org/abs/2511.07332","title":"Grounding Computer Use Agents on Human Demonstrations","topics":["grounding","data-generation","evaluation","reinforcement-learning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.09127","kind":"paper","record_sha256":"3bb6aad63fe956826e1f3f7dc76bc7244cb3c27bc2a6adc16b9e9c18d6e367b1","scope":"core","source_url":"https://arxiv.org/abs/2511.09127","title":"History-Aware Reasoning for GUI Agents","topics":["interaction-history","memory","self-reflection","reinforcement-learning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.09157","kind":"paper","record_sha256":"8906803a4591e200f1765a51a2ac33ff9d35f62b6f68d292c5eaf2b798922afc","scope":"core","source_url":"https://arxiv.org/abs/2511.09157","title":"ProBench: Benchmarking GUI Agents with Accurate Process Information","topics":["evaluation","mobile-navigation","observability"],"year":"2025"},{"environments":["web","android"],"id":"arxiv:2511.10705","kind":"paper","record_sha256":"192047a92def0132be2dc21f7c3872d0225a43f1fa0e1611797bea05a9344325","scope":"core","source_url":"https://arxiv.org/abs/2511.10705","title":"Co-EPG: A Framework for Co-Evolution of Planning and Grounding in Autonomous GUI Agents","topics":["grounding","planning","reinforcement-learning","data-generation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.11672","kind":"paper","record_sha256":"e2a70e1137053235c7cd3c0574f25da59b66885e0f1f501d5f3280b492484146","scope":"core","source_url":"https://arxiv.org/abs/2511.11672","title":"OSGym: Scalable OS Infra for Computer Use Agents","topics":["sandbox","efficiency","reinforcement-learning","data-generation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.12254","kind":"paper","record_sha256":"edbf6269ccf490335257f8378902eafdced39fbb53cd673f551b7848ca3e9509","scope":"core","source_url":"https://arxiv.org/abs/2511.12254","title":"Mobile-Agent-RAG: Driving Smart Multi-Agent Coordination with Contextual Knowledge Empowerment for Long-Horizon Mobile Automation","topics":["mobile-navigation","memory","long-horizon"],"year":"2025"},{"environments":["web"],"id":"arxiv:2511.12997","kind":"paper","record_sha256":"592e05607e633338120da7bd3312638da0d8802371e886316510df57202b6b40","scope":"core","source_url":"https://arxiv.org/abs/2511.12997","title":"WebCoach: Self-Evolving Web Agents with Cross-Session Memory Guidance","topics":["memory","interaction-history","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.13087","kind":"paper","record_sha256":"93b3acc43dfa9fbe99959460294e106ca16462fbe8856e287a2c024d82c23c44","scope":"core","source_url":"https://arxiv.org/abs/2511.13087","title":"MEGA-GUI: Multi-stage Enhanced Grounding Agents for GUI Elements","topics":["grounding","screen-understanding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2511.15567","kind":"paper","record_sha256":"4e0a55e69ed839f21ed1b9540e10c7a8e2172549e091f793c8d22000f8820811","scope":"core","source_url":"https://arxiv.org/abs/2511.15567","title":"Computer-Use Agents as Judges for Generative User Interface","topics":["evaluation","ui-design","observability"],"year":"2025"},{"environments":["android"],"id":"arxiv:2511.16590","kind":"paper","record_sha256":"710c008749c73be4b0cd150736067d971f3a050e9b86f90fc46b05655e8f264b","scope":"core","source_url":"https://arxiv.org/abs/2511.16590","title":"D-GARA: A Dynamic Benchmarking Framework for GUI Agent Robustness in Real-World Anomalies","topics":["evaluation","robustness","mobile-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.17131","kind":"paper","record_sha256":"cc1867f285cd688d7a6968cf2c366470ea824ef962d4c5e8618b23a484c6f4ed","scope":"core","source_url":"https://arxiv.org/abs/2511.17131","title":"UI-CUBE: Enterprise-Grade Computer Use Agent Benchmarking Beyond Task Accuracy to Operational Reliability","topics":["evaluation","industry","workflow","robustness"],"year":"2025"},{"environments":["web"],"id":"arxiv:2511.19477","kind":"paper","record_sha256":"1999ba9b3876ba3bbc69260e1e634885ce83e9d3453318c4950432f80f3c60cd","scope":"core","source_url":"https://arxiv.org/abs/2511.19477","title":"Building Browser Agents: Architecture, Security, and Practical Solutions","topics":["web-navigation","safety-security","tool-integration"],"year":"2025"},{"environments":["web"],"id":"arxiv:2511.19663","kind":"paper","record_sha256":"f6531cb44bf5e6863623abe3f7846f732a2c9b4bb2f115270d1f9b7c17461e0e","scope":"core","source_url":"https://arxiv.org/abs/2511.19663","title":"Fara-7B: An Efficient Agentic Model for Computer Use","topics":["data-generation","efficiency","web-navigation"],"year":"2025"},{"environments":["desktop","macos"],"id":"arxiv:2511.20067","kind":"paper","record_sha256":"e979470fd6dfdfcd1dc833c143ebe3f2134410293be3f67b609e365abb5b975c","scope":"core","source_url":"https://arxiv.org/abs/2511.20067","title":"\"Are We Done Yet?\": A Vision-Based Judge for Autonomous Task Completion of Computer Use Agents","topics":["evaluation","self-reflection"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.20766","kind":"paper","record_sha256":"b968fcc5519597c4287931fd0c1ccc56d64cfd4e09d742be52fb0938c3c42d5d","scope":"core","source_url":"https://arxiv.org/abs/2511.20766","title":"OpenApps: Simulating Environment Variations to Measure UI-Agent Reliability","topics":["evaluation","robustness","data-generation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2511.21398","kind":"paper","record_sha256":"1e5a0e4903c9711b7961fa4eba8b49a0f8380c367fa7eb819754002a0abc1e89","scope":"core","source_url":"https://arxiv.org/abs/2511.21398","title":"Prune4Web: DOM Tree Pruning Programming for Web Agent","topics":["grounding","screen-understanding","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2511.22235","kind":"paper","record_sha256":"093caa4c10d280aeac443a48f84a00c9464fd13b778a17912ed6f2b111ecd8c4","scope":"core","source_url":"https://arxiv.org/abs/2511.22235","title":"Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation","topics":["planning","state-tracking","reinforcement-learning","long-horizon"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.00756","kind":"paper","record_sha256":"2edb28b141f24ac0aee0f6a0d899c6613da1294fbf893500c079bf98af191fb6","scope":"core","source_url":"https://arxiv.org/abs/2512.00756","title":"MPR-GUI: Benchmarking and Enhancing Multilingual Perception and Reasoning in GUI Agents","topics":["evaluation","screen-understanding","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.00846","kind":"paper","record_sha256":"79af123a28d2ddb31c436e450d113ce5ee29f5cf01e1caa12019c58f8bb1746d","scope":"core","source_url":"https://arxiv.org/abs/2512.00846","title":"AFRAgent : An Adaptive Feature Renormalization Based High Resolution Aware GUI agent","topics":["grounding","mobile-navigation","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.01174","kind":"paper","record_sha256":"84340fe72ab3bb370a5aaa3687eb76f461a6b7235b9cfc6b709d51dca96e2272","scope":"core","source_url":"https://arxiv.org/abs/2512.01174","title":"DrawingBench: Evaluating Spatial Reasoning and UI Interaction Capabilities of Large Language Models through Mouse-Based Drawing Tasks","topics":["evaluation","grounding","human-in-the-loop","state-tracking"],"year":"2025"},{"environments":["android"],"id":"arxiv:2512.01763","kind":"paper","record_sha256":"46fd9e7821f04bd076c9f67670757bb996a40ce901d690e92c0ca89b78f0a0f9","scope":"core","source_url":"https://arxiv.org/abs/2512.01763","title":"HiconAgent: History Context-aware Policy Optimization for GUI Agents","topics":["interaction-history","memory","efficiency"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.01979","kind":"paper","record_sha256":"7fe296eb15289bd11a39d87ff8008909b79746263881483b874dc6cb1c359a62","scope":"core","source_url":"https://arxiv.org/abs/2512.01979","title":"Chain-of-Ground: Improving GUI Grounding via Iterative Reasoning and Reference Feedback","topics":["grounding","training-free","robustness","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.02423","kind":"paper","record_sha256":"360556634de52cd132285d8006b812bac0492d6db15a00d880e3323319ad024c","scope":"core","source_url":"https://arxiv.org/abs/2512.02423","title":"GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning","topics":["reinforcement-learning","evaluation","data-generation"],"year":"2025"},{"environments":["web"],"id":"arxiv:2512.04105","kind":"paper","record_sha256":"a57213947e71ce402071da682a7579e50010dcf508247d5fec5099a2d88747d9","scope":"core","source_url":"https://arxiv.org/abs/2512.04105","title":"LegalWebAgent: Empowering Access to Justice via LLM-Based Web Agents","topics":["industry","web-navigation","human-in-the-loop","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.05941","kind":"paper","record_sha256":"7a6aa63d0beed2a3acab732a7bfd44f66f708659f7f1ef51d133bef3765e7941","scope":"core","source_url":"https://arxiv.org/abs/2512.05941","title":"Zoom in, Click out: Unlocking and Evaluating the Potential of Zooming for GUI Grounding","topics":["grounding","training-free","evaluation","screen-understanding"],"year":"2025"},{"environments":["web"],"id":"arxiv:2512.07725","kind":"paper","record_sha256":"6500e8cfb739adaf8fe4ae7886ac5bb8ee2f99fd49ee42928f125175ef4a80d0","scope":"core","source_url":"https://arxiv.org/abs/2512.07725","title":"Privacy Practices of Browser Agents","topics":["safety-security","evaluation","web-navigation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.08529","kind":"paper","record_sha256":"7bd7ae61f9e25ce6dcd0e0c1795554209507c9c858dd2d8684be01a840deb032","scope":"core","source_url":"https://arxiv.org/abs/2512.08529","title":"MVP: Multiple View Prediction Improves GUI Grounding","topics":["grounding","training-free","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.09396","kind":"paper","record_sha256":"d3f8903e633c5f970018a3beafe9769c7df6e0fc4dc92f5b57ae834c2accf3d5","scope":"core","source_url":"https://arxiv.org/abs/2512.09396","title":"GAIR: GUI Automation via Information-Joint Reasoning and Group Reflection","topics":["planning","self-reflection","automation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2512.10371","kind":"paper","record_sha256":"cd50553af816ed4625aca1962a56ad784139eb6920d8c7275c9cc13cbd915547","scope":"core","source_url":"https://arxiv.org/abs/2512.10371","title":"AgentProg: Empowering Long-Horizon GUI Agents with Program-Guided Context Management","topics":["interaction-history","memory","long-horizon"],"year":"2025"},{"environments":["web"],"id":"arxiv:2512.10962","kind":"paper","record_sha256":"edfa1a84dadd2883c30ab0ed643f5d1c45cbc9ab6c9dca18885a1bbfaf657e5c","scope":"core","source_url":"https://arxiv.org/abs/2512.10962","title":"WebSTAR: Scalable Data Synthesis for Computer Use Agents with Step-Level Filtering","topics":["data-generation","reward-models","imitation-learning"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.11611","kind":"paper","record_sha256":"d28a831b53f569f03f8f5066063dbb1adef1fe9a766af376859a8d75c041b0a9","scope":"core","source_url":"https://arxiv.org/abs/2512.11611","title":"Using GUI Agent for Electronic Design Automation","topics":["industry","data-generation","evaluation","workflow"],"year":"2025"},{"environments":["web"],"id":"arxiv:2512.12594","kind":"paper","record_sha256":"0392f3d1f89019ef7f684066fa64b2ecdd421e23cce3513c82f45150d34a2196","scope":"core","source_url":"https://arxiv.org/abs/2512.12594","title":"ceLLMate: Sandboxing Browser AI Agents","topics":["safety-security","sandbox","web-navigation","robustness","evaluation"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.12634","kind":"paper","record_sha256":"21778e9fb102491ea87fcdbcd9b5d3edff3ef153fb27a5622f651ddfd5770d9c","scope":"core","source_url":"https://arxiv.org/abs/2512.12634","title":"MobiBench: Multi-Branch, Modular Benchmark for Mobile GUI Agents","topics":["evaluation","mobile-navigation","observability"],"year":"2025"},{"environments":["android"],"id":"arxiv:2512.14014","kind":"paper","record_sha256":"5b39891720ed55119c48f196b83cd9f1ddc3e24cf60d7616a9ffc9ebf1b29e99","scope":"core","source_url":"https://arxiv.org/abs/2512.14014","title":"MobileWorldBench: Towards Semantic World Modeling For Mobile Agents","topics":["world-models","planning","evaluation"],"year":"2025"},{"environments":["android","desktop"],"id":"arxiv:2512.15431","kind":"paper","record_sha256":"6860fa3a736a6a414956e74aae77d6721b36a4c67c8bb99649d31b81c9687d34","scope":"core","source_url":"https://arxiv.org/abs/2512.15431","title":"Step-GUI Technical Report","topics":["data-generation","reward-models","tool-integration","evaluation","mobile-navigation"],"year":"2025"},{"environments":["web","desktop","android"],"id":"arxiv:2512.16295","kind":"paper","record_sha256":"7405146cbf2a0f246629d1012c5f5e357f88bbfe0b6548316dda02a249d92d57","scope":"core","source_url":"https://arxiv.org/abs/2512.16295","title":"OS-Oracle: A Comprehensive Framework for Cross-Platform GUI Critic Models","topics":["reward-models","reinforcement-learning","data-generation","evaluation"],"year":"2025"},{"environments":["cross-platform"],"id":"arxiv:2512.16501","kind":"paper","record_sha256":"f02589d966e469763e23fe7097e0fab03f3393316aaff174c505e51250fcbd3a","scope":"core","source_url":"https://arxiv.org/abs/2512.16501","title":"VenusBench-GD: A Comprehensive Multi-Platform GUI Benchmark for Diverse Grounding Tasks","topics":["grounding","evaluation","data-generation","robustness"],"year":"2025"},{"environments":["web"],"id":"arxiv:2512.17221","kind":"paper","record_sha256":"2c63f7eb0436de428c3456742f612a70bcee569f4a1f007c39dcd186d9964661","scope":"core","source_url":"https://arxiv.org/abs/2512.17221","title":"DAVE: A VLM Vision Encoder for Document Understanding and Web Agents","topics":["screen-understanding","grounding","evaluation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2512.19396","kind":"paper","record_sha256":"51bd80c3784117288eeec5542e965226fd7d4b7238634a087e912dbf92768427","scope":"core","source_url":"https://arxiv.org/abs/2512.19396","title":"EchoTrail-GUI: Building Actionable Memory for GUI Agents via Critic-Guided Self-Exploration","topics":["memory","data-generation","reward-models","mobile-navigation"],"year":"2025"},{"environments":["android"],"id":"arxiv:2512.19432","kind":"paper","record_sha256":"19322b7e1d31a6166613a25b39bc8ae6752bf5b9ac5ad99e60e0e0b0ca9b4cc8","scope":"core","source_url":"https://arxiv.org/abs/2512.19432","title":"MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive and MCP-Augmented Environments","topics":["evaluation","tool-integration","long-horizon"],"year":"2025"},{"environments":["android"],"id":"arxiv:2512.21302","kind":"paper","record_sha256":"94a8b5bf55e2b9ee0f97204e7d249b620e32005e503df707ce84e361639dbce7","scope":"core","source_url":"https://arxiv.org/abs/2512.21302","title":"AndroidLens: Long-latency Evaluation with Nested Sub-targets for Android GUI Agents","topics":["evaluation","long-horizon","mobile-navigation","robustness"],"year":"2025"},{"environments":["unspecified"],"id":"arxiv:2512.22009","kind":"paper","record_sha256":"6244c81183e06fa225ed40f65e0f89babf9fae9e13687b95382b912580f77fa6","scope":"core","source_url":"https://arxiv.org/abs/2512.22009","title":"iSHIFT: Lightweight Slow-Fast GUI Agent with Adaptive Perception","topics":["efficiency","grounding","screen-understanding"],"year":"2025"},{"environments":["android"],"id":"arxiv:2512.22047","kind":"paper","record_sha256":"da27acd392d824eb741fb7dbbd68bc4fc4daf36d53e7525eba1e5d87a6fc3c42","scope":"core","source_url":"https://arxiv.org/abs/2512.22047","title":"MAI-UI Technical Report: Real-World Centric Foundation GUI Agents","topics":["human-in-the-loop","tool-integration","reinforcement-learning","data-generation","efficiency"],"year":"2025"},{"environments":["web"],"id":"arxiv:2512.23128","kind":"paper","record_sha256":"e977dfaed27ef771ecef41e1a857cc5e5df2f76636d3ce653e850ec391efcdde","scope":"core","source_url":"https://arxiv.org/abs/2512.23128","title":"It's a TRAP! Task-Redirecting Agent Persuasion Benchmark for Web Agents","topics":["safety-security","evaluation","ui-design"],"year":"2025"},{"environments":["web"],"id":"arxiv:2512.23647","kind":"paper","record_sha256":"c3b8e4c4dad81667e06bba242928202ff480f699e3cc643af1438b4d02382e74","scope":"core","source_url":"https://arxiv.org/abs/2512.23647","title":"Nested Browser-Use Learning for Agentic Information Seeking","topics":["web-navigation","planning","efficiency","tool-integration"],"year":"2025"},{"environments":["desktop"],"id":"arxiv:2512.24965","kind":"paper","record_sha256":"ac81594d05c5e29f4f1a8b84dcdc2c551cc59ad858da7a67378c8abd18637859","scope":"core","source_url":"https://arxiv.org/abs/2512.24965","title":"ShowUI-$π$: Flow-based Generative Models as GUI Dexterous Hands","topics":["grounding","evaluation","data-generation"],"year":"2025"},{"environments":["windows","desktop"],"id":"arxiv:2601.02399","kind":"paper","record_sha256":"0c349ba8e3dca737a59aefd1cafe114fcd526608bd9c658ca2313d7a67687c55","scope":"core","source_url":"https://arxiv.org/abs/2601.02399","title":"ProSoftArena: Benchmarking Hierarchical Capabilities of Multimodal Agents in Professional Software Environments","topics":["evaluation","industry","science","cross-app"],"year":"2025"},{"environments":["web"],"id":"arxiv:2601.02439","kind":"paper","record_sha256":"d368e207c1045f3bda559a905cb2e1315e8dfcd3fee810ad57e84501e78acf42","scope":"core","source_url":"https://arxiv.org/abs/2601.02439","title":"WebGym: Scaling Training Environments for Visual Web Agents with Realistic Tasks","topics":["web-navigation","reinforcement-learning","data-generation","evaluation","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.03928","kind":"paper","record_sha256":"7a615e326e6ef76baa2de296107d78ed40a349ab5808e2d6acc6f3ec81cd038a","scope":"core","source_url":"https://arxiv.org/abs/2601.03928","title":"FocusUI: Efficient UI Grounding via Position-Preserving Visual Token Selection","topics":["grounding","efficiency","screen-understanding"],"year":"2026"},{"environments":["android"],"id":"arxiv:2601.04035","kind":"paper","record_sha256":"3eedb2e3a316afe451f58805b4cdd4b6312ff559a721c145a622bcb76bb0e169","scope":"core","source_url":"https://arxiv.org/abs/2601.04035","title":"MobileDreamer: Generative Sketch World Model for GUI Agent","topics":["world-models","planning","mobile-navigation"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2601.04126","kind":"paper","record_sha256":"0a729d31ac1bb13a8ec0944e844454c5a85b533bef25d8bda3e5f10185ca0c53","scope":"core","source_url":"https://arxiv.org/abs/2601.04126","title":"InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training","topics":["data-generation","reinforcement-learning","evaluation","sandbox"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.04500","kind":"paper","record_sha256":"ecf0656a3e9b974ff1f60312920039eadba757a9bdd93d2f781acf96febfeebe","scope":"core","source_url":"https://arxiv.org/abs/2601.04500","title":"GUITester: Enabling GUI Agents for Exploratory Defect Discovery","topics":["automation","evaluation","self-reflection"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2601.05787","kind":"paper","record_sha256":"cb7fe1b2e346dd9748675a72aea4fc989dc4d4361761b875163171dbe6ef72f8","scope":"core","source_url":"https://arxiv.org/abs/2601.05787","title":"From Off-Policy to On-Policy: Enhancing GUI Agents via Bi-level Expert-to-Policy Assimilation","topics":["reinforcement-learning","data-generation","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2601.07181","kind":"paper","record_sha256":"555a26777c42b684e8c543f235653a7280f06e4c38feacf4aa52435892f3296c","scope":"core","source_url":"https://arxiv.org/abs/2601.07181","title":"ShowUI-Aloha: Human-Taught GUI Agent","topics":["recording","workflow","data-generation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2601.07262","kind":"paper","record_sha256":"2c1dba468372412d808cf6308037efd41cf4063f59da4a269ae5d67b40bee1f0","scope":"core","source_url":"https://arxiv.org/abs/2601.07262","title":"ColorBrowserAgent: Complex Long-Horizon Browser Agent with Adaptive Knowledge Evolution","topics":["memory","human-in-the-loop","long-horizon","web-navigation"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2601.07779","kind":"paper","record_sha256":"2cd5223cace25d15a66c2c823bfccf73e7743310d75c9c10ebc31bf63f555c80","scope":"core","source_url":"https://arxiv.org/abs/2601.07779","title":"OS-Symphony: A Holistic Framework for Robust and Generalist Computer-Using Agent","topics":["memory","self-reflection","planning","tool-integration"],"year":"2026"},{"environments":["web"],"id":"arxiv:2601.08406","kind":"paper","record_sha256":"ba31ba2b2a333532896f5c474abafadc37a8dcd64cc956ac737b80f2d063c451","scope":"core","source_url":"https://arxiv.org/abs/2601.08406","title":"WebTrap Park: An Automated Platform for Systematic Security Evaluation of Web Agents","topics":["safety-security","evaluation","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2601.08605","kind":"paper","record_sha256":"82543779714aa39902a8ef5408ec65a49d3d2ddc70b8903371c72a4ec9a70fdc","scope":"core","source_url":"https://arxiv.org/abs/2601.08605","title":"ExpSeek: Self-Triggered Experience Seeking for Web Agents","topics":["memory","planning","web-navigation","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2601.09636","kind":"paper","record_sha256":"fc0d1baa4845a50acb36fe398d1f17211ca9294860762f3be1d059465eeb72c5","scope":"core","source_url":"https://arxiv.org/abs/2601.09636","title":"PersonalAlign: Hierarchical Implicit Intent Alignment for Personalized GUI Agent with Long-Term User-Centric Records","topics":["memory","evaluation","mobile-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.09770","kind":"paper","record_sha256":"dd0691aaee49ed52e4a4bd6fa5ee66e4d44ae2b44db3468b63d3d5d139c5e538","scope":"core","source_url":"https://arxiv.org/abs/2601.09770","title":"GUI-Eyes: Tool-Augmented Perception for Visual Grounding in GUI Agents","topics":["grounding","reinforcement-learning","tool-integration","screen-understanding"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2601.09923","kind":"paper","record_sha256":"30bc5e8ac19ca93e64f008dfdcd7d0413f6d0235348e1934e0bcae85b21ddff7","scope":"core","source_url":"https://arxiv.org/abs/2601.09923","title":"CaMeLs Can Use Computers Too: System-level Security for Computer Use Agents","topics":["safety-security","planning","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.11631","kind":"paper","record_sha256":"b892848be6e2b33b630474ee9d650db9ed984459432cae51b1f236e02b5b3160","scope":"core","source_url":"https://arxiv.org/abs/2601.11631","title":"Compress to Focus: Efficient Coordinate Compression for Policy Optimization in Multi-Turn GUI Agents","topics":["grounding","reinforcement-learning","efficiency","interaction-history"],"year":"2026"},{"environments":["android"],"id":"arxiv:2601.12349","kind":"paper","record_sha256":"fbb8235e5f96432e9737aaf617ffd1c5e3cda81246f6f9586d786a044bf19d7d","scope":"core","source_url":"https://arxiv.org/abs/2601.12349","title":"Mind the Gap: Action Rebinding Attacks against Android GUI Agents","topics":["safety-security","cross-app","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.12822","kind":"paper","record_sha256":"f94f2d674e321f047dbae525039c721788131b38f2d9cae3915026d2b4c4ce6f","scope":"core","source_url":"https://arxiv.org/abs/2601.12822","title":"MirrorGuard: Toward Secure Computer-Use Agents via Simulation-to-Real Reasoning Correction","topics":["safety-security","data-generation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.13060","kind":"paper","record_sha256":"d78f5f671a8fbd7a78d6da68a32fc9a65bb71a9ebcf6c35f29958b94e32b0ce2","scope":"core","source_url":"https://arxiv.org/abs/2601.13060","title":"MagicGUI-RMS: A Multi-Agent Reward Model System for Self-Evolving GUI Agents via Automated Feedback Reflux","topics":["reward-models","data-generation","evaluation","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2601.15876","kind":"paper","record_sha256":"ab45de501057bb5886fc3f837b3f6bf74b0acdd7453b52b2969dc43ce96864dd","scope":"core","source_url":"https://arxiv.org/abs/2601.15876","title":"EvoCUA: Evolving Computer Use Agents via Learning from Scalable Synthetic Experience","topics":["data-generation","sandbox","long-horizon","self-reflection"],"year":"2026"},{"environments":["web"],"id":"arxiv:2601.16356","kind":"paper","record_sha256":"eef0a1b6750392692b731fab53e8f3bb8797cdd89715dd5bbf78120ebd71269b","scope":"core","source_url":"https://arxiv.org/abs/2601.16356","title":"The Behavioral Fabric of LLM-Powered GUI Agents: Human Values and Interaction Outcomes","topics":["evaluation","ui-design","human-in-the-loop"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.17418","kind":"paper","record_sha256":"14bf61b2028b2ea32ad9d3dbe3a036221394a1bf57ddc8ef23437192c0f346a1","scope":"core","source_url":"https://arxiv.org/abs/2601.17418","title":"GraphPilot: GUI Task Automation with One-Step LLM Reasoning Powered by Knowledge Graph","topics":["mobile-navigation","planning","memory","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.17722","kind":"paper","record_sha256":"487c3929075209022b5e77f5a83bf32b2805134b5fd298e26fc039223b92468d","scope":"core","source_url":"https://arxiv.org/abs/2601.17722","title":"EntWorld: A Holistic Environment and Benchmark for Verifiable Enterprise GUI Agents","topics":["evaluation","industry","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.18092","kind":"paper","record_sha256":"7c1bb8832689d0a6910c71259cc94bd5d17c4a6a6d63fe5a877f5b61b05ab98a","scope":"adjacent","source_url":"https://arxiv.org/abs/2601.18092","title":"From Struggle to Success: Context-Aware Guidance for Screen Reader Users in Computer Use","topics":["human-in-the-loop","evaluation","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.18197","kind":"paper","record_sha256":"c75dda3f5615a625575c9ed137762421e79b015dfe28150dd1990f4c1b7222c8","scope":"core","source_url":"https://arxiv.org/abs/2601.18197","title":"GAIA: A Data Flywheel System for Training GUI Test-Time Scaling Critic Models","topics":["reward-models","data-generation","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.18305","kind":"paper","record_sha256":"f4ace9d6f73334f1faaf21c183f0b09089517550217498d859da7ee1f3a2268f","scope":"core","source_url":"https://arxiv.org/abs/2601.18305","title":"SwipeGen: Bridging the Execution Gap in GUI Agents via Human-like Swipe Synthesis","topics":["action-prediction","data-generation","evaluation"],"year":"2026"},{"environments":["android","desktop"],"id":"arxiv:2601.18842","kind":"paper","record_sha256":"febf16ff885b9dbd27961911a12a141c6d18939ebf52c41464c8d43fe1809b4a","scope":"core","source_url":"https://arxiv.org/abs/2601.18842","title":"GUIGuard-Bench: Toward a General Evaluation for Privacy-Preserving GUI Agents","topics":["safety-security","evaluation","grounding"],"year":"2026"},{"environments":["android"],"id":"arxiv:2601.19199","kind":"paper","record_sha256":"d56948cc3712c46a738bfe24b07f8811d8ba87aa1e1036aeacabdfcfb0565bb5","scope":"core","source_url":"https://arxiv.org/abs/2601.19199","title":"MAGNET: Towards Adaptive GUI Agents with Memory-Driven Knowledge Evolution","topics":["memory","grounding","robustness","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.20335","kind":"paper","record_sha256":"46520d6b02449b60f5e660f3c0e0147f7551a58f8f294d1c26e6e189f0d0e6d5","scope":"core","source_url":"https://arxiv.org/abs/2601.20335","title":"MobileBench-OL: A Comprehensive Chinese Benchmark for Evaluating Mobile GUI Agents in Real-World Environment","topics":["evaluation","mobile-navigation","robustness"],"year":"2026"},{"environments":["android","desktop","linux"],"id":"arxiv:2601.20380","kind":"paper","record_sha256":"d338a8186880c6980581850d1cb3cdf950624883ce98463b300d7f33daf5d54b","scope":"core","source_url":"https://arxiv.org/abs/2601.20380","title":"OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Execution","topics":["data-generation","reinforcement-learning","grounding","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.20650","kind":"paper","record_sha256":"8bee359fca20dbb0f017269cecd6f23a6897f6ae071c796045375d26b6ed7ca1","scope":"core","source_url":"https://arxiv.org/abs/2601.20650","title":"OS-Marathon: Benchmarking Computer-Use Agents on Vast-Horizon, Repetitive Tasks","topics":["evaluation","long-horizon","human-in-the-loop"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.20732","kind":"paper","record_sha256":"3332453dbe5164ad86840f3bae3549ead6801305d3fa281b40e4d7e1f3e32e0f","scope":"core","source_url":"https://arxiv.org/abs/2601.20732","title":"Continual GUI Agents","topics":["grounding","reinforcement-learning","robustness","evaluation"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2601.21123","kind":"paper","record_sha256":"b4c9d8d011627a62bd984c66c44fe47a814648cd43ab8636226b580cabc1291e","scope":"core","source_url":"https://arxiv.org/abs/2601.21123","title":"CUA-Skill: Develop Skills for Computer Using Agent","topics":["tool-integration","planning","memory","automation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2601.21253","kind":"paper","record_sha256":"aa05a979a6b997c67ffc06aaa6183837e5cf6fad3739786244c2d3021175839b","scope":"core","source_url":"https://arxiv.org/abs/2601.21253","title":"CovAgent: Overcoming the 30% Curse of Mobile Application Coverage with Agentic AI and Dynamic Instrumentation","topics":["automation","mobile-navigation","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2601.21352","kind":"paper","record_sha256":"34953d40e08d30ac04793998267a49be65ba7254d2cb93cf05c4022a2ee38f4c","scope":"core","source_url":"https://arxiv.org/abs/2601.21352","title":"BEAP-Agent: Backtrackable Execution and Adaptive Planning for GUI Agents","topics":["planning","state-tracking","robustness","long-horizon"],"year":"2026"},{"environments":["web"],"id":"arxiv:2601.21872","kind":"paper","record_sha256":"4a24865168cda7ea87e0f6a61a6cfbd4446e7c17b9ac0f98666d409e1783f0b1","scope":"core","source_url":"https://arxiv.org/abs/2601.21872","title":"WebArbiter: A Principle-Guided Reasoning Process Reward Model for Web Agents","topics":["reward-models","reinforcement-learning","evaluation","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2601.21961","kind":"paper","record_sha256":"b7f6a618a898afbcda74ad212c47da4286e14b77f2528484ce196a1b35881cb0","scope":"core","source_url":"https://arxiv.org/abs/2601.21961","title":"How do Visual Attributes Influence Web Agents? A Comprehensive Evaluation of User Interface Design Factors","topics":["ui-design","evaluation","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2601.22149","kind":"paper","record_sha256":"0b888260ad8f0f5ca0713c763ca4c8955c9641d1584fd9293ecb35ada3ff8e29","scope":"core","source_url":"https://arxiv.org/abs/2601.22149","title":"DynaWeb: Model-Based Reinforcement Learning of Web Agents","topics":["world-models","reinforcement-learning","data-generation","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.22528","kind":"paper","record_sha256":"260deae849904a9fa1c0715d64b6011684b88a86b02b22b86a572a3bb622354e","scope":"core","source_url":"https://arxiv.org/abs/2601.22528","title":"Darwinian Memory: A Training-Free Self-Regulating Memory System for GUI Agent Evolution","topics":["memory","training-free","cross-app","long-horizon"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2601.22781","kind":"paper","record_sha256":"ca9b32ff010f2543b11b0987fc655099ff50e0fb9dd5e02f0fa8b82dd62829ad","scope":"core","source_url":"https://arxiv.org/abs/2601.22781","title":"Learning with Challenges: Adaptive Difficulty-Aware Data Generation for Mobile GUI Agent Training","topics":["data-generation","mobile-navigation","evaluation"],"year":"2026"},{"environments":["desktop","android"],"id":"arxiv:2602.00575","kind":"paper","record_sha256":"e788769cd2cbd7df2d2aff8fa63b08de113120b1ec9bde04862d07fd5823d24d","scope":"core","source_url":"https://arxiv.org/abs/2602.00575","title":"Agentic Reward Modeling: Verifying GUI Agent via Progressive Trajectory-Grounded Interaction","topics":["reward-models","tool-integration","evaluation","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.01576","kind":"paper","record_sha256":"34f3bacadb274b022f66c6d0ed4627c03fea16ef61291ea577cb9c5f8fdc3621","scope":"core","source_url":"https://arxiv.org/abs/2602.01576","title":"Generative Visual Code Mobile World Models","topics":["world-models","data-generation","screen-understanding","mobile-navigation"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2602.01725","kind":"paper","record_sha256":"5d8a3271c562cbc5d1bce7332d530733a7fa1c8913cffbbe1f5f93deb7add849","scope":"core","source_url":"https://arxiv.org/abs/2602.01725","title":"SafePred: A Predictive Guardrail for Computer-Using Agents via World Models","topics":["safety-security","world-models","planning","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.02419","kind":"paper","record_sha256":"155ccb3895fe337cb2a06e1bf87975e2392cbf545d8f6afe57cdd353030b6a60","scope":"core","source_url":"https://arxiv.org/abs/2602.02419","title":"SafeGround: Know When to Trust GUI Grounding Models via Uncertainty Calibration","topics":["grounding","robustness","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.02468","kind":"paper","record_sha256":"5518f3ccc6233c1f99c9560a75496c9152754366d2db1e949c710fe16d0b747c","scope":"core","source_url":"https://arxiv.org/abs/2602.02468","title":"Avenir-Web: Human-Experience-Imitating Multimodal Web Agents with Mixture of Grounding Experts","topics":["grounding","planning","memory","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.02548","kind":"paper","record_sha256":"60ef5c36738f1e1869b3b6a40acecfd933707f895a8a8cc791e0213deb26cdd7","scope":"core","source_url":"https://arxiv.org/abs/2602.02548","title":"ToolTok: Tool Tokenization for Efficient and Generalizable GUI Agents","topics":["tool-integration","grounding","data-generation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2602.02995","kind":"paper","record_sha256":"cd6db06216c5e82728828bbe60bfa95f51f94c9f4aaaa4ef9acf32f36ca163d6","scope":"core","source_url":"https://arxiv.org/abs/2602.02995","title":"Agent Alpha: Tree Search Unifying Generation, Exploration and Evaluation for Computer-Use Agents","topics":["planning","efficiency","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.03792","kind":"paper","record_sha256":"35059dde12900957052ff95069eca97716ef05b0025f63080a390435b5c39d6e","scope":"core","source_url":"https://arxiv.org/abs/2602.03792","title":"WebSentinel: Detecting and Localizing Prompt Injection Attacks for Web Agents","topics":["safety-security","evaluation","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.05354","kind":"paper","record_sha256":"83e0605f86be47c57eb418ae6ff733557dd831f74826c57e417cc62e00439524","scope":"core","source_url":"https://arxiv.org/abs/2602.05354","title":"Do Web Agents Investigate Before They Decide?","topics":["evaluation","web-navigation","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.05429","kind":"paper","record_sha256":"40255856e8bd77f68afc60357fa4b02d4776441bfcaaf02358347c62c8a6c7fe","scope":"core","source_url":"https://arxiv.org/abs/2602.05429","title":"M$^2$-Miner: Multi-Agent Enhanced MCTS for Mobile GUI Agent Data Mining","topics":["data-generation","mobile-navigation","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.05832","kind":"paper","record_sha256":"03de73ab8abf529ee5be24df1e24d2f76c229b6b010926235f79984c38391e4a","scope":"core","source_url":"https://arxiv.org/abs/2602.05832","title":"UI-Mem: Self-Evolving Experience Memory for Online Reinforcement Learning in Mobile GUI Agents","topics":["memory","reinforcement-learning","cross-app"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.06075","kind":"paper","record_sha256":"bd397e6f469dc7c6c2c31a2a5fc097c515e4092f164de642f1308d10aa106870","scope":"core","source_url":"https://arxiv.org/abs/2602.06075","title":"MemGUI-Bench: Benchmarking Memory of Mobile GUI Agents in Dynamic Environments","topics":["memory","evaluation","mobile-navigation","cross-app"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.06351","kind":"paper","record_sha256":"890391feeb2b6eec8433b4db5666f70468375c5ad1241dbd45cad58794101b62","scope":"core","source_url":"https://arxiv.org/abs/2602.06351","title":"Trifuse: Enhancing Attention-Based GUI Grounding via Multimodal Fusion","topics":["grounding","screen-understanding","training-free"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2602.06391","kind":"paper","record_sha256":"406356f24b54091f7303965b5290ca56b4743278fa151933d279a01bf10382a5","scope":"core","source_url":"https://arxiv.org/abs/2602.06391","title":"POINTS-GUI-G: GUI-Grounding Journey","topics":["grounding","reinforcement-learning"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2602.07153","kind":"paper","record_sha256":"276ded089468c3073eed03d8ab4c6ca41c5e786a282191d109f0e066313009ee","scope":"core","source_url":"https://arxiv.org/abs/2602.07153","title":"ANCHOR: Branch-Point Data Generation for GUI Agents","topics":["data-generation","planning","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.07283","kind":"paper","record_sha256":"cd5ea59d40cba4362281abc62f34619fc8339f8219d3a00e7e807a7ab7c9202f","scope":"core","source_url":"https://arxiv.org/abs/2602.07283","title":"Mapping the Design Space of User Experience for Computer Use Agents","topics":["human-in-the-loop","ui-design","evaluation"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2602.08995","kind":"paper","record_sha256":"9bba5bc25f4f35b1213308ed7e398de97ba8204ee13501e70498096f12345d91","scope":"core","source_url":"https://arxiv.org/abs/2602.08995","title":"When Actions Go Off-Task: Detecting and Correcting Misaligned Actions in Computer-Use Agents","topics":["safety-security","evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.09007","kind":"paper","record_sha256":"75589b85498188f8542000b5308f64d6e04328c4258097b54a15107a7ca48e81","scope":"core","source_url":"https://arxiv.org/abs/2602.09007","title":"GUI-GenBench: Evaluating Image Generation Models as Interactive GUI Environments","topics":["evaluation","world-models","grounding","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.09012","kind":"paper","record_sha256":"02fe393eabbb896c89ea6bded7f115ff49c6dd3132c0964935d2096ba384fff4","scope":"core","source_url":"https://arxiv.org/abs/2602.09012","title":"Next-Gen CAPTCHAs: Leveraging the Cognitive Gap for Scalable and Diverse GUI-Agent Defense","topics":["evaluation","safety-security","data-generation"],"year":"2026"},{"environments":["cross-platform","web","android"],"id":"arxiv:2602.09082","kind":"paper","record_sha256":"e28bb84b9219d213ba57be24273f5d41a9f10fe3bcb09e822126db7b5468d242","scope":"core","source_url":"https://arxiv.org/abs/2602.09082","title":"UI-Venus-1.5 Technical Report","topics":["reinforcement-learning","grounding","mobile-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.09310","kind":"paper","record_sha256":"5514c4e4107a4428042c75482b467478eb0e793cb2d264f6cde236375d3d092d","scope":"core","source_url":"https://arxiv.org/abs/2602.09310","title":"A11y-CUA Dataset: Characterizing the Accessibility Gap in Computer Use Agents","topics":["human-in-the-loop","evaluation","data-generation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.09662","kind":"paper","record_sha256":"07ae5b3944ce37a93eefdfecec77d1605dd6707cde40bf8cb7d7f8d6ac6be843","scope":"core","source_url":"https://arxiv.org/abs/2602.09662","title":"TreeCUA: Efficiently Scaling GUI Automation with Tree-Structured Verifiable Evolution","topics":["data-generation","planning","memory","efficiency"],"year":"2026"},{"environments":["android"],"id":"arxiv:2602.09856","kind":"paper","record_sha256":"ffabbdf363a8d14c7258da15fb57732563fe9319295160754198769299a8d8ac","scope":"core","source_url":"https://arxiv.org/abs/2602.09856","title":"Code2World: A GUI World Model via Renderable Code Generation","topics":["world-models","data-generation","reinforcement-learning","mobile-navigation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2602.10139","kind":"paper","record_sha256":"6349620003b8b26c94c318e6a66b0a44d7f4c4370ca1cee6f3f8f504e27434f0","scope":"core","source_url":"https://arxiv.org/abs/2602.10139","title":"Anonymization-Enhanced Privacy Protection for Mobile GUI Agents: Available but Invisible","topics":["safety-security","screen-understanding","mobile-navigation"],"year":"2026"},{"environments":["desktop","linux","windows"],"id":"arxiv:2602.10356","kind":"paper","record_sha256":"335f896f877231f7880a700bda6179d432aa458899864618f7e4b2c233b98303","scope":"core","source_url":"https://arxiv.org/abs/2602.10356","title":"Autonomous Continual Learning for Environment Adaptation of Computer-Use Agents","topics":["reinforcement-learning","data-generation","reward-models","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.10814","kind":"paper","record_sha256":"e7b818ddcf22a2aed3b30247ba0f01acfc473cf62f6f1f3813e08325203911cf","scope":"core","source_url":"https://arxiv.org/abs/2602.10814","title":"See, Plan, Snap: Evaluating Multimodal GUI Agents in Scratch","topics":["evaluation","grounding","automation"],"year":"2026"},{"environments":["android","web"],"id":"arxiv:2602.11524","kind":"paper","record_sha256":"6b110fd5892183f88654d1126d5bff6185eb249f4b3a97d2fe3040d8424142ea","scope":"core","source_url":"https://arxiv.org/abs/2602.11524","title":"Adaptive Milestone Reward for GUI Agents","topics":["reinforcement-learning","reward-models","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.11724","kind":"paper","record_sha256":"f0dde059800d0f85be405a86d624ca985d820121038a28661f26dc4997a40418","scope":"core","source_url":"https://arxiv.org/abs/2602.11724","title":"WebTestPilot: Agentic End-to-End Web Testing against Natural Language Specification by Inferring Oracles with Symbolized GUI Elements","topics":["automation","evaluation","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.11750","kind":"paper","record_sha256":"d92d26f9d14552b0ad8cfee957487928d327a7f0fef51f624e8bf3766a03f8ad","scope":"core","source_url":"https://arxiv.org/abs/2602.11750","title":"AmbiBench: Benchmarking Mobile GUI Agents Beyond One-Shot Instructions in the Wild","topics":["evaluation","human-in-the-loop","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.12276","kind":"paper","record_sha256":"95d102f4e1da27adce453f555731351edde544a317db038bd5f3b07148c50f8e","scope":"core","source_url":"https://arxiv.org/abs/2602.12276","title":"Agentic Test-Time Scaling for WebAgents","topics":["efficiency","planning","evaluation","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.12430","kind":"survey","record_sha256":"d738a59cfefe564cf59fedbd4db58229a40d91150ea870b8ebbdda9e09721866","scope":"adjacent","source_url":"https://arxiv.org/abs/2602.12430","title":"Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward","topics":["survey","tool-integration","safety-security","planning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.12544","kind":"paper","record_sha256":"2f63fb6db303acf3f95a99483ec31e933965d17846b45a2b8fce5ecce82b0a92","scope":"core","source_url":"https://arxiv.org/abs/2602.12544","title":"Scaling Web Agent Training through Automatic Data Generation and Fine-grained Evaluation","topics":["data-generation","evaluation","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.13559","kind":"paper","record_sha256":"2a7f3a92007a329a24935d65970613d229f7739dcc985a77f85339901b253b14","scope":"core","source_url":"https://arxiv.org/abs/2602.13559","title":"OpAgent: Operator Agent for Web Navigation","topics":["web-navigation","reinforcement-learning","planning","self-reflection"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2602.13653","kind":"paper","record_sha256":"25e096136f5e994a6e9617ddef070c29e75fc15b39283e0a25e8bdeb3db95d28","scope":"core","source_url":"https://arxiv.org/abs/2602.13653","title":"Building Autonomous GUI Navigation via Agentic-Q Estimation and Step-Wise Policy Optimization","topics":["reinforcement-learning","reward-models","grounding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.14093","kind":"paper","record_sha256":"1723a3af23fc7cf2b36501a8c8be9246a26ef9d03a9f56202fc4271b43df2c3c","scope":"core","source_url":"https://arxiv.org/abs/2602.14093","title":"GUI-GENESIS: Automated Synthesis of Efficient Environments with Verifiable Rewards for GUI Agent Post-Training","topics":["data-generation","sandbox","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.14276","kind":"paper","record_sha256":"e045d2b3e1eeba0dae3bfd5edb06232869df066818930bda0065cf68b8518688","scope":"core","source_url":"https://arxiv.org/abs/2602.14276","title":"ScreenParse: Moving Beyond Sparse Grounding with Complete Screen Parsing Supervision","topics":["screen-understanding","data-generation","grounding","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.14296","kind":"paper","record_sha256":"aecba5d8f6d0b787970d114d58055537098955b0b090ffb181624b5935cb3c51","scope":"core","source_url":"https://arxiv.org/abs/2602.14296","title":"AutoWebWorld: Synthesizing Infinite Verifiable Web Environments via Finite State Machines","topics":["data-generation","sandbox","web-navigation","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.14721","kind":"paper","record_sha256":"bb356e5ed0f5b5897b6fe4732380a0035b02a28a34101e7e6f5ea24b24c99b35","scope":"core","source_url":"https://arxiv.org/abs/2602.14721","title":"WebWorld: A Large-Scale World Model for Web Agent Training","topics":["world-models","data-generation","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.14865","kind":"paper","record_sha256":"64db2e5becb155e3c7791c0806622c11d034904b23f2d2c39e18c2007ada1327","scope":"core","source_url":"https://arxiv.org/abs/2602.14865","title":"EmbeWebAgent: Embedding Web Agents into Any Customized UI","topics":["web-navigation","tool-integration","automation","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.16313","kind":"paper","record_sha256":"150346314db54af91f10a963be6d0a90cce84be97489da989357e3d01dc042e4","scope":"core","source_url":"https://arxiv.org/abs/2602.16313","title":"MemoryArena: Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks","topics":["memory","interaction-history","planning","evaluation","long-horizon"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2602.16855","kind":"paper","record_sha256":"0ab27a628bcc5e27d6446a37291b2b292f006032531e63c51b4d893a905d807f","scope":"core","source_url":"https://arxiv.org/abs/2602.16855","title":"Mobile-Agent-v3.5: Multi-platform Fundamental GUI Agents","topics":["tool-integration","memory","reinforcement-learning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.17003","kind":"paper","record_sha256":"05cc8e916d3dffdb836ce2b510d94d5867c37decccd683350ff697a39b16612f","scope":"core","source_url":"https://arxiv.org/abs/2602.17003","title":"Persona2Web: Benchmarking Personalized Web Agents for Contextual Reasoning with User History","topics":["evaluation","memory","web-navigation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2602.17365","kind":"paper","record_sha256":"5caca9c5ec17739eeee9db3e2a3255d8da735b6da3ecda544c69f4d5f05090fc","scope":"core","source_url":"https://arxiv.org/abs/2602.17365","title":"Computer-Using World Model","topics":["world-models","planning","action-prediction"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.17588","kind":"paper","record_sha256":"581b67338f289014d1a678d244c98b7683ddcd87f1fc60545fb282a619424b83","scope":"core","source_url":"https://arxiv.org/abs/2602.17588","title":"Modeling Distinct Human Interaction in Web Agents","topics":["human-in-the-loop","web-navigation","data-generation","action-prediction"],"year":"2026"},{"environments":["web"],"id":"arxiv:2602.20502","kind":"paper","record_sha256":"d18dded34c637ee5bc0945e640baa1e530713fae6772e088ab55d688eb056e95","scope":"core","source_url":"https://arxiv.org/abs/2602.20502","title":"ActionEngine: From Reactive to Programmatic Web Agents via State Machine Memory","topics":["memory","planning","efficiency","web-navigation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2603.02951","kind":"paper","record_sha256":"12e57399732eb8b539495f0ad63bd17d23c28e396996736f9ed87773152af686","scope":"core","source_url":"https://arxiv.org/abs/2603.02951","title":"CGL: Advancing Continual GUI Learning via Reinforcement Fine-Tuning","topics":["reinforcement-learning","robustness","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.04364","kind":"paper","record_sha256":"2b03e4fb1d467f60bcfc9d8f54811bebd8bbca66af50eee82955bf1d0fffa64a","scope":"core","source_url":"https://arxiv.org/abs/2603.04364","title":"Dual-Modality Multi-Stage Adversarial Safety Training: Robustifying Multimodal Web Agents Against Cross-Modal Attacks","topics":["safety-security","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.04949","kind":"paper","record_sha256":"aec9851d399b426d8fcfb83a95366cfd3cd3ff2b973b68ce6a5696d7471d8e8c","scope":"core","source_url":"https://arxiv.org/abs/2603.04949","title":"TimeWarp: Evaluating Web Agents by Revisiting the Past","topics":["evaluation","robustness","web-navigation","data-generation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.05044","kind":"paper","record_sha256":"9e5314c27e3929b8518f9085aec5a692e4685d39288def3b479ff42bd76b13d2","scope":"core","source_url":"https://arxiv.org/abs/2603.05044","title":"WebFactory: Automated Compression of Foundational Language Intelligence into Grounded Web Agents","topics":["data-generation","reinforcement-learning","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.05295","kind":"paper","record_sha256":"d93936277371d808aa4573ca03f429590ae69ea6fd3479f6584d9bb54fd7b42a","scope":"core","source_url":"https://arxiv.org/abs/2603.05295","title":"WebChain: A Large-Scale Human-Annotated Dataset of Real-World Web Interaction Traces","topics":["data-generation","evaluation","grounding","planning"],"year":"2026"},{"environments":["android"],"id":"arxiv:2603.07432","kind":"paper","record_sha256":"bd4d74d3d3f7d35db3d066628443b34ca1d09c635ba42c7ef5e3c844e84103fe","scope":"core","source_url":"https://arxiv.org/abs/2603.07432","title":"Generalization in Online Reinforcement Learning for Mobile Agents","topics":["reinforcement-learning","evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.07978","kind":"paper","record_sha256":"5501eade05d0d88e146eded22de3f4a86cd4aa1e8aec09e8ba577978354584d5","scope":"core","source_url":"https://arxiv.org/abs/2603.07978","title":"OSExpert: Computer-Use Agents Learning Professional Skills via Exploration","topics":["planning","data-generation","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.08013","kind":"paper","record_sha256":"08145b0c3189adb79cbd999c52e5b605e602cf7a6ee944d138a763ada6a07d63","scope":"core","source_url":"https://arxiv.org/abs/2603.08013","title":"PIRA-Bench: A Transition from Reactive GUI Agents to GUI-based Proactive Intent Recommendation Agents","topics":["evaluation","screen-understanding","memory","state-tracking","human-in-the-loop"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.08533","kind":"paper","record_sha256":"d7b7e1908b5fd8dffc9499be7b399687833983234b5b583553b6c584c24c1731","scope":"core","source_url":"https://arxiv.org/abs/2603.08533","title":"SecAgent: Efficient Mobile GUI Agent with Semantic Context","topics":["mobile-navigation","memory","reinforcement-learning"],"year":"2026"},{"environments":["desktop","linux","macos","windows","android"],"id":"arxiv:2603.10178","kind":"paper","record_sha256":"65398d25dc6044045f4ebf6ac33ea387a73b2db051e8171ed60552d7e5a511a6","scope":"core","source_url":"https://arxiv.org/abs/2603.10178","title":"Video-Based Reward Modeling for Computer-Use Agents","topics":["reward-models","video","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.10268","kind":"paper","record_sha256":"153e86cc63078e55e6b35e8e60da7c6d76a078ae945fbb30d8f26d7ebf8fcae3","scope":"core","source_url":"https://arxiv.org/abs/2603.10268","title":"SpecOps: A Fully Automated AI Agent Testing Framework in Real-World GUI Environments","topics":["automation","evaluation","planning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.10505","kind":"paper","record_sha256":"01fa65ee13d9467daa1d72019138afe2c55b2d9510069fa154992c6891a1c582","scope":"core","source_url":"https://arxiv.org/abs/2603.10505","title":"Safe and Scalable Web Agent Learning via Recreated Websites","topics":["data-generation","sandbox","reinforcement-learning"],"year":"2026"},{"environments":["desktop","macos","windows","linux"],"id":"arxiv:2603.10577","kind":"paper","record_sha256":"592814cc486450db4c39f2dffdf5f11202bd23a15b09c50641419a59b1618465","scope":"core","source_url":"https://arxiv.org/abs/2603.10577","title":"CUAAudit: Meta-Evaluation of Vision-Language Models as Auditors of Autonomous Computer-Use Agents","topics":["evaluation","observability","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.10664","kind":"paper","record_sha256":"d1882720486881b23c852bf7bbb9e445851f61d581f817c188b831afd871cea7","scope":"adjacent","source_url":"https://arxiv.org/abs/2603.10664","title":"Terminal Is All You Need: Design Properties for Human-AI Agent Collaboration","topics":["human-in-the-loop","ui-design","observability","tool-integration"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.12138","kind":"paper","record_sha256":"a3fc06fa92566ac0830f76b8fd8c8e24e463c9f553e3ab8bc71149f925cfc154","scope":"core","source_url":"https://arxiv.org/abs/2603.12138","title":"HATS: Hardness-Aware Trajectory Synthesis for GUI Agents","topics":["data-generation","robustness","action-prediction"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.12823","kind":"paper","record_sha256":"518a80de5d1bbb94fb757cb4eb6098434c9ca3a737e1f7aae601fbd475bcca7d","scope":"core","source_url":"https://arxiv.org/abs/2603.12823","title":"Adaptive Vision-Language Model Routing for Computer Use Agents","topics":["efficiency","grounding","memory","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.14448","kind":"paper","record_sha256":"fcb71cac9a9df13effd00d38262d70cc3bbf55c6ca09f822c54d33de5f1e93b2","scope":"core","source_url":"https://arxiv.org/abs/2603.14448","title":"Zoom to Essence: Trainless GUI Grounding by Inferring upon Interface Elements","topics":["grounding","training-free","screen-understanding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.14707","kind":"paper","record_sha256":"13ea476faf608c64796767de5c4879635d223c54f11288d9cc48051ec7a4122e","scope":"core","source_url":"https://arxiv.org/abs/2603.14707","title":"Visual Confused Deputy: Exploiting and Defending Perception Failures in Computer-Using Agents","topics":["safety-security","grounding","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.15039","kind":"paper","record_sha256":"305fae851ee697268b5efdd4d3cfc795c28a77fc2287f68251e6d44620f31f2d","scope":"core","source_url":"https://arxiv.org/abs/2603.15039","title":"GUI-CEval: A Hierarchical and Comprehensive Chinese Benchmark for Mobile GUI Agents","topics":["evaluation","mobile-navigation","self-reflection","screen-understanding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.15714","kind":"paper","record_sha256":"72d5e33a01c4684b6462e36f407d6778b0d1b548c30c82e29c916e86367bc6c2","scope":"core","source_url":"https://arxiv.org/abs/2603.15714","title":"How Vulnerable Are AI Agents to Indirect Prompt Injections? Insights from a Large-Scale Public Competition","topics":["safety-security","evaluation","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.17357","kind":"paper","record_sha256":"40c47ccb3464283960ede540d0d96dd676aa59941a51ff0e21187b747c550340","scope":"core","source_url":"https://arxiv.org/abs/2603.17357","title":"WebPII: Benchmarking Visual PII Detection for Computer-Use Agents","topics":["evaluation","data-generation","safety-security","screen-understanding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.17441","kind":"paper","record_sha256":"bf83694e2efad7ccb649aca693a8313d9a55153cba2bd41e93dca1e84126100d","scope":"core","source_url":"https://arxiv.org/abs/2603.17441","title":"AdaZoom-GUI: Adaptive Zoom-based GUI Grounding with Instruction Refinement","topics":["grounding","reinforcement-learning","screen-understanding"],"year":"2026"},{"environments":["android"],"id":"arxiv:2603.18429","kind":"paper","record_sha256":"2d0a96deb786366566f5d18ac49d277f1c3d2da6deafb02ced3c3bad6d33e8af","scope":"core","source_url":"https://arxiv.org/abs/2603.18429","title":"AndroTMem: From Interaction Trajectories to Anchored Memory in Long-Horizon GUI Agents","topics":["memory","long-horizon","evaluation"],"year":"2026"},{"environments":["android","cross-platform"],"id":"arxiv:2603.19191","kind":"paper","record_sha256":"839261aa6745d9be18d6e0ede4b754af9c9e0dad89ec707ff0b5b5733844b92c","scope":"core","source_url":"https://arxiv.org/abs/2603.19191","title":"OS-Themis: A Scalable Critic Framework for Generalist GUI Rewards","topics":["reward-models","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.19685","kind":"paper","record_sha256":"165ec6ab9e62467563de4b171211a4d15113f1d5062ea4f43e14ceceaa6a3c37","scope":"core","source_url":"https://arxiv.org/abs/2603.19685","title":"A Subgoal-driven Framework for Improving Long-Horizon LLM Agents","topics":["planning","reinforcement-learning","reward-models","long-horizon","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2603.19993","kind":"paper","record_sha256":"70701cf823b59f8f08934b76618c0d5494e92db6d1dbf61636283e4ed5c977a2","scope":"core","source_url":"https://arxiv.org/abs/2603.19993","title":"MedSPOT: A Workflow-Aware Sequential Grounding Benchmark for Clinical GUI","topics":["grounding","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.20340","kind":"paper","record_sha256":"8731744acfe32c573110c18547cb6921e548a4dccf50383ab91c785680b7a55f","scope":"core","source_url":"https://arxiv.org/abs/2603.20340","title":"ContractSkill: Repairable Contract-Based Skills for Multimodal Web Agents","topics":["web-navigation","planning","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.20351","kind":"paper","record_sha256":"645a30cc4fb20c4a42b4f366c0dae1f2526d4bc36a2089b179b3d586825a62d5","scope":"core","source_url":"https://arxiv.org/abs/2603.20351","title":"MANA: Towards Efficient Mobile Ad Detection via Multimodal Agentic UI Navigation","topics":["mobile-navigation","screen-understanding","efficiency","safety-security"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.22529","kind":"paper","record_sha256":"a18409a5d54679e7d59e121004af0fd8c7474c1d730c4ed8236d93a9d992cf92","scope":"core","source_url":"https://arxiv.org/abs/2603.22529","title":"Ego2Web: A Web Agent Benchmark Grounded in Egocentric Videos","topics":["evaluation","video","web-navigation","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.23007","kind":"paper","record_sha256":"e4d878d2d08cf3f85e114aeeb029889fcd5d4c0a496f584eb055c0ec023fed0b","scope":"core","source_url":"https://arxiv.org/abs/2603.23007","title":"AgentRAE: Remote Action Execution through Notification-based Visual Backdoors against Screenshots-based Mobile GUI Agents","topics":["safety-security","robustness","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.24157","kind":"paper","record_sha256":"fed0cfa10de6b3eed1ee9ff67cd3bdd949f82ca03be7268ed48e380e0a7d16c9","scope":"core","source_url":"https://arxiv.org/abs/2603.24157","title":"CarePilot: A Multi-Agent Framework for Long-Horizon Computer Task Automation in Healthcare","topics":["automation","planning","memory","evaluation","long-horizon","workflow"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2603.24440","kind":"paper","record_sha256":"2c03d2f32b0ba2d7311dd81334c2a6b79e4ad2fab8f78596c872f6eb7ac65fe6","scope":"core","source_url":"https://arxiv.org/abs/2603.24440","title":"CUA-Suite: Massive Human-annotated Video Demonstrations for Computer-Use Agents","topics":["data-generation","video","evaluation","grounding","workflow"],"year":"2026"},{"environments":["android"],"id":"arxiv:2603.24533","kind":"paper","record_sha256":"cc180add6df9f0792a3a2af6e588fb80706b5c70f6c5f4c9f50b2cd1bea141c5","scope":"core","source_url":"https://arxiv.org/abs/2603.24533","title":"UI-Voyager: A Self-Evolving GUI Agent Learning via Failed Experience","topics":["imitation-learning","data-generation","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.25226","kind":"paper","record_sha256":"b690b5728733df531154d0c17dc175bd90ceb8dfbbd4984deacbdee360f4b0a5","scope":"core","source_url":"https://arxiv.org/abs/2603.25226","title":"WebTestBench: Evaluating Computer-Use Agents towards End-to-End Automated Web Testing","topics":["evaluation","web-navigation","automation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.25864","kind":"paper","record_sha256":"91876ed0ce7cd2c43d2142ab108fc53afe6fdee003446e0c44c4cb9cccd02450","scope":"core","source_url":"https://arxiv.org/abs/2603.25864","title":"GUIDE: A Benchmark for Understanding and Assisting Users in Open-Ended GUI Tasks","topics":["evaluation","human-in-the-loop","video","screen-understanding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.26041","kind":"paper","record_sha256":"fccd3f52f214bc20f3596a1ffc715b40bb2db8b5cc69c502daf3b14c45b1b924","scope":"core","source_url":"https://arxiv.org/abs/2603.26041","title":"Where and How to Prune: An Empirical Study of Visual Token Pruning for GUI Agent Navigation","topics":["efficiency","interaction-history","screen-understanding"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2603.26211","kind":"paper","record_sha256":"e2965e4f1fa2f61613cfd22413d2553b65733b243c214f8cebf45b3a20827cb3","scope":"core","source_url":"https://arxiv.org/abs/2603.26211","title":"Towards GUI Agents: Vision-Language Diffusion Models for GUI Grounding","topics":["grounding","action-prediction","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.26266","kind":"paper","record_sha256":"e293f64f788b1bd59dfb112a8455693d8ad52fbcebf0a4b3e28d93e53f9a4331","scope":"core","source_url":"https://arxiv.org/abs/2603.26266","title":"GUIDE: Resolving Domain Bias in GUI Agents through Real-Time Web Video Retrieval and Plug-and-Play Annotation","topics":["video","planning","grounding","training-free"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.28069","kind":"paper","record_sha256":"aaf85228cb9897bc43be2859b808aa634de7d72e5f285c9a43a694da9c03fd8e","scope":"adjacent","source_url":"https://arxiv.org/abs/2603.28069","title":"MolmoPoint: Better Pointing for VLMs with Grounding Tokens","topics":["grounding","screen-understanding","video","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2603.29020","kind":"paper","record_sha256":"5ae26fcd283c147d8a0195959520a43dd4e76146a94dd6cebb294652d0c7b67f","scope":"core","source_url":"https://arxiv.org/abs/2603.29020","title":"Emergence WebVoyager: Toward Consistent and Transparent Evaluation of (Web) Agents in The Wild","topics":["evaluation","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2603.29318","kind":"paper","record_sha256":"fb930e8a3767c5e004563063e24690677036b0937bb2579c567d99f65dcae52e","scope":"core","source_url":"https://arxiv.org/abs/2603.29318","title":"PSPA-Bench: A Personalized Benchmark for Smartphone GUI Agent","topics":["evaluation","mobile-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.00886","kind":"paper","record_sha256":"aeea72180720d0f9a6d2f24f5c290ad63ed4f4250cce135b6c7b86ad1db8a7c5","scope":"adjacent","source_url":"https://arxiv.org/abs/2604.00886","title":"PixelPrune: Pixel-Level Adaptive Visual Token Reduction via Predictive Coding","topics":["efficiency","screen-understanding","training-free","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.01676","kind":"paper","record_sha256":"ca65b04d0d7b0741dfc67f0607d87100e1a02e7fc27e614f77e60876c18d054e","scope":"core","source_url":"https://arxiv.org/abs/2604.01676","title":"GPA: Learning GUI Process Automation from Demonstrations","topics":["automation","workflow","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.02345","kind":"paper","record_sha256":"e04dcec5e0e0011b881c52ce0758f83432f6349f7ba5594469bd350f30eea00e","scope":"core","source_url":"https://arxiv.org/abs/2604.02345","title":"UI-Oceanus: Scaling GUI Agents with Synthetic Environmental Dynamics","topics":["world-models","data-generation","state-tracking","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.02623","kind":"paper","record_sha256":"6c577aa28b48f3980bcfdf821fbd82119f407ed7005d8f7cbb8368319cff4d61","scope":"core","source_url":"https://arxiv.org/abs/2604.02623","title":"Poison Once, Exploit Forever: Environment-Injected Memory Poisoning Attacks on Web Agents","topics":["safety-security","memory","web-navigation"],"year":"2026"},{"environments":["web","android"],"id":"arxiv:2604.04399","kind":"paper","record_sha256":"b986017ebe90bbe7370af6a6398c0259da76fc864cbdc76b605befa50d61f57d","scope":"core","source_url":"https://arxiv.org/abs/2604.04399","title":"GUIDE: Interpretable GUI Agent Evaluation via Hierarchical Diagnosis","topics":["evaluation","long-horizon","state-tracking"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.04918","kind":"paper","record_sha256":"08a80713f03f5c1dfae41b843cd1cddb8c8fa3fa9f9c315f3fd61684e8516e23","scope":"core","source_url":"https://arxiv.org/abs/2604.04918","title":"Comparing Human Oversight Strategies for Computer-Use Agents","topics":["human-in-the-loop","safety-security","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2604.05157","kind":"paper","record_sha256":"0244122f2bd145dac87a288ed170edf0976a52aae80a2f26287f03492408121b","scope":"core","source_url":"https://arxiv.org/abs/2604.05157","title":"IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents","topics":["reward-models","planning","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2604.05477","kind":"paper","record_sha256":"dc8cb3b904788f64b46244e5bd9696b5ff9f9cfb3a4d47cef062424fa7a48817","scope":"core","source_url":"https://arxiv.org/abs/2604.05477","title":"Don't Act Blindly: Robust GUI Automation via Action-Effect Verification and Self-Correction","topics":["robustness","reinforcement-learning","state-tracking"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.06126","kind":"paper","record_sha256":"55cb63c0bc78509ccd58cea8251cd7dfe369e5f794acd4c140a433a201a603a8","scope":"core","source_url":"https://arxiv.org/abs/2604.06126","title":"Gym-Anything: Turn any Software into an Agent Environment","topics":["data-generation","sandbox","long-horizon","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.06240","kind":"paper","record_sha256":"d56e72dde160b19dc42d6afce590160c481a644afd5217f46e700107fde629e6","scope":"core","source_url":"https://arxiv.org/abs/2604.06240","title":"The Art of Building Verifiers for Computer Use Agents","topics":["reward-models","evaluation","long-horizon"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.06367","kind":"paper","record_sha256":"036b4c74bf1c6b49be2d2cd0d48b3560deb71d273eb271950473815ce7ecfda4","scope":"core","source_url":"https://arxiv.org/abs/2604.06367","title":"WebSP-Eval: Evaluating Web Agents on Website Security and Privacy Tasks","topics":["safety-security","evaluation","web-navigation","state-tracking"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.06995","kind":"paper","record_sha256":"fadecef84a2f3defb0d036be455c6cd90532ff44f91b077b1930c7cf79cb23c9","scope":"core","source_url":"https://arxiv.org/abs/2604.06995","title":"What's Missing in Screen-to-Action? Towards a UI-in-the-Loop Paradigm for Multimodal GUI Reasoning","topics":["screen-understanding","grounding","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2604.07277","kind":"paper","record_sha256":"26208294d74f068cb748ef66c66f2e066e193f3f3081a69b0cdd8f1e178bd120","scope":"core","source_url":"https://arxiv.org/abs/2604.07277","title":"Android Coach: Improve Online Agentic Training Efficiency with Single State Multiple Actions","topics":["reinforcement-learning","reward-models","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.07831","kind":"paper","record_sha256":"dd61494eb92093521689a3d91eb3c6315686de0769ff779f420c6f1d8ad748b8","scope":"core","source_url":"https://arxiv.org/abs/2604.07831","title":"Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element Injection","topics":["safety-security","grounding","robustness","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.08516","kind":"paper","record_sha256":"82270cb3545f6c81a4ce380feff3c8906af41174808d82db0179cde3971c4083","scope":"core","source_url":"https://arxiv.org/abs/2604.08516","title":"MolmoWeb: Open Visual Web Agent and Open Data for the Open Web","topics":["web-navigation","data-generation","imitation-learning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.09155","kind":"paper","record_sha256":"81570060a6c32d78c925e7b3006dbbec62ac2fc8d12dc75959096168dec9f16d","scope":"core","source_url":"https://arxiv.org/abs/2604.09155","title":"CORA: Conformal Risk-Controlled Agents for Safeguarded Mobile GUI Automation","topics":["safety-security","mobile-navigation","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.09581","kind":"paper","record_sha256":"655e84b8fcc11bceb9202db33f87ed38ae8cddc621d227cf37598433908a16fc","scope":"core","source_url":"https://arxiv.org/abs/2604.09581","title":"Avenir-UX: Automated UX Evaluation via Simulated Human Web Interaction with GUI Grounding","topics":["automation","evaluation","ui-design","grounding"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2604.09815","kind":"paper","record_sha256":"078b26541e5bf3d6b3cd12333e4131b9ff75f75d6d2c0bc0b0f6cf86cab27fd7","scope":"core","source_url":"https://arxiv.org/abs/2604.09815","title":"EE-MCP: Self-Evolving MCP-GUI Agents via Automated Environment Generation and Experience Learning","topics":["tool-integration","data-generation","imitation-learning","memory"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.09937","kind":"paper","record_sha256":"7c0fe188ca592cb891d244cd8b93695aa5ed26c6b398d2543738ce91d46a1f6e","scope":"core","source_url":"https://arxiv.org/abs/2604.09937","title":"HealthAdminBench: Evaluating Computer-Use Agents on Healthcare Administration Tasks","topics":["evaluation","workflow","industry"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.10988","kind":"paper","record_sha256":"beae6c6ae8a4aa48993174aa43c987a77c1071b9039d49fee46492e05482bc1e","scope":"core","source_url":"https://arxiv.org/abs/2604.10988","title":"WebForge: Breaking the Realism-Reproducibility-Scalability Trilemma in Browser Agent Benchmark","topics":["data-generation","evaluation","web-navigation"],"year":"2026"},{"environments":["android","harmonyos","ios"],"id":"arxiv:2604.11784","kind":"paper","record_sha256":"900cf86fedf96aa3d6e992af4054cb283689a0e180545d9fd5a675591723f351","scope":"core","source_url":"https://arxiv.org/abs/2604.11784","title":"ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agents","topics":["reinforcement-learning","evaluation","tool-integration","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.12284","kind":"paper","record_sha256":"5d6480e271a133a2854a11e245df1014e75fe18c8706a785272f5604a6b233e7","scope":"core","source_url":"https://arxiv.org/abs/2604.12284","title":"WebAgentGuard: A Reasoning-Driven Guard Model for Detecting Prompt Injection Attacks in Web Agents","topics":["safety-security","data-generation","imitation-learning","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.13019","kind":"paper","record_sha256":"bc8e64d3c9fb31bb80fe574692973967a3607ca31317aba1a2c3091e7123bd6b","scope":"core","source_url":"https://arxiv.org/abs/2604.13019","title":"PrecisionCUA: Iterative Visual Refinement for Pixel-Precise Cursor Grounding in Code Editors","topics":["grounding","robustness","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.13318","kind":"paper","record_sha256":"7a40b3e0282244c26ba2feb49f7cf39a0a48a8bd957bdb3c6713392e78ae0569","scope":"core","source_url":"https://arxiv.org/abs/2604.13318","title":"WebXSkill: Skill Learning for Autonomous Web Agents","topics":["memory","planning","web-navigation","automation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2604.13822","kind":"paper","record_sha256":"f4c2a731cc51b534fdbfde60a58b71abfe7539c2218022f06dfdd26bce4307ce","scope":"core","source_url":"https://arxiv.org/abs/2604.13822","title":"UI-Copilot: Advancing Long-Horizon GUI Automation via Tool-Integrated Policy Optimization","topics":["tool-integration","memory","planning","long-horizon","mobile-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.14113","kind":"paper","record_sha256":"852b6aaf4caf53f087a8bfbedd31b9d2f0b93b5ad3a1a26914cca06f05ff4db2","scope":"core","source_url":"https://arxiv.org/abs/2604.14113","title":"UI-Zoomer: Uncertainty-Driven Adaptive Zoom-In for GUI Grounding","topics":["grounding","training-free","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.14262","kind":"paper","record_sha256":"ebe4630abf9571a9d8127ccb303b81bd236e88f16b04046018413c5667784955","scope":"core","source_url":"https://arxiv.org/abs/2604.14262","title":"GUI-Perturbed: Domain Randomization Reveals Systematic Brittleness in GUI Grounding Models","topics":["grounding","robustness","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.14668","kind":"paper","record_sha256":"6ebc99d864dda93c2067db9b179a445c9ff55221469d7b606393b61e8cb8e6a6","scope":"core","source_url":"https://arxiv.org/abs/2604.14668","title":"Beyond Chat and Clicks: GUI Agents for In-Situ Assistance via Live Interface Transformation","topics":["human-in-the-loop","ui-design","grounding","evaluation"],"year":"2026"},{"environments":["web","desktop","cross-platform"],"id":"arxiv:2604.14956","kind":"paper","record_sha256":"f866d78beac607b5fc075291af73b80f30b83c4feb259f9edb9bcb9a91cf5258","scope":"core","source_url":"https://arxiv.org/abs/2604.14956","title":"FedGUI: Benchmarking Federated GUI Agents across Heterogeneous Platforms, Devices, and Operating Systems","topics":["evaluation","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2604.15093","kind":"paper","record_sha256":"50cfeb6fa986b31863ae867dd5a2b4b79e4f7f0961bb1fb4466bec3b5cb82ef8","scope":"core","source_url":"https://arxiv.org/abs/2604.15093","title":"OpenMobile: Building Open Mobile Agents with Task and Trajectory Synthesis","topics":["data-generation","imitation-learning","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.16520","kind":"paper","record_sha256":"4fdd68c6f25f92c38cb37fa72549b935ec52a7ed6ec0390eb8a387dcf38cd38c","scope":"adjacent","source_url":"https://arxiv.org/abs/2604.16520","title":"AgentClick: A Skill-Based Human-in-the-Loop Review Layer for Terminal AI Agents","topics":["human-in-the-loop","observability","workflow","memory","tool-integration"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.17284","kind":"paper","record_sha256":"818f8fe1ee0878fa5da8031cfcae3dcff6082f78f0a9a0629771133924761329","scope":"core","source_url":"https://arxiv.org/abs/2604.17284","title":"HalluClear: Diagnosing, Evaluating and Mitigating Hallucinations in GUI Agents","topics":["robustness","grounding","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2604.17817","kind":"paper","record_sha256":"d476da7b43fcf60342f2d8b10c02fa6801e7f5ccb4a14ac92930d8bc606688a0","scope":"core","source_url":"https://arxiv.org/abs/2604.17817","title":"Do LLMs Need to See Everything? A Benchmark and Study of Failures in LLM-driven Smartphone Automation using Screentext vs. Screenshots","topics":["evaluation","mobile-navigation","screen-understanding","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.17821","kind":"paper","record_sha256":"f4fbbbd45f5a8ad881eb03a64ca96d4737d773fca79cbd74a9ee2793d6f684ec","scope":"core","source_url":"https://arxiv.org/abs/2604.17821","title":"WebUncertainty: Dual-Level Uncertainty Driven Planning and Reasoning For Autonomous Web Agent","topics":["planning","web-navigation","long-horizon","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2604.17849","kind":"paper","record_sha256":"4e345ed75c845e668b4a20ac3260eff31b21bef4f8315f0f0e1bd5ce5f354fe0","scope":"core","source_url":"https://arxiv.org/abs/2604.17849","title":"On the Reliability of Computer Use Agents","topics":["evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.18240","kind":"paper","record_sha256":"855b87502be62ae6c462c154e5cf2c07c862e8a4cd0a889c3ddcc8fa44921357","scope":"core","source_url":"https://arxiv.org/abs/2604.18240","title":"AJ-Bench: Benchmarking Agent-as-a-Judge for Environment-Aware Evaluation","topics":["evaluation","reward-models","state-tracking"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2604.18860","kind":"paper","record_sha256":"cba4e9b361946feaab32fc604e0715c5d69677576b45c4dae9e4213776616eb9","scope":"core","source_url":"https://arxiv.org/abs/2604.18860","title":"Temporal UI State Inconsistency in Desktop GUI Agents: Formalizing and Defending Against TOCTOU Attacks on Computer-Use Agents","topics":["safety-security","robustness","evaluation","state-tracking"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.19742","kind":"paper","record_sha256":"106903476799a5dc9612c9b37bf4d87f38eaef16f021c8517e556d8868acd855","scope":"core","source_url":"https://arxiv.org/abs/2604.19742","title":"PlayCoder: Making LLM-Generated GUI Code Playable","topics":["evaluation","automation","ui-design","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.21268","kind":"paper","record_sha256":"8019089614f272b40fff4482b86e63a1beb3fb111bc584b53b7ec63b71074be5","scope":"core","source_url":"https://arxiv.org/abs/2604.21268","title":"Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding","topics":["grounding","reinforcement-learning","self-reflection"],"year":"2026"},{"environments":["desktop","linux","windows"],"id":"arxiv:2604.21375","kind":"paper","record_sha256":"13e49e4bb114cc33acee2624d1457ece3b4a0d5a099171b8290fc2fe236a3bd2","scope":"core","source_url":"https://arxiv.org/abs/2604.21375","title":"VLAA-GUI: Knowing When to Stop, Recover, and Search, A Modular Framework for GUI Automation","topics":["robustness","self-reflection","tool-integration","evaluation"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2604.23941","kind":"paper","record_sha256":"2a802f207650b87b77463aea03c6fe280057505910ee4a908bf7cb7f23d2f8f8","scope":"core","source_url":"https://arxiv.org/abs/2604.23941","title":"GoClick: Lightweight Element Grounding Model for Autonomous GUI Interaction","topics":["grounding","efficiency"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2604.24441","kind":"paper","record_sha256":"f38d0c7d1a2545bdffc24be108516d6245892973512c4d0bea9438e759c0b08f","scope":"core","source_url":"https://arxiv.org/abs/2604.24441","title":"AutoGUI-v2: A Comprehensive Multi-Modal GUI Functionality Understanding Benchmark","topics":["screen-understanding","grounding","state-tracking","evaluation","data-generation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.24964","kind":"paper","record_sha256":"a0a7f7db1d58451d7aa9b5e634100ed51a0a8bd38ae0742023ffd0f81ea3a9c1","scope":"core","source_url":"https://arxiv.org/abs/2604.24964","title":"Odysseys: Benchmarking Web Agents on Realistic Long Horizon Tasks","topics":["evaluation","long-horizon","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.25380","kind":"paper","record_sha256":"3bb88b570435e62691eb187ba548527f77d70ba7ce91cf11395fc0410900c195","scope":"core","source_url":"https://arxiv.org/abs/2604.25380","title":"Benchmarking and Improving GUI Agents in High-Dynamic Environments","topics":["video","state-tracking","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.25562","kind":"paper","record_sha256":"68e3bb225ccc8bd5ee3566c7c009bcaa021ae800ab4144f0be7958255fb624f3","scope":"core","source_url":"https://arxiv.org/abs/2604.25562","title":"SnapGuard: Lightweight Prompt Injection Detection for Screenshot-Based Web Agents","topics":["safety-security","screen-understanding","evaluation","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.26020","kind":"paper","record_sha256":"20c053661727f04591c88ab1dbe22e0c8cd86b136b54723896fdbdfd50039222","scope":"core","source_url":"https://arxiv.org/abs/2604.26020","title":"Training Computer Use Agents to Assess the Usability of Graphical User Interfaces","topics":["ui-design","evaluation","automation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.26148","kind":"paper","record_sha256":"d63c7c59775f2bd26842875079cfc8c21fd8f4f445361994dc26f32e3f7d18df","scope":"core","source_url":"https://arxiv.org/abs/2604.26148","title":"Beyond Screenshots: Evaluating VLMs' Understanding of UI Animations","topics":["evaluation","screen-understanding","video","ui-design"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.27151","kind":"paper","record_sha256":"ba06ccde0a95bc8ea8a1f0daa93d5ecd3b1ded2dd96564ed6f7ea252f27cd9a0","scope":"core","source_url":"https://arxiv.org/abs/2604.27151","title":"Step-level Optimization for Efficient Computer-use Agents","topics":["efficiency","robustness","state-tracking","interaction-history","long-horizon"],"year":"2026"},{"environments":["web"],"id":"arxiv:2604.27253","kind":"paper","record_sha256":"bec2ced8048d13272f35c5243a7285181cabc973b7ecf956e65fe1d2155e1373","scope":"core","source_url":"https://arxiv.org/abs/2604.27253","title":"AutoSurfer -- Teaching Web Agents through Comprehensive Surfing, Learning, and Modeling","topics":["web-navigation","data-generation","imitation-learning","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2604.27776","kind":"paper","record_sha256":"0fee642849e1e323adb4b8061bd91109fa310379f1569fb549a17c6da2b69907","scope":"core","source_url":"https://arxiv.org/abs/2604.27776","title":"WindowsWorld: A Process-Centric Benchmark of Autonomous GUI Agents in Professional Cross-Application Environments","topics":["evaluation","cross-app","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2604.27996","kind":"paper","record_sha256":"c42e4bfb381e9bfd953028f26132730f426a171716f4f778859821d06ac6194b","scope":"core","source_url":"https://arxiv.org/abs/2604.27996","title":"Exploring LLM Agent Designs and Interaction Modalities for Scientific Visualization","topics":["science","evaluation","tool-integration","memory"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.00551","kind":"paper","record_sha256":"63f4c2802859e81745df565cd0cd83ceafac0671215a5b3f1277e7c7961741fa","scope":"core","source_url":"https://arxiv.org/abs/2605.00551","title":"A11y-Compressor: A Framework for Enhancing the Efficiency of GUI Agent Observations through Visual Context Reconstruction and Redundancy Reduction","topics":["screen-understanding","efficiency","grounding","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.01208","kind":"paper","record_sha256":"c089f9ed3082aac35604208f7497b511884a4fe2c9c7d703f836445394ff7935","scope":"core","source_url":"https://arxiv.org/abs/2605.01208","title":"Faithful Mobile GUI Agents with Guided Advantage Estimator","topics":["reinforcement-learning","robustness","mobile-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.02630","kind":"paper","record_sha256":"3aaa40cd1fb65d97560851ab3cf688dc7a3645adb15595204f5eb83e0101fb46","scope":"core","source_url":"https://arxiv.org/abs/2605.02630","title":"AutoFocus: Uncertainty-Aware Active Visual Search for GUI Grounding","topics":["grounding","training-free","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.06761","kind":"paper","record_sha256":"d373b7675b3b84f5cdcf931ca856b522f8f8c64b27ffcc579c98e96cd7ba32e6","scope":"core","source_url":"https://arxiv.org/abs/2605.06761","title":"Weblica: Scalable and Reproducible Training Environments for Visual Web Agents","topics":["sandbox","data-generation","reinforcement-learning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.07134","kind":"paper","record_sha256":"72043f97296b29ddf5cfa8a552ca448bb0f45c69f5666baa05ed469a2e284251","scope":"core","source_url":"https://arxiv.org/abs/2605.07134","title":"Region4Web: Rethinking Observation Space Granularity for Web Agents","topics":["screen-understanding","web-navigation","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.09317","kind":"paper","record_sha256":"3cb954e0b5d1a4dfd80bcf0ba2a6c74e1181d7cdea3d64b454141c1e4b448656","scope":"core","source_url":"https://arxiv.org/abs/2605.09317","title":"Mem-W: Latent Memory-Native GUI Agents","topics":["memory","long-horizon","imitation-learning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.09497","kind":"paper","record_sha256":"519f011ac9dffffa2acc0d21961cd4be355f372c361b83b21399b316f95a6c7f","scope":"core","source_url":"https://arxiv.org/abs/2605.09497","title":"Don't Click That: Teaching Web Agents to Resist Deceptive Interfaces","topics":["safety-security","reinforcement-learning","evaluation","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2605.10347","kind":"paper","record_sha256":"2d797608c1f015105cc8f645a39441ab76385157af77f1570d39450c08f8d881","scope":"core","source_url":"https://arxiv.org/abs/2605.10347","title":"How Mobile World Model Guides GUI Agents?","topics":["world-models","data-generation","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.10966","kind":"paper","record_sha256":"7ee4fdf850ba62139409b8a32bd6dc67c0fb247698ea5143ca7fcc0f0562a63c","scope":"adjacent","source_url":"https://arxiv.org/abs/2605.10966","title":"MMTB: Evaluating Terminal Agents on Multimedia-File Tasks","topics":["evaluation","automation","video","tool-integration"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2605.11212","kind":"paper","record_sha256":"91ddc5efc22942158ed9de3821e080424e8b50a4b7032467e9559bf83437667a","scope":"core","source_url":"https://arxiv.org/abs/2605.11212","title":"ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction","topics":["efficiency","interaction-history","screen-understanding"],"year":"2026"},{"environments":["android"],"id":"arxiv:2605.12294","kind":"paper","record_sha256":"771818b6633dae56eea1bc35312251f6e7ade5c6c0af64a5238927bac1e85c1c","scope":"core","source_url":"https://arxiv.org/abs/2605.12294","title":"Executable Agentic Memory for GUI Agent","topics":["memory","planning","long-horizon","efficiency"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2605.12481","kind":"paper","record_sha256":"47a4e1a8882a783a9233fcf84f3589de8eefee6ca8666961a6ae4b8146bf8865","scope":"core","source_url":"https://arxiv.org/abs/2605.12481","title":"ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents","topics":["tool-integration","data-generation","imitation-learning","reinforcement-learning","planning","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.12493","kind":"paper","record_sha256":"b8bdaad7743a3e6f2e597c1cb45926f70ac7875cfeb6f664dc203419a788ad2d","scope":"core","source_url":"https://arxiv.org/abs/2605.12493","title":"LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues","topics":["memory","interaction-history","state-tracking","workflow","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.12501","kind":"paper","record_sha256":"99491a4c5af3e5e465da6744f948ed92d9e37e1f7a6354588fc4cfc0807f8cf7","scope":"core","source_url":"https://arxiv.org/abs/2605.12501","title":"Covering Human Action Space for Computer Use: Data Synthesis and Benchmark","topics":["grounding","data-generation","evaluation","action-prediction"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.12549","kind":"paper","record_sha256":"8771ac541ed643c7e8d1017790a170a56cbf4460b57107c74341b89dde55ff0f","scope":"core","source_url":"https://arxiv.org/abs/2605.12549","title":"What Happens Before Decoding? Prefill Determines GUI Grounding in VLMs","topics":["grounding","training-free","screen-understanding","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.14290","kind":"paper","record_sha256":"ab4afe11528ac80cb00d4bf5479c3142c411b0c20917f5579f6ef0e8d7a5d61f","scope":"adjacent","source_url":"https://arxiv.org/abs/2605.14290","title":"Web Agents Should Adopt the Plan-Then-Execute Paradigm","topics":["planning","safety-security","tool-integration"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.15542","kind":"paper","record_sha256":"afce34c1f8cfbf1fa3fdb6ecd722c1b67aee77261991ae3fd97de2f91dbd569a","scope":"core","source_url":"https://arxiv.org/abs/2605.15542","title":"DRS-GUI: Dynamic Region Search for Training-Free GUI Grounding","topics":["grounding","training-free","planning","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.15777","kind":"paper","record_sha256":"c6c78bab164a590983b6f91919c778575040527f8c94ad2396450642e3d99efe","scope":"core","source_url":"https://arxiv.org/abs/2605.15777","title":"SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows?","topics":["web-navigation","evaluation","long-horizon","workflow","cross-app","state-tracking"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2605.16402","kind":"paper","record_sha256":"2672e1aad43c2328210792a899cc0d50f6981cc58de3eddc5ea53720a22be3c7","scope":"core","source_url":"https://arxiv.org/abs/2605.16402","title":"WinDeskGround: A Benchmark for Robust GUI Grounding in Complex Multi-Window Desktop Environments","topics":["grounding","evaluation","robustness","data-generation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2605.16883","kind":"paper","record_sha256":"6c6ad0042e00e431fc6b7c671009314ec04095ce06667c02edde0e3522227a8a","scope":"core","source_url":"https://arxiv.org/abs/2605.16883","title":"SE-GA: Memory-Augmented Self-Evolution for GUI Agents","topics":["memory","interaction-history","mobile-navigation","grounding","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.17320","kind":"paper","record_sha256":"b928fa01ef1003bfc977ed43f6d360eee804487d6f7bbcb61349b8bde1c8aa55","scope":"core","source_url":"https://arxiv.org/abs/2605.17320","title":"TClone: Low-Latency Forking of Live GUI Environments for Computer-Use Agents","topics":["sandbox","efficiency","safety-security","automation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.17656","kind":"paper","record_sha256":"44042d6ce792b1b0303ffbe19545f420bea8647be0bc4076065e1645b373c6f8","scope":"core","source_url":"https://arxiv.org/abs/2605.17656","title":"MUIAnno: An Expert-Annotated Dataset and Evaluation Benchmark for Mobile UI Understanding","topics":["screen-understanding","grounding","evaluation","data-generation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.18048","kind":"paper","record_sha256":"45cc06cbff5a0e6c27a672a4520ad3f4dd3223c32b8a15c4fe5ca63b2090ca90","scope":"core","source_url":"https://arxiv.org/abs/2605.18048","title":"DocOS: Towards Proactive Document-Guided Actions in GUI Agents","topics":["web-navigation","grounding","evaluation","planning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.18652","kind":"paper","record_sha256":"cbcdc53f80c54d943ffaf145d84ff88d9226b23732732df8dc2c3179aabc68b6","scope":"core","source_url":"https://arxiv.org/abs/2605.18652","title":"MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI Agents","topics":["memory","interaction-history","long-horizon","data-generation","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.18758","kind":"paper","record_sha256":"582bac0df8bdebc81d003d69dfe87392dd23cf4404355a7d6dc7513d2b4de84d","scope":"core","source_url":"https://arxiv.org/abs/2605.18758","title":"OmniGUI: Benchmarking GUI Agents in Omni-Modal Smartphone Environments","topics":["evaluation","action-prediction","video","mobile-navigation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2605.19769","kind":"paper","record_sha256":"a2aea10082e355274489f6146bab38b12e534a21018abddaa538015f1c3881f1","scope":"core","source_url":"https://arxiv.org/abs/2605.19769","title":"OpenComputer: Verifiable Software Worlds for Computer-Use Agents","topics":["evaluation","data-generation","observability"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.20291","kind":"paper","record_sha256":"6dd5ef3b22a9ecaa8d97924857c6fe5a874734e9ba0690876d71b3b6d57f0a2d","scope":"core","source_url":"https://arxiv.org/abs/2605.20291","title":"Weasel: Out-of-Domain Generalization for Web Agents via Importance-Diversity Data Selection","topics":["web-navigation","imitation-learning","efficiency","robustness","data-generation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.21470","kind":"paper","record_sha256":"d6bfd949f58e31a61b3be4de2efd2348d87fff4cf9a68d99d17e06c6f78017fb","scope":"core","source_url":"https://arxiv.org/abs/2605.21470","title":"Agent JIT Compilation for Latency-Optimizing Web Agent Planning and Scheduling","topics":["planning","efficiency","tool-integration","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.25160","kind":"paper","record_sha256":"a0a0e79b5f4849a090efbaea057b092cbff8e33d6c7bf151063e44539064611f","scope":"core","source_url":"https://arxiv.org/abs/2605.25160","title":"ScaleWoB: Guiding GUI Agents with Coding Agents via Large-Scale Environmental Synthesis","topics":["data-generation","evaluation","sandbox"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2605.25624","kind":"paper","record_sha256":"53a7c7a9aba824d84c6c5bea606ee37c1076beffb8bd4dbb42c933e8464aaa53","scope":"core","source_url":"https://arxiv.org/abs/2605.25624","title":"CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents","topics":["data-generation","reinforcement-learning","sandbox"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.26086","kind":"paper","record_sha256":"ce782e16fc3852d260441052d8f128b4e78f0ab3f3e26a6295acb270254cd194","scope":"core","source_url":"https://arxiv.org/abs/2605.26086","title":"Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World","topics":["evaluation","interaction-history","long-horizon","robustness","data-generation","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.26114","kind":"paper","record_sha256":"26172989b7f02fe0b2d545716e7aedeefb718b1583c261fdfc11d4c50fcf8388","scope":"core","source_url":"https://arxiv.org/abs/2605.26114","title":"MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research","topics":["sandbox","evaluation","reinforcement-learning"],"year":"2026"},{"environments":["web","android"],"id":"arxiv:2605.26144","kind":"paper","record_sha256":"e1c025d856f279a01a347fbfd5053a239f6c19752a23bce94edcd9dad2e51ca1","scope":"adjacent","source_url":"https://arxiv.org/abs/2605.26144","title":"VISTA: An End-to-End Benchmark for Visual Spec-to-Web-App Coding Agents","topics":["evaluation","ui-design","automation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2605.26546","kind":"paper","record_sha256":"5e5b1638fb5766bda05153cbf1d51b480fb6f9d664fc6cfc9bd6460ee512dee3","scope":"core","source_url":"https://arxiv.org/abs/2605.26546","title":"MobileExplorer: Accelerating On-Device Inference for Mobile GUI Agents via Online Exploration","topics":["mobile-navigation","efficiency","memory","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2605.27761","kind":"paper","record_sha256":"99dbe09a5e30f98e64edf1242ab5f5fa8ee580714a3df887f59819a52ecb498b","scope":"core","source_url":"https://arxiv.org/abs/2605.27761","title":"AndroidDaily: A Verifiable Benchmark for Mobile GUI Agents on Real-World Closed-Source Applications","topics":["mobile-navigation","evaluation","long-horizon","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.28116","kind":"paper","record_sha256":"75eb154ab63a6d214fce1c906329e0be66d356a985e1d6d97f305888718e2ccf","scope":"core","source_url":"https://arxiv.org/abs/2605.28116","title":"MIRAGE: Context-Aware Prompt Injection against Mobile GUI Agents via User-Generated Content","topics":["safety-security","data-generation","evaluation","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2605.28258","kind":"paper","record_sha256":"bdab83686508e4e3b34474d9d59d8ec94f7bc0c439eeef2bd48160a58653732e","scope":"adjacent","source_url":"https://arxiv.org/abs/2605.28258","title":"GUI Agents for Continual Game Generation","topics":["evaluation","automation","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.28534","kind":"paper","record_sha256":"e3ba18877b4e2712f136b77cf97904b82e0d85a8ce6702a85f3dc6efa40eb813","scope":"core","source_url":"https://arxiv.org/abs/2605.28534","title":"GUI-CIDER: Mid-training GUI Agents via Causal Internalization and Density-aware Exemplar Reselection","topics":["data-generation","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.28646","kind":"paper","record_sha256":"b9c3e782eee68b505df7f6c1bda1ac47a7974c59f53ce526f3bc4e7b6b7d3a19","scope":"core","source_url":"https://arxiv.org/abs/2605.28646","title":"MaskClaw: Edge-Side Personalized Privacy Arbitration for GUI Agents with Behavior-Driven Skill Evolution","topics":["safety-security","memory","human-in-the-loop","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2605.29486","kind":"paper","record_sha256":"2188fbdaca19816759036694939310d297999b48062065cf272bbc416bbc31a1","scope":"core","source_url":"https://arxiv.org/abs/2605.29486","title":"PhoneWorld: From Real-App Trajectories to Dynamic and Verifiable Environments for Phone-Use Agents","topics":["data-generation","sandbox","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2605.29534","kind":"paper","record_sha256":"5d92a25468ccaddd6ca2310fd506dd792b45add3ced9404990aef28384628164","scope":"core","source_url":"https://arxiv.org/abs/2605.29534","title":"UI-KOBE: Knowledge-Oriented Behavior Exploration for Lightweight Graph-Guided GUI Agents","topics":["mobile-navigation","planning","state-tracking","memory","efficiency"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2606.01533","kind":"paper","record_sha256":"2d489d32cee9c4d379f39974a52d1b2bb5981410ee9af26283d0bf34b8261db1","scope":"core","source_url":"https://arxiv.org/abs/2606.01533","title":"Multi-Agent Computer Use","topics":["planning","long-horizon","workflow","state-tracking","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.01790","kind":"paper","record_sha256":"e9a3e3edc9b49da2c07adef50be3aea49ad2c87d59d9d1d263591c39e8aebdb7","scope":"core","source_url":"https://arxiv.org/abs/2606.01790","title":"STaR-KV: Spatio-Temporal Adaptive Re-weighting for KV Cache Compression in GUI Vision-Language Models","topics":["efficiency","training-free","screen-understanding","interaction-history"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.03103","kind":"paper","record_sha256":"6edbd416136e24f2aeef3fbbad6ad04b6c14a5df594e38ed466ef29706b6dad6","scope":"core","source_url":"https://arxiv.org/abs/2606.03103","title":"DeskCraft: Benchmarking Desktop Agents on Professional Workflows and Human-in-the-Loop Collaboration","topics":["evaluation","human-in-the-loop","long-horizon","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.03203","kind":"paper","record_sha256":"0c2d700b3747dc4ea296f2b220ad4471e741ecc13e3430a21bba4aba8230b378","scope":"core","source_url":"https://arxiv.org/abs/2606.03203","title":"MedCUA-Bench: A Screenshot-Only Benchmark for Clinical Computer-Use Agents","topics":["evaluation","safety-security","science","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.03854","kind":"paper","record_sha256":"c80c86c54f44aeb7d8e1ad78171e38048bb35a828ebeb93dfbb2846e2d1223f5","scope":"adjacent","source_url":"https://arxiv.org/abs/2606.03854","title":"CLI-Anything: Towards Agent-Native Computer Use","topics":["automation","tool-integration","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.04391","kind":"paper","record_sha256":"5214a72544f390762af7a6621f6586cf71b2d76d0d0c7f0f236ccce8a8253e88","scope":"core","source_url":"https://arxiv.org/abs/2606.04391","title":"Online Skill Learning for Web Agents via State-Grounded Dynamic Retrieval","topics":["web-navigation","memory","interaction-history","planning","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.04701","kind":"paper","record_sha256":"74c936a706af85849da607cd8ac4ccb0ed4de7a2d2974e32b5f64eda558b7384","scope":"core","source_url":"https://arxiv.org/abs/2606.04701","title":"Benchmarking Living-Screen-Native GUI Agents on Short-Video Platforms","topics":["evaluation","video","efficiency","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.05233","kind":"paper","record_sha256":"c98bd8ac440ca7ac870f811ef1ab6731733561dd7280375abf0c04e8c0d3fb97","scope":"core","source_url":"https://arxiv.org/abs/2606.05233","title":"Domain-Conditioned Safety in Frontier Computer-Using Agents: A 793-Episode Browser Benchmark, a Coding-Domain Cross-Reference, and a Reproducibility Audit of Recent Red-Teaming","topics":["safety-security","evaluation","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.05597","kind":"paper","record_sha256":"57950fdb51e6aa4cc868efb6ae937f7f6ba35de15094ebac00ba52da40704982","scope":"core","source_url":"https://arxiv.org/abs/2606.05597","title":"AsyncWebRL: Efficient Asynchronous Reinforcement Learning for Multi-Step Visual Web Agents","topics":["reinforcement-learning","efficiency","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.06322","kind":"paper","record_sha256":"30ce1ae8a92347047a90503ab613ada6d53c614dba0159693f895a640b878f91","scope":"core","source_url":"https://arxiv.org/abs/2606.06322","title":"DragOn: A Benchmark and Dataset for Drag-Based GUI Interactions","topics":["grounding","data-generation","evaluation","screen-understanding"],"year":"2026"},{"environments":["desktop","macos"],"id":"arxiv:2606.06560","kind":"paper","record_sha256":"c041ae786197e3f13b96759d1951acfe31dda0959348858611d57f83c3ce3178","scope":"core","source_url":"https://arxiv.org/abs/2606.06560","title":"MacArena: Benchmarking Computer Use Agents on an Online macOS Environment","topics":["evaluation","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.06708","kind":"paper","record_sha256":"abc20a16cd209d597b698c253d16acb45dfea7809d684e40d831d1d0776583cd","scope":"core","source_url":"https://arxiv.org/abs/2606.06708","title":"Signal-Driven Observation for Long-Horizon Web Agents","topics":["efficiency","screen-understanding","long-horizon"],"year":"2026"},{"environments":["android"],"id":"arxiv:2606.07027","kind":"paper","record_sha256":"46683990cfd4a27697db73c49ccad13b0775ab93c0e5725e7f04fc9b6ee059b2","scope":"core","source_url":"https://arxiv.org/abs/2606.07027","title":"StainFlow: Entity-Stain Tracking and Evidence Linking for Process Rewards in GUI Agents","topics":["reward-models","state-tracking","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["desktop","macos"],"id":"arxiv:2606.07594","kind":"paper","record_sha256":"ffce948e2f1112b5993b338fefb89aca77980225e2081457b3f029d831209dd2","scope":"core","source_url":"https://arxiv.org/abs/2606.07594","title":"Syll: Open-Source Personal Automation with Cross-Surface Execution","topics":["automation","human-in-the-loop","workflow","memory","recording","observability"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.09399","kind":"paper","record_sha256":"1c5a7e865f5181011326de6b36cf96dcdbc1877615a11db01d02511fa5953239","scope":"core","source_url":"https://arxiv.org/abs/2606.09399","title":"RunAgent SuperBrowser: A Theory of Autonomous Web Navigation Grounded in Human Browsing Behaviour","topics":["web-navigation","grounding","planning","memory","efficiency"],"year":"2026"},{"environments":["desktop","linux"],"id":"arxiv:2606.09426","kind":"paper","record_sha256":"8fb8ea40eb8c051e6ecb29cf513250096e9a026635a89f58df062122cc6f45cf","scope":"core","source_url":"https://arxiv.org/abs/2606.09426","title":"WeaveBench: A Long-Horizon, Real-World Benchmark for Computer-Use Agents with Hybrid Interfaces","topics":["evaluation","long-horizon","workflow","tool-integration"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.09447","kind":"paper","record_sha256":"b989ba0d8c8b68dfa607c0454fe78d5951c71cba4c5d5cd4a22e7a14d80dd628","scope":"core","source_url":"https://arxiv.org/abs/2606.09447","title":"AliyunConsoleAgent: Training Web Agents in Real-World Cloud Environments via Distillation and Reinforcement Learning","topics":["web-navigation","imitation-learning","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["ios"],"id":"arxiv:2606.09764","kind":"paper","record_sha256":"2cef6e9c87141bcc3f40a5a7bfa9ea6d9a376dba9b1085d979da1feff6313809","scope":"core","source_url":"https://arxiv.org/abs/2606.09764","title":"iOSWorld: A Benchmark for Personally Intelligent Phone Agents","topics":["evaluation","cross-app","memory"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.10423","kind":"paper","record_sha256":"ea9c471d3a24d9960b59d2e80f55c2fc5e7a4dfe288867c884e5316b8ab2a624","scope":"core","source_url":"https://arxiv.org/abs/2606.10423","title":"WebChallenger: A Reliable and Efficient Generalist Web Agent","topics":["web-navigation","memory","efficiency","training-free","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.10522","kind":"paper","record_sha256":"710a4b81005f786dd9b8c2093cd9dd597723e17adb8946efa81baf1f9e2d108a","scope":"core","source_url":"https://arxiv.org/abs/2606.10522","title":"GUI-AC: Enhancing Continual Learning in GUI Agents","topics":["reinforcement-learning","grounding","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.11042","kind":"paper","record_sha256":"d07aa8dfc7fde468bcdd827caab22b24a0663c746b69c99d0c505c2bf945b8b0","scope":"core","source_url":"https://arxiv.org/abs/2606.11042","title":"Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional Fields","topics":["evaluation","long-horizon","workflow","robustness"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2606.11078","kind":"paper","record_sha256":"fb44084216f61b57f82ec22f5c725ee02aa0d6aa412d0f4d8e7386ef0f6d51ac","scope":"core","source_url":"https://arxiv.org/abs/2606.11078","title":"A History-Aware Visually Grounded Critic for Computer Use Agents","topics":["interaction-history","memory","grounding","reward-models","long-horizon","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.12485","kind":"paper","record_sha256":"c9012af0c92cac33fb786d448368c148d4e568c3394735efe0886238d9b187ce","scope":"core","source_url":"https://arxiv.org/abs/2606.12485","title":"Speculative Rollback Correction for Quality-Diverse Web Agent Imitation","topics":["imitation-learning","data-generation","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2606.12666","kind":"paper","record_sha256":"7a8d68116ca92d32eb8cb251b40bbd6ee77f00d14e73dcfc79b8f78fd5b952ba","scope":"core","source_url":"https://arxiv.org/abs/2606.12666","title":"CAPED: Context-Aware Privacy Exposure Defense for Mobile GUI Agents","topics":["safety-security","screen-understanding","mobile-navigation"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2606.13239","kind":"paper","record_sha256":"d38ff12731317fc65d4f21e6a5dc12042d08e376dcdf185ae16f7aa2d2e58f2c","scope":"adjacent","source_url":"https://arxiv.org/abs/2606.13239","title":"ComAct: Reframing Professional Software Manipulation via COM-as-Action Paradigm","topics":["automation","evaluation","tool-integration","science"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.13385","kind":"paper","record_sha256":"2827cb00bebef84ba6e37ef3c17bf59500990e412c5478cf9b325f1206b458ba","scope":"core","source_url":"https://arxiv.org/abs/2606.13385","title":"Who Pays the Price? Stakeholder-Centric Prompt Injection Benchmarking for Real-world Web Agents","topics":["evaluation","safety-security","web-navigation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.14106","kind":"paper","record_sha256":"443c1533984516fb121d9b836bff0bb47db45e485d8b2c95727e59768b8836b3","scope":"core","source_url":"https://arxiv.org/abs/2606.14106","title":"Naive Visual Memory is Not Enough: A Failure-Mode Study of GUI Agents","topics":["memory","interaction-history","grounding","robustness","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.14579","kind":"paper","record_sha256":"23a6e1e199931115746ee99480b8df9b80dca2d904e9bfdfd774832b755a957c","scope":"core","source_url":"https://arxiv.org/abs/2606.14579","title":"VISTA: View-Consistent Self-Verified Training for GUI Grounding","topics":["grounding","reinforcement-learning","robustness","data-generation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.14832","kind":"paper","record_sha256":"8260206560332a7b3a5c6714b99fbf7d7fcc5f002f1beb379e6e2fe08325fdf7","scope":"core","source_url":"https://arxiv.org/abs/2606.14832","title":"PhoneHarness: Harnessing Phone-Use Agents through Mixed GUI, CLI, and Tool Actions","topics":["mobile-navigation","tool-integration","evaluation","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.15017","kind":"paper","record_sha256":"0bfc2d8abd76f33eb29b4912d5dd85cf6d357cfb98b113a9b209af7766b807bf","scope":"core","source_url":"https://arxiv.org/abs/2606.15017","title":"Are Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web Agents","topics":["memory","efficiency","evaluation","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.15034","kind":"paper","record_sha256":"bd1afa86a5a97f4f28e906b72a5b98418bb8708740595124d233fc261bd29931","scope":"core","source_url":"https://arxiv.org/abs/2606.15034","title":"OSGuard: A Benchmark for Safety in Computer-Use Agents","topics":["safety-security","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.16659","kind":"paper","record_sha256":"5db979400b22323b0460947c876394ed883844d8dbd417f871bb868b95574839","scope":"core","source_url":"https://arxiv.org/abs/2606.16659","title":"FraudSMSWalker: Benchmarking Agentic Large Language Models for SMS-to-Webpage Fraud Detection","topics":["safety-security","evaluation","robustness"],"year":"2026"},{"environments":["desktop","linux","web"],"id":"arxiv:2606.16748","kind":"paper","record_sha256":"c591b1cf7680bf57f9ba00695ad23961da3baa6139ccc052c583ffafc006eb20","scope":"core","source_url":"https://arxiv.org/abs/2606.16748","title":"MyPCBench: A Benchmark for Personally Intelligent Computer-Use Agents","topics":["evaluation","cross-app","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.16802","kind":"paper","record_sha256":"a9a56d5dabc4e7b26c54905a34f501ed384e7dd3f0624b09047d82ab4f825919","scope":"core","source_url":"https://arxiv.org/abs/2606.16802","title":"LabOSBench: Benchmarking Computer Use Agents for Scientific Instrument Control","topics":["evaluation","science","long-horizon"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.17321","kind":"paper","record_sha256":"dedf79737f62155aafa6f633418e9d664c8e983cf08ee4a86595af884c527eb8","scope":"core","source_url":"https://arxiv.org/abs/2606.17321","title":"ProCUA-SFT Technical Report","topics":["data-generation","imitation-learning","evaluation"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2606.17929","kind":"paper","record_sha256":"294239690a41f6e427c6da25bd39959d55cf187cbe46af9e1062dd72d1aba7e8","scope":"core","source_url":"https://arxiv.org/abs/2606.17929","title":"PreAct: Computer-Using Agents that Get Faster on Repeated Tasks","topics":["recording","memory","efficiency","automation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.18101","kind":"paper","record_sha256":"ec21053e0c88d89a4a264c2d485dd3f694eeb6fb48a70bee5947c9a84149bf58","scope":"core","source_url":"https://arxiv.org/abs/2606.18101","title":"Trust the Right Teacher: Quality-Aware Self-Distillation for GUI Grounding","topics":["grounding","imitation-learning","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.18448","kind":"paper","record_sha256":"2ecd265ce3836bac212d857886695c7fab98612d5a6d27a659a2cc8d36111e17","scope":"core","source_url":"https://arxiv.org/abs/2606.18448","title":"VISUALSKILL: Multimodal Skills for Computer-Use Agents","topics":["memory","tool-integration","screen-understanding","long-horizon"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.18671","kind":"paper","record_sha256":"6e666e4247e3376c2ab896d5dcd9ed8a263ad9bee5f5eabe1d24ec25f1bfedb9","scope":"core","source_url":"https://arxiv.org/abs/2606.18671","title":"HANSEL: Extracting Breadcrumbs from Web Agent Trajectories for Interactive Verification","topics":["human-in-the-loop","observability","recording","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.18890","kind":"paper","record_sha256":"6054c25cd17d9b12e3f5134017da0e825c089d2b311da4f70c3ff741b6e615f1","scope":"core","source_url":"https://arxiv.org/abs/2606.18890","title":"Skill-Guided Continuation Distillation for GUI Agents","topics":["imitation-learning","data-generation","robustness","planning"],"year":"2026"},{"environments":["android"],"id":"arxiv:2606.19388","kind":"paper","record_sha256":"c3b0992dfaf99c9f9969d1e50d59b942972b89da027b43d29f4630cc1da44ddc","scope":"adjacent","source_url":"https://arxiv.org/abs/2606.19388","title":"Beyond the GUI Paradigm: Do Mobile Agents Need the Phone Screen?","topics":["evaluation","automation","cross-app","tool-integration"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.19926","kind":"paper","record_sha256":"b5bae45c3519e539e5cfb02bd9d7a6c4a7fe6271c178c3d1a0ccb8f1fb866ed5","scope":"core","source_url":"https://arxiv.org/abs/2606.19926","title":"MemGUI-Agent: An End-to-End Long-Horizon Mobile GUI Agent with Proactive Context Management","topics":["memory","interaction-history","long-horizon","imitation-learning","data-generation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2606.19930","kind":"paper","record_sha256":"06b69a857f76d6bce58b316b15842f4d617d9add80b1a797b5fadfb03fadbc1f","scope":"core","source_url":"https://arxiv.org/abs/2606.19930","title":"MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization","topics":["mobile-navigation","reinforcement-learning","data-generation","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.20363","kind":"paper","record_sha256":"383a2021d98ddb5b9881bd7e519932eaac187b1761f95f325ab56c4bea704439","scope":"core","source_url":"https://arxiv.org/abs/2606.20363","title":"Automating SKILL.md Generation for Computer-Using Agents via Interaction Trajectory Mining","topics":["memory","data-generation","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.20717","kind":"paper","record_sha256":"8c3231b76d8331a3a8779e5602feec0ab429f70dce20d57378fa2f109232fbfa","scope":"core","source_url":"https://arxiv.org/abs/2606.20717","title":"MIRAGE: Stealthy Visual Prompt Injection for Vulnerability Detection in Web Agents","topics":["safety-security","robustness","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.20785","kind":"paper","record_sha256":"16ab490c447050c6e4c5a11949887975a6f552e6e64e136c9e37bbd3f58e7863","scope":"core","source_url":"https://arxiv.org/abs/2606.20785","title":"Fara-1.5: Scalable Learning Environments for Computer Use Agents","topics":["data-generation","web-navigation","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.20910","kind":"paper","record_sha256":"2cbbf1b7961baba9528b82117eb99b1ccdedfa25bce96900f4885d3b83368f6b","scope":"core","source_url":"https://arxiv.org/abs/2606.20910","title":"Whose Agent Are You? Multi-Layer Fingerprinting and Attribution of Autonomous Web Agents","topics":["safety-security","observability","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.21654","kind":"paper","record_sha256":"c37f1dffa249700e6f3267903d7ff2ed1b50db4ae5072f51be64bcc17ddc690f","scope":"core","source_url":"https://arxiv.org/abs/2606.21654","title":"ChainWorld: Composing Long-Horizon Desktop Workloads from Atomic OSWorld Tasks","topics":["evaluation","long-horizon","state-tracking","workflow"],"year":"2026"},{"environments":["desktop","macos"],"id":"arxiv:2606.22557","kind":"paper","record_sha256":"cf4e59e9f9bfd906c33da205e8a8e8e8c7411debfe43fdff24337d33b03e23e3","scope":"core","source_url":"https://arxiv.org/abs/2606.22557","title":"MacAgentBench: Benchmarking AI Agents on Real-World macOS Desktop","topics":["evaluation","cross-app","tool-integration"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.22948","kind":"paper","record_sha256":"c863ee2adabad3fe16196ded8ba2ab8f65aae54ed28904a58fa0935008d2da43","scope":"core","source_url":"https://arxiv.org/abs/2606.22948","title":"ENVS: Environment-Native Verified Search for Long-Horizon GUI Agents","topics":["data-generation","reinforcement-learning","long-horizon","robustness","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.23189","kind":"paper","record_sha256":"b75d0aff9628d73d7b7fe9921e929df0e923714cadaf1cf55a802ceee5907687","scope":"core","source_url":"https://arxiv.org/abs/2606.23189","title":"Capable but Careless: Do Computer-Use Agents Follow Contextual Integrity?","topics":["evaluation","safety-security","cross-app"],"year":"2026"},{"environments":["desktop","macos","windows"],"id":"arxiv:2606.24515","kind":"paper","record_sha256":"6e34baa110134b041614c96b1db1f280775d85edffdc36fefefd81d29e5f6662","scope":"core","source_url":"https://arxiv.org/abs/2606.24515","title":"Reinforcement Learning for Computer-Use Agents with Autonomous Evaluation","topics":["reinforcement-learning","reward-models","evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.24525","kind":"paper","record_sha256":"266f73094c88e62e96f8b9666ece7baa54b9e97b982be636289cf9cf9b31f04c","scope":"core","source_url":"https://arxiv.org/abs/2606.24525","title":"VisCritic: Visual State Comparison as Process Reward for GUI Agents","topics":["reward-models","state-tracking","data-generation","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.24551","kind":"paper","record_sha256":"6181db7aa83335e1664d5958239de8562ada116e7e7e410139ff6dc739a1a825","scope":"core","source_url":"https://arxiv.org/abs/2606.24551","title":"GUI vs. CLI: Execution Bottlenecks in Screen-Only and Skill-Mediated Computer-Use Agents","topics":["evaluation","tool-integration","robustness","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.25705","kind":"paper","record_sha256":"83037e7c0411d721783a5ec03e459963c890484e28b4b05077891f161ff9cb97","scope":"core","source_url":"https://arxiv.org/abs/2606.25705","title":"GUI agent: Guided Exploration of User-Sensitive Screens","topics":["safety-security","human-in-the-loop","data-generation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.25760","kind":"paper","record_sha256":"9c0c7afcadc3b55377b8696a08b2c644f086955dbcc9965458cab8d64f94f3aa","scope":"core","source_url":"https://arxiv.org/abs/2606.25760","title":"Uncertainty Quantification for Computer-Use Agents: A Benchmark across Vision-Language Models and GUI Grounding Datasets","topics":["grounding","evaluation","robustness","safety-security"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.25964","kind":"paper","record_sha256":"740df38caf1e56953cc7a6d541a09eea9773e51e65d1c00b3a17fea56c72554c","scope":"core","source_url":"https://arxiv.org/abs/2606.25964","title":"WinDOM: Self-Family Distillation for Small-Model GUI Grounding","topics":["grounding","data-generation","imitation-learning","reinforcement-learning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.27330","kind":"paper","record_sha256":"84688dac4f84e36f61617d6426ceb75dfb9777d340433c8618dd04a026a470ec","scope":"core","source_url":"https://arxiv.org/abs/2606.27330","title":"Empowering GUI Agents via Autonomous Experience Exploration and Hindsight Experience Utilization for Task Planning","topics":["planning","data-generation","imitation-learning","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.28480","kind":"paper","record_sha256":"1c996c92ffa47e30f9450e4b968b70f3dda76187e312f0233101be841ffad31b","scope":"adjacent","source_url":"https://arxiv.org/abs/2606.28480","title":"TUA-Bench: A Benchmark for General-Purpose Terminal-Use Agents","topics":["evaluation","automation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.29472","kind":"paper","record_sha256":"dd26da8a5d11fade605e48fc0259cbd1eb717ef3652b11a7a9ff3c1347002871","scope":"core","source_url":"https://arxiv.org/abs/2606.29472","title":"Agent-Computer Observation Interfaces Enable Dynamic Computer Use","topics":["evaluation","video","screen-understanding","memory","training-free"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2606.29537","kind":"paper","record_sha256":"f3e9e41680dc0e801b85320d9c24f9c1feaf008a190f635804b9f2298abd3859","scope":"core","source_url":"https://arxiv.org/abs/2606.29537","title":"OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks","topics":["evaluation","long-horizon","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.29705","kind":"paper","record_sha256":"327ce86c28b22c8c8367c6585dc594b5f246aa46532d657092432fb097d02442","scope":"core","source_url":"https://arxiv.org/abs/2606.29705","title":"GUICrafter: Weakly-Supervised GUI Agent Leveraging Massive Unannotated Screenshots","topics":["grounding","data-generation","reinforcement-learning","screen-understanding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.30084","kind":"paper","record_sha256":"37bdeb3b77fecc94c1972d3ecc3c3d5a698d6b582c44b1f39dc8a80f6994a78c","scope":"core","source_url":"https://arxiv.org/abs/2606.30084","title":"One Forward Beats Two: InnerZoom for Accurate and Efficient GUI Grounding","topics":["grounding","efficiency","screen-understanding","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.30119","kind":"paper","record_sha256":"caefb6eb9e26cb0fe2b5c077a9018c294a99a2143b89080b7a1802613d3a05d2","scope":"core","source_url":"https://arxiv.org/abs/2606.30119","title":"On the Internet, Nobody Knows You're an LLM Bot: Unmasking Web Agents with Multi-Layer Fingerprinting","topics":["safety-security","observability","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2606.30697","kind":"paper","record_sha256":"6e5c17802e87a2492462430807bf54a5181b734df5943e8bee468a28b101789b","scope":"core","source_url":"https://arxiv.org/abs/2606.30697","title":"LUMOS: A Semantic Operating-System Layer for Accessibility-Grounded AI Agents","topics":["grounding","screen-understanding","tool-integration","automation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.31270","kind":"paper","record_sha256":"66ec1694a0bba860f8ae3905f0ac0d47f5aa4c4ae213c6fba060c0a2f9b9aef6","scope":"core","source_url":"https://arxiv.org/abs/2606.31270","title":"Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents","topics":["self-reflection","robustness","training-free","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2606.31410","kind":"paper","record_sha256":"60bc69f2278d976cd885950de2dd9a042ec80575bee404693ca13a8a487f30da","scope":"core","source_url":"https://arxiv.org/abs/2606.31410","title":"Xiaomi-GUI-0 Technical Report","topics":["mobile-navigation","reinforcement-learning","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2606.31980","kind":"paper","record_sha256":"e0a9d6ad4f46617efb94ae9166d75ed19491af6cd47ad648c65a68f9060e9c5c","scope":"adjacent","source_url":"https://arxiv.org/abs/2606.31980","title":"DigitalCoach: Communication and Grounding Gaps in Human and Agentic Computer Use Coaching","topics":["human-in-the-loop","video","evaluation","grounding"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2607.04334","kind":"paper","record_sha256":"4f9f08ea4fd352affbd450d094a82dad760835ea8aa70af18e1e9c5529d800b7","scope":"core","source_url":"https://arxiv.org/abs/2607.04334","title":"Do GUI Agents Believe Their Eyes? Diagnosing State-Belief Reliance on Pixels versus Structure","topics":["screen-understanding","state-tracking","evaluation","robustness","safety-security"],"year":"2026"},{"environments":["desktop","cross-platform"],"id":"arxiv:2607.04425","kind":"paper","record_sha256":"d8452145b3317149ada6974b9e334b6b766e9369e1ea26a8bdfaf6deb5d9a3f3","scope":"core","source_url":"https://arxiv.org/abs/2607.04425","title":"UI-MOPD: Multi-Platform On-Policy Distillation for Unified GUI Agents","topics":["data-generation","imitation-learning","action-prediction"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.05120","kind":"paper","record_sha256":"5818a09ca5e97887ab0bdfd2bddcae189c0c0af66d1137419811efcee8bd3512","scope":"core","source_url":"https://arxiv.org/abs/2607.05120","title":"Agent Data Injection Attacks are Realistic Threats to AI Agents","topics":["safety-security","evaluation","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.05277","kind":"paper","record_sha256":"324647dbe52daf6bea2e5c7dc2539fdb5145d077f49e8bae6fe782163c253ecd","scope":"core","source_url":"https://arxiv.org/abs/2607.05277","title":"Untrusted Content Masking for Web Agents with Security Guarantees","topics":["safety-security","sandbox","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.06118","kind":"paper","record_sha256":"a8d53caacaa1de2313e046e9f9370e3493dc2d0ebee0de418baf525cfa706593","scope":"core","source_url":"https://arxiv.org/abs/2607.06118","title":"WebRetriever: A Large-Scale Comprehensive Benchmark for Efficient Web Agent Evaluation","topics":["evaluation","web-navigation","observability"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2607.09773","kind":"paper","record_sha256":"ef4d253857e031f72b444de1bb8a5da79a7b7b4bad0282e22adbfae9c6e19e0f","scope":"core","source_url":"https://arxiv.org/abs/2607.09773","title":"EvoCUA-1.5: Online Reinforcement Learning for Multi-turn Computer-Use Agents","topics":["reinforcement-learning","long-horizon","sandbox","data-generation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.10079","kind":"paper","record_sha256":"a2b6f649b9d8ec36ea269d8253a696dfb56991217ea77c2bd24528133e2fcad3","scope":"core","source_url":"https://arxiv.org/abs/2607.10079","title":"MAG: A Web-Agent Benchmark and Harness for Multimodal Action and Guide Generation","topics":["evaluation","grounding","reinforcement-learning","action-prediction"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.10608","kind":"paper","record_sha256":"8864f1745f8d98e726b1233744c2b95e1d2f6323b047e4b9f13771469f01efb6","scope":"core","source_url":"https://arxiv.org/abs/2607.10608","title":"The Compliance Trap: Diagnosing How AI Agents Consume Conflicting Memory","topics":["memory","safety-security","robustness","evaluation","long-horizon"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2607.11185","kind":"paper","record_sha256":"768008dd640036c0af69e3010e3ab4d394202e973d1198f0b3e4c2ce9b603fa5","scope":"core","source_url":"https://arxiv.org/abs/2607.11185","title":"SCALECUA: Scaling Computer Use Agents with Verifiable Task Synthesis and Efficient Online RL","topics":["data-generation","reinforcement-learning","efficiency"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2607.11388","kind":"paper","record_sha256":"7bbc0dc5a154da926e371e9c5f29a3f33d610114be967567cc740d4d910056e6","scope":"core","source_url":"https://arxiv.org/abs/2607.11388","title":"StructAgent: Harness Long-horizon Digital Agents with Unified Causal Structure","topics":["state-tracking","planning","long-horizon","robustness"],"year":"2026"},{"environments":["desktop","android","ios","harmonyos","windows"],"id":"arxiv:2607.12625","kind":"paper","record_sha256":"329c515f7b01cba59aa7d2c4314eda42aa03c53da0febf15c95983ef5f7d3f13","scope":"core","source_url":"https://arxiv.org/abs/2607.12625","title":"KnowAct-GUIClaw: Know Deeply, Act Perfectly, Personal GUI Assistant with Self-Evolving Memory and Skill","topics":["memory","planning","interaction-history","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.12640","kind":"paper","record_sha256":"c0c29410b98b899014898998b7e4f6a0fc6c16a769a51e740f5ba22074fd0610","scope":"core","source_url":"https://arxiv.org/abs/2607.12640","title":"A Learning-Rate-Gated Failure of GRPO in a Small Language and Vision-Language Model Web Agent: A Controlled Null and Its Mechanism","topics":["reinforcement-learning","evaluation","robustness"],"year":"2026"},{"environments":["desktop","macos"],"id":"arxiv:2607.14443","kind":"paper","record_sha256":"d2f89dba167d95153bd04412e8d6f3af6cb67f68b629cbd0a70036e70d861b12","scope":"core","source_url":"https://arxiv.org/abs/2607.14443","title":"Tactile: Giving Computer-Using Agents Hands and Feet","topics":["grounding","automation","observability","tool-integration","robustness","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2607.14548","kind":"paper","record_sha256":"0151250e6948b7c014829c08e471f57256fa866ef0eaf5887ad6c8072f2ec0de","scope":"core","source_url":"https://arxiv.org/abs/2607.14548","title":"HyMobileAgent: Data-Environment Co-Scaling for Efficient GUI Agents","topics":["mobile-navigation","data-generation","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2607.15193","kind":"paper","record_sha256":"f0dbb37e83ef218317c5fa852a65229cfd626011a39d30e456f8a43646873bca","scope":"core","source_url":"https://arxiv.org/abs/2607.15193","title":"Plover: Steering GUI Agents through Plan-Centric Interaction","topics":["planning","human-in-the-loop","observability","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2607.15550","kind":"paper","record_sha256":"a3afe3778fc358e6539318e9d1db002bf180b8c3a9f33614dbbadf944f0aa815","scope":"core","source_url":"https://arxiv.org/abs/2607.15550","title":"SeerGuard: A Safety Framework for Mobile GUI Agents via World Model Prediction","topics":["safety-security","world-models"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.17050","kind":"paper","record_sha256":"5494bf6567521986d4bd3672fa0136fa8528180688a8db7afe2846324aa52b7c","scope":"core","source_url":"https://arxiv.org/abs/2607.17050","title":"EvoGUI: An Evolution-Aware Benchmark for GUI State-Transition Understanding","topics":["evaluation","world-models","action-prediction"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2607.17527","kind":"paper","record_sha256":"d89c7c293b365a899aef2629210aed7a7a77dbf4729a9b505f265a191c5c99c9","scope":"core","source_url":"https://arxiv.org/abs/2607.17527","title":"Sidekick: Designing Communication for Effective Multitasking with Computer Use Agents","topics":["human-in-the-loop","observability","long-horizon"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2607.18046","kind":"paper","record_sha256":"0fa63251eb58a1dd2dab45b25c726a97bf3a960417300440e13568a42e2f45f7","scope":"core","source_url":"https://arxiv.org/abs/2607.18046","title":"SEE: Structure-aware Exploring & Exploiting for Long-horizon GUI Agent Trajectory Synthesis","topics":["data-generation","long-horizon","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.21557","kind":"paper","record_sha256":"bd0f1f1cd5ac520c9d780da0669994d5cac70e5a1b41e0db48d37e07734afb71","scope":"core","source_url":"https://arxiv.org/abs/2607.21557","title":"OpenForgeRL: Train Harness-native Agents in Any Environment","topics":["reinforcement-learning","sandbox","tool-integration","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2607.22689","kind":"paper","record_sha256":"670e9fab6c6a0584a72555644ca58d8b4a0c987573556ea7e9d240e122b8073c","scope":"core","source_url":"https://arxiv.org/abs/2607.22689","title":"Beyond Sequential Interaction: Benchmarking Parallel Execution and Coordination for GUI Agents","topics":["evaluation","planning","efficiency","long-horizon"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2607.22798","kind":"paper","record_sha256":"8ac0981be3aee93b70e915639cce697196ada4b9a864dc0eb1c0c132e519fbf2","scope":"core","source_url":"https://arxiv.org/abs/2607.22798","title":"StateAct: Program State, before Pixels, for Long-Horizon Computer-Use Agents","topics":["tool-integration","state-tracking","long-horizon"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2607.23263","kind":"paper","record_sha256":"170031606adbe730acdec2d746439ba539aae033008e78345385a320ff129d6b","scope":"core","source_url":"https://arxiv.org/abs/2607.23263","title":"SeekJudge: A Practical Reward Framework for Reinforcement Learning in Computer-Use Agents","topics":["reward-models","evaluation","data-generation"],"year":"2026"},{"environments":["desktop","android"],"id":"arxiv:2607.24112","kind":"paper","record_sha256":"73c96fe08d05224fca45dd105e42467ad55a08313fe1ac5ee4a902ddeed6d470","scope":"core","source_url":"https://arxiv.org/abs/2607.24112","title":"Scaling GUI Agents with Visual State Transitions","topics":["world-models","action-prediction","data-generation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.24167","kind":"paper","record_sha256":"d7f97c285414433f985d78b85a63c23e65d5f6cdd188f53d60c4441806b47d8e","scope":"core","source_url":"https://arxiv.org/abs/2607.24167","title":"Falsifiable Commitment Planning for Self-Correcting Web Agents","topics":["planning","robustness","self-reflection"],"year":"2026"},{"environments":["desktop","linux"],"id":"arxiv:2607.25904","kind":"paper","record_sha256":"8e60208180a3afe2e93f3021a3ace297c9b5be58970d5bc38ac97a9d6824b9bc","scope":"core","source_url":"https://arxiv.org/abs/2607.25904","title":"Interactive Reward Agent: GUI Task Evaluation via Environment-State Verification","topics":["reward-models","evaluation","tool-integration"],"year":"2026"},{"environments":["desktop","linux"],"id":"arxiv:2607.26041","kind":"paper","record_sha256":"f9abd0545716f483a6b8771c566c0c0691a7d37f067fac403380562ef49097c5","scope":"core","source_url":"https://arxiv.org/abs/2607.26041","title":"Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?","topics":["evaluation","screen-understanding","state-tracking"],"year":"2026"},{"environments":["web"],"id":"arxiv:2607.28074","kind":"paper","record_sha256":"9e3c6b8d3a7f126af062eeb15c4a6b262aaa320a7c85069d272af78fddbba38a","scope":"core","source_url":"https://arxiv.org/abs/2607.28074","title":"Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale","topics":["sandbox","data-generation","reinforcement-learning"],"year":"2026"},{"environments":["cross-platform","web","desktop"],"id":"arxiv:2607.28227","kind":"paper","record_sha256":"05a83adcef1b249d5889e474ee8dc20be2fd04d726e54fdabe2cd56c28721676","scope":"core","source_url":"https://arxiv.org/abs/2607.28227","title":"Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents","topics":["reinforcement-learning","long-horizon","tool-integration"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2607.28367","kind":"paper","record_sha256":"ff4b587064a45ab8a82fab2c6cbe19451ff51f17827a7b0cab5774e57c808485","scope":"core","source_url":"https://arxiv.org/abs/2607.28367","title":"How Benchmarks Mis-Score Computer-Use Agents","topics":["evaluation","observability","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2607.28399","kind":"paper","record_sha256":"54385a5992a3c8de591a81c206e3192e555e5c22fefd021828a860e0dc9e5df7","scope":"core","source_url":"https://arxiv.org/abs/2607.28399","title":"Why Are GUI Agents Correct but Late? Decode on the Decision-Time Critical Path, Tested with Pre-Compiled Policy Trees","topics":["efficiency","planning","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2607.28573","kind":"paper","record_sha256":"78acf1369aa96b784584acf2f6f9ad5687538584cbe689c1153b5fac55886a80","scope":"core","source_url":"https://arxiv.org/abs/2607.28573","title":"Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs","topics":["efficiency","evaluation","robustness"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2607.28609","kind":"paper","record_sha256":"f90bba85a6d13f94adc9b7c5339d6fdb6522712af938598f9510572b0bd19683","scope":"core","source_url":"https://arxiv.org/abs/2607.28609","title":"OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models","topics":["reward-models","evaluation","data-generation"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2607.29320","kind":"paper","record_sha256":"3f5b7dacc2b799a8bc0d8c9ebee0d6b3c079b9488f25f0d5238e10bca0e302d5","scope":"core","source_url":"https://arxiv.org/abs/2607.29320","title":"MAGA: Multi-Platform Self-Fusion of GUI Agents via Structured Action Distillation","topics":["imitation-learning","grounding","automation"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2608.02352","kind":"paper","record_sha256":"36c4502e7f0254be8aca9a1b6b8708f9c460ff58fc0b4f94763d4867650c53f3","scope":"core","source_url":"https://arxiv.org/abs/2608.02352","title":"Qwen-CUA: Native Computer Use for (almost) Everything","topics":["reinforcement-learning","imitation-learning","data-generation","interaction-history","long-horizon","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.02643","kind":"paper","record_sha256":"0edc0e5fcb07193ebf32f6f9a19438c4367b058a59231c0b69249e359edd01dc","scope":"core","source_url":"https://arxiv.org/abs/2608.02643","title":"CUADebug: Diagnosing and Repairing Computer-Use Agent Failures","topics":["evaluation","robustness","self-reflection","state-tracking"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.03270","kind":"paper","record_sha256":"0ad5eb99e1ac879cc2984fa1e8ddac3d9b4708ed8c1823898604136f6179acb1","scope":"core","source_url":"https://arxiv.org/abs/2608.03270","title":"GUI-Lens: Coarse-to-Fine Cropping for GUI Grounding with General-Purpose VLMs","topics":["grounding","screen-understanding","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2608.03327","kind":"paper","record_sha256":"923a8ae9c985a529315955cfbdb1cb62de32e23cb8fe98e841dda10c6d1ed1d5","scope":"core","source_url":"https://arxiv.org/abs/2608.03327","title":"Screenshots or Tools? Eliciting Tool Use and Managing Multimodal Context in Hybrid GUI-MCP Computer-Use Agents","topics":["tool-integration","reinforcement-learning","efficiency","interaction-history","evaluation"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2608.03740","kind":"paper","record_sha256":"b4be089b3c4264fdd4790dd2f817add0c86af925c7a626785b78aced9ff82b55","scope":"core","source_url":"https://arxiv.org/abs/2608.03740","title":"MissClick: Execution-Aware Adversarial Attacks on Coordinate Generation in GUI Grounding Models","topics":["safety-security","grounding","robustness","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.04741","kind":"paper","record_sha256":"0d8d477ed11845ca9937b93a938e34b6d19d1bacb76da92ffa845d1e41f4eda6","scope":"core","source_url":"https://arxiv.org/abs/2608.04741","title":"LoginTrap: Uncovering Task-Agnostic Phishing-Style Indirect Prompt Injection Attacks against LLM-based Web Agents","topics":["safety-security","robustness","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.04755","kind":"paper","record_sha256":"11324d702c79359011f13b569d46b1a49b9b3857ad38feb95dcc014350a81865","scope":"core","source_url":"https://arxiv.org/abs/2608.04755","title":"\"Allow\" to Achieve, Over-Privileged Inadvertently: The Unintended Cost of Task-Completion-Driven Pop-up Decisions in Mobile GUI Agents","topics":["safety-security","evaluation","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2608.05587","kind":"paper","record_sha256":"a70b6949d862194f0c844aa0c81368ff765983d286e2dd326cca146c1cfb2bde","scope":"core","source_url":"https://arxiv.org/abs/2608.05587","title":"StepReflect: Structured UI Transition Reflection for Mobile GUI Agents","topics":["self-reflection","state-tracking","imitation-learning","reward-models","evaluation","efficiency"],"year":"2026"},{"environments":["android"],"id":"arxiv:2608.05891","kind":"paper","record_sha256":"277d2a00e1000943141943d2afb0c1c3d2bf305a0a14574be6f69b29e3ba5d5e","scope":"core","source_url":"https://arxiv.org/abs/2608.05891","title":"AppDeltaWorld: Transition-Grounded Delta Code World Model for Mobile GUI Agents","topics":["world-models","data-generation","imitation-learning","reinforcement-learning","mobile-navigation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2608.06065","kind":"paper","record_sha256":"95bb8b7e805255cd157ce18cb5dc6f7f0d584b3cb25272e10b9e3bc317014570","scope":"core","source_url":"https://arxiv.org/abs/2608.06065","title":"The Next Screenshot Knows: Gated Hindsight Distillation for Mobile GUI Agents","topics":["imitation-learning","interaction-history","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.06171","kind":"paper","record_sha256":"8f2837b0e48a232c5b3436f3764f9da292e8e7827dcdcf006345009bce6f159f","scope":"core","source_url":"https://arxiv.org/abs/2608.06171","title":"Routing Is Least Learnable Where It Is Most Valuable: Bounds on Representation Routing for Web Agents","topics":["screen-understanding","efficiency","evaluation","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.08392","kind":"paper","record_sha256":"775d5d9559f51c8fdb6c079aed5db3b1465fdd7e0570ddaee8ba64fa58a76a0f","scope":"core","source_url":"https://arxiv.org/abs/2608.08392","title":"CAP: A Scalable Benchmark for Evaluating Cross-Site Browser Agents with Complex Actions and Perception","topics":["evaluation","web-navigation","cross-app","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.09654","kind":"paper","record_sha256":"c5ee5a660c374a9db3e3f8a5c62c02f7975827db38c8dd3a2068640116a2a734","scope":"core","source_url":"https://arxiv.org/abs/2608.09654","title":"Hallucination-Free GUI Grounding via Regression-Free Layout-Aware Matching","topics":["grounding","screen-understanding","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.09723","kind":"paper","record_sha256":"928dbe30b8e9ef94646ab74556e350f8083f339c5aabe7053590f22d3cfe20e0","scope":"core","source_url":"https://arxiv.org/abs/2608.09723","title":"LookAgain: Closed-Loop GUI Grounding with Visually Grounded Reflection","topics":["grounding","self-reflection","reinforcement-learning","imitation-learning"],"year":"2026"},{"environments":["desktop","linux"],"id":"arxiv:2608.10366","kind":"paper","record_sha256":"f4ae1d40b3c0c9a1e6cd9fabb570da0c57dbb91e4f6b462a00d0846c92c2b499","scope":"core","source_url":"https://arxiv.org/abs/2608.10366","title":"DSAgentBench: Can Agents Automate End-to-End Data-Science Workflows in Real Computer Environments?","topics":["evaluation","workflow","science","long-horizon"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.11434","kind":"paper","record_sha256":"dbeadd9f41ba67c4061cebacf577726e55fbba00356134810f41edfd96cd60b5","scope":"core","source_url":"https://arxiv.org/abs/2608.11434","title":"Benchmarking LLM Judges for Mobile Agent Evaluation","topics":["evaluation","reward-models","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.12429","kind":"paper","record_sha256":"6a03c07accc8e2fce292e8172790edf0378ba6cded0e216cd58f21a64b5e9d8a","scope":"core","source_url":"https://arxiv.org/abs/2608.12429","title":"SynWeaver: Website-Prior Task and Trajectory Co-Synthesis for Web Agents","topics":["data-generation","web-navigation","state-tracking","robustness"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2608.14131","kind":"paper","record_sha256":"6e743330a8b696ac5e59c35e85ec09ef3d9ee806e59278c416b5a30dafb4fdb4","scope":"core","source_url":"https://arxiv.org/abs/2608.14131","title":"LegacyWorld: Atomicity-Aware Evaluation of GUI Agents for Legacy Workflows","topics":["evaluation","safety-security","workflow","recording"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2608.15930","kind":"paper","record_sha256":"f8cc76f793eaaa6d9f711e71da76741e048fc9acfe0cfb2ffe464cd32b217ed5","scope":"core","source_url":"https://arxiv.org/abs/2608.15930","title":"UI-Mate: Advancing Open-Weight Foundation GUI Agents with In-Context Demonstrations","topics":["workflow","data-generation","long-horizon"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.17319","kind":"paper","record_sha256":"036ab326f1f365d730eb3fcc663b18a971a53a0f9a05ad33c8d27600d5ef726f","scope":"core","source_url":"https://arxiv.org/abs/2608.17319","title":"Wuying-Browser-Agent: Real-World Centric Fundamental Long-Horizon Browser Agents","topics":["web-navigation","long-horizon","reinforcement-learning"],"year":"2026"},{"environments":["android"],"id":"arxiv:2608.17659","kind":"paper","record_sha256":"6359e1cf9ed90c25704534fc92acd07999a57c4a9856e0e5f1ebf078df14c0ee","scope":"core","source_url":"https://arxiv.org/abs/2608.17659","title":"MobileWorldSafety: Benchmarking GUI Agent Safety Against Environmental Injection Attacks in Android Apps","topics":["safety-security","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.18307","kind":"paper","record_sha256":"56d73cb486651fa3d2cd98aec760bdccd0b6fba87e17380cb3dc1e9481768fa1","scope":"core","source_url":"https://arxiv.org/abs/2608.18307","title":"ComponentBench: Diagnosing Component-Level Failures in Computer-Use Agents","topics":["evaluation","screen-understanding","efficiency","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.20319","kind":"paper","record_sha256":"620656547a75991984cacfca794f54b8d7496a3213a9517511439385ad326876","scope":"core","source_url":"https://arxiv.org/abs/2608.20319","title":"Inducing Task Models from Computer-Use Traces","topics":["recording","workflow","memory","interaction-history"],"year":"2026"},{"environments":["web","cross-platform"],"id":"arxiv:2608.21794","kind":"paper","record_sha256":"61fb4f832f96b80e8f7d593da8ce8f5af04e4a9bfce9dfb84aecc7f5c4025ae4","scope":"core","source_url":"https://arxiv.org/abs/2608.21794","title":"Lexical Coupling in GUI Element Grounding: Sentence Embeddings Track Labels across Mobile and Web","topics":["grounding","evaluation","screen-understanding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.21830","kind":"paper","record_sha256":"e7792f1f0188ff02c748cfeaecda96ef93d4b490527553246c8c732f2e5e642f","scope":"core","source_url":"https://arxiv.org/abs/2608.21830","title":"Beyond Success and Failure: Length-Aware Contrastive Learning for GUI Agents","topics":["reinforcement-learning","efficiency","action-prediction"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.21832","kind":"paper","record_sha256":"0bbff2e9232f5bda60316538df23a12d0f7ef86170a78f43078fc96d5844a785","scope":"core","source_url":"https://arxiv.org/abs/2608.21832","title":"GUI-Primitives: Diagnosing Spatial Reasoning Failures in Vision-Language GUI Grounding","topics":["grounding","evaluation","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.21898","kind":"paper","record_sha256":"6f684b48357a56d300e402aa39beaadc75435589cf12ba3cbd5db1a09ae04875","scope":"core","source_url":"https://arxiv.org/abs/2608.21898","title":"Training Needs Trustworthy Worlds: Verified Synthetic Web Environments for Agent Learning","topics":["data-generation","sandbox","reinforcement-learning"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2608.22577","kind":"paper","record_sha256":"5b943f9cc1922bca6ab5f9148a34687e9c1b30538364043fcd58d71af84b0905","scope":"core","source_url":"https://arxiv.org/abs/2608.22577","title":"CausalCache: Conditional High-Fidelity Restoration for Long-Horizon GUI Agents","topics":["memory","long-horizon","efficiency"],"year":"2026"},{"environments":["android"],"id":"arxiv:2608.22847","kind":"paper","record_sha256":"f0eb38d3796eb33bb749bece2170183a100e4e15d77cadf47b91e6154425f6be","scope":"core","source_url":"https://arxiv.org/abs/2608.22847","title":"GSAR: Goal-State-Anchor Rewards for Mobile GUI Agents with Self-Evolving Data Synthesis","topics":["data-generation","reward-models","reinforcement-learning"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2608.24015","kind":"paper","record_sha256":"6e0c986f7d0e8416dc17c1099ee4b1990ac5f7a3d45f1c8be7255ceea88fff97","scope":"core","source_url":"https://arxiv.org/abs/2608.24015","title":"Reflection with Action-Induced Visual Differences for Desktop GUI Agents","topics":["self-reflection","screen-understanding","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.24017","kind":"paper","record_sha256":"4064a539cec1baec2fe66a26f17b9028164eacd8e7c0e65a94ab60fe1768fead","scope":"adjacent","source_url":"https://arxiv.org/abs/2608.24017","title":"WebMCP-Phalanx: Enforcing and Characterizing Trust Boundaries for Browser-Integrated LLM Agents","topics":["safety-security","tool-integration"],"year":"2026"},{"environments":["android"],"id":"arxiv:2608.24099","kind":"paper","record_sha256":"99554ed3705c76ebf1e3b15dc6a6e572e91554128ed494a428e4116430b5f9da","scope":"core","source_url":"https://arxiv.org/abs/2608.24099","title":"Are Android GUI Agents Robust Against Runtime Anomalies? AnTrap: Evaluating Agents in Dynamic Adversarial Environments","topics":["evaluation","robustness","reinforcement-learning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.24174","kind":"paper","record_sha256":"ebcab7d0fdd9d271556bc8c572559b7d741749762aff4cf9f7ae41fdfc6ce1b5","scope":"core","source_url":"https://arxiv.org/abs/2608.24174","title":"Task-Adaptive Rubrics for GUI Reward Modeling","topics":["reward-models","evaluation","reinforcement-learning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.24848","kind":"paper","record_sha256":"cb37414b9c4a604e728fd5446c8f4b45aa9418ce44e547272f881e44920d7e91","scope":"core","source_url":"https://arxiv.org/abs/2608.24848","title":"BrowserForge: Scaling Web Episode via Parallel Browser Sandboxes","topics":["data-generation","sandbox","web-navigation"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2608.25777","kind":"paper","record_sha256":"37772c97a8a3fadd57cfead8e49d8534792e1095bc0742675386fc99218fa416","scope":"core","source_url":"https://arxiv.org/abs/2608.25777","title":"LocalLSTC: A Long Short-Term Control Architecture for Locally Deployed GUI Agents","topics":["state-tracking","planning","training-free"],"year":"2026"},{"environments":["desktop","cross-platform"],"id":"arxiv:2608.26204","kind":"paper","record_sha256":"b11793c95a12c179de6de2d6840e55e9cec46a749b8b3ae0ffd01aadfd5bba74","scope":"core","source_url":"https://arxiv.org/abs/2608.26204","title":"ADeptS-Bench: Measuring the Trustworthiness of Computer Use Agents Across Devices","topics":["safety-security","evaluation","human-in-the-loop"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.26991","kind":"paper","record_sha256":"688773b718f69952a869012adc67ee8aec6c7b15f8c44f399e7918b4d3bcbfa1","scope":"core","source_url":"https://arxiv.org/abs/2608.26991","title":"ASIL: Replacing Screenshot-and-Click with Structured State and Semantic Actions","topics":["automation","tool-integration","evaluation","reinforcement-learning","imitation-learning"],"year":"2026"},{"environments":["android"],"id":"arxiv:2608.27508","kind":"paper","record_sha256":"d3c38ca31a729386b2b910e7a02abe550b2271e22ffb6c1a639762d206ef6d7a","scope":"core","source_url":"https://arxiv.org/abs/2608.27508","title":"WM-R1: Training GUI Agents to Reason and leverage World Models with Reinforcement Learning","topics":["world-models","reinforcement-learning"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2608.27808","kind":"paper","record_sha256":"d05d98d3cfc536e75c90f11fe37f3e22435b6bd3d8766ea748c4f33b2d47eae6","scope":"core","source_url":"https://arxiv.org/abs/2608.27808","title":"CURA: Certified Runtime Alarms for Computer-Use Agents","topics":["observability","safety-security","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.27866","kind":"paper","record_sha256":"cb0ee67028c3e0720dd2f86a7b23e5902dd2ce3c55dfeb72fb11918d1f080b88","scope":"core","source_url":"https://arxiv.org/abs/2608.27866","title":"Iron: Intent-Aligned and Retrospective Dual Learning Framework for Enhancing Generalist Virtual Agents","topics":["reinforcement-learning","data-generation","grounding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2608.29305","kind":"paper","record_sha256":"ce8ce2fb6c963532e5b6c4e49693bd896b8151d94be24ed9989414a23f02d15d","scope":"core","source_url":"https://arxiv.org/abs/2608.29305","title":"Learning Simple Test-Time Environments for LLM Web Agents","topics":["planning","robustness","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2608.30038","kind":"paper","record_sha256":"783db91fefc824eb3e98f57d62da47846d12ef6af1a1ef190d72e7c4680f0cdc","scope":"core","source_url":"https://arxiv.org/abs/2608.30038","title":"ActReal: System-Level Mobile Agents Challenge Mobile Automation Detection","topics":["safety-security","mobile-navigation","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2608.30207","kind":"paper","record_sha256":"5a876cf180ea9746d03de8acf7342093d56dd3640430286a9b964c9b17d773fa","scope":"core","source_url":"https://arxiv.org/abs/2608.30207","title":"SIR: Self-improving Red-teaming for Compute Use Agents","topics":["safety-security","evaluation","data-generation"],"year":"2026"},{"environments":["web","desktop","cross-platform"],"id":"arxiv:2609.00028","kind":"paper","record_sha256":"1333281c61baa5719f9655a01cd6e59a78109faf34db87efdc5b333cc01a3179","scope":"core","source_url":"https://arxiv.org/abs/2609.00028","title":"UI-Venus-2 Technical Report","topics":["automation","reinforcement-learning","safety-security"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.00048","kind":"paper","record_sha256":"557774ef60993c0a5042a5301b6fdfa27d444bf37b1ba88d241e33bb19a9edb0","scope":"core","source_url":"https://arxiv.org/abs/2609.00048","title":"GUI-CC: Benchmarking Contextual Consistency of GUI World Models as Agent Environments","topics":["world-models","evaluation","long-horizon"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.00524","kind":"paper","record_sha256":"ba6d7995b6c0e9bd538fd37cb7b4f91f5e8e7136e3283dbf18da6e790f9d935b","scope":"core","source_url":"https://arxiv.org/abs/2609.00524","title":"Are We There Yet? Assessing Computer-Use Agents for Blind Users' Accessible Interaction with Desktop Applications","topics":["evaluation","human-in-the-loop","grounding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.02057","kind":"paper","record_sha256":"cb923aa3c020257626d03275a2b71c1e0a30c4b1ca872ee46ad49e603b65015d","scope":"core","source_url":"https://arxiv.org/abs/2609.02057","title":"Monitoring Web Agents Without Internal Signals: Observable Trajectories and Key-Step Supervision","topics":["observability","evaluation","robustness","interaction-history"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.02149","kind":"paper","record_sha256":"b9a30d1550362ba6b80b46cacc03b4a23f50cbe94d35b99f1f90c703a707c154","scope":"core","source_url":"https://arxiv.org/abs/2609.02149","title":"OmegaUse-SOP: SOP Engineering for Professional Computer Use from Human Demonstrations","topics":["human-in-the-loop","recording","workflow","automation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.02154","kind":"paper","record_sha256":"ea1c1026b408dd52e55a13e347856b3a66530090d78c281a889a0d93cba198c2","scope":"core","source_url":"https://arxiv.org/abs/2609.02154","title":"WiP: Characterizing and Defending Against Mobile-Agent-Driven MFA Automation","topics":["safety-security","evaluation"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2609.02309","kind":"survey","record_sha256":"adc44dfe425f8d31332b77465a748bc793d28076450fe15d2c5b0e8d42f86ee3","scope":"core","source_url":"https://arxiv.org/abs/2609.02309","title":"Efficient GUI Agents: A Systems Survey of Observation, Memory, Action, and Runtime Optimization","topics":["survey"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.02885","kind":"paper","record_sha256":"10663e585ae18e342710e1b8c49b1e2eb55051e5773f6d8e14c123ae42814d21","scope":"core","source_url":"https://arxiv.org/abs/2609.02885","title":"Discriminative World Models for Web Agents","topics":["world-models","reward-models","planning","data-generation","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.03438","kind":"paper","record_sha256":"2cf61c3ff6eeb945f0a721004278566af88633e52713c35b96406744ca3439c5","scope":"core","source_url":"https://arxiv.org/abs/2609.03438","title":"Do GUI Agents Know When Not to Act? Enabling Conflict-Aware Termination for Multimodal GUI Agents","topics":["safety-security","evaluation","robustness","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.04850","kind":"paper","record_sha256":"f878c2e65fbb1b83f4b551257c69c595d3e06597b61f076f44245453c8e97d79","scope":"core","source_url":"https://arxiv.org/abs/2609.04850","title":"ElderBench: Benchmarking Autonomous Mobile Agents for Older Adults","topics":["evaluation","human-in-the-loop","mobile-navigation","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.04869","kind":"paper","record_sha256":"9bb437964d3519ac7dfd2354851c01924cb88f3a3bbbe34ca738b90d20557eae","scope":"core","source_url":"https://arxiv.org/abs/2609.04869","title":"From Interaction Traces to Persistent Skills: Online Evolution for Computer-Use Agents","topics":["memory","workflow","evaluation","training-free"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.05374","kind":"paper","record_sha256":"9cdf2e823f39b52121212cdee7af5339bf5ef601424b8b5e7081495a45be7ea0","scope":"core","source_url":"https://arxiv.org/abs/2609.05374","title":"CUA-Universe: A Scalable and Dynamic Environment for Hybrid GUI+CLI Agents","topics":["tool-integration","data-generation","sandbox"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.05511","kind":"paper","record_sha256":"4528e260f41fb860e47e7f4d522b41d522c363d7443d6e935081ac4c55428f35","scope":"core","source_url":"https://arxiv.org/abs/2609.05511","title":"SCAFFOLD: Self-Improving Web Agents via Recursive Parametric Skill Abstraction","topics":["memory","automation","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.05513","kind":"paper","record_sha256":"6afc644a8fc3f82cee775e43443099616cc94cec87ab393dcbd31d489b2b4bf0","scope":"core","source_url":"https://arxiv.org/abs/2609.05513","title":"When and What to Teach: Budget-Aware Online Adaptation for Web Agents","topics":["efficiency","data-generation","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.06530","kind":"paper","record_sha256":"90449b620f9611ca9b47b517f8eaa4acee86c4b38b4bb5f411977256bb8ac765","scope":"core","source_url":"https://arxiv.org/abs/2609.06530","title":"Selective Knowledge Control for Continual GUI Agent Learning over Application Streams","topics":["memory","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.06792","kind":"paper","record_sha256":"7941fa615d39aeeca4e53387681e57fce536fa5f8a123b2cbd14fee65a4b7eca","scope":"core","source_url":"https://arxiv.org/abs/2609.06792","title":"Improving Proficiency and Efficiency of Android GUI Agents via Self-Generating Tool Actions","topics":["tool-integration","automation","mobile-navigation","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.07712","kind":"paper","record_sha256":"c98d2728695e6c88c31691f40899f6b3565abecff3409ac5f970e642618f04b1","scope":"core","source_url":"https://arxiv.org/abs/2609.07712","title":"APPSim-Bench: Bridging Real-world Apps and Reproducible Evaluation for Mobile GUI Agents","topics":["evaluation","sandbox","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.09212","kind":"paper","record_sha256":"629b914464775f5594f74d025f816b3b0d0d624a96d6794ddccef06aa566ff29","scope":"core","source_url":"https://arxiv.org/abs/2609.09212","title":"AgentHijack: Visual Patch Attacks on Multimodal Computer-Use Agents","topics":["safety-security","evaluation","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.10031","kind":"paper","record_sha256":"79227c7b2ec2b84530c48fa2aa13391e3afde0207281e30544b2b3bfa32c75a8","scope":"core","source_url":"https://arxiv.org/abs/2609.10031","title":"GraphDroid: Asynchronous LLM-Based Mobile App GUI Testing via History-Aware Exploration and Hybrid Intent Fulfillment","topics":["evaluation","mobile-navigation","memory","interaction-history","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.10297","kind":"paper","record_sha256":"985db8c8e032157f3a9539257bac5894b8bf38fba49417081fa94af57c143183","scope":"core","source_url":"https://arxiv.org/abs/2609.10297","title":"TRACE: Trajectory-robust Admission with Evidence Ordering for Efficient GUI Agents","topics":["efficiency","interaction-history","screen-understanding","training-free"],"year":"2026"},{"environments":["desktop","android","windows","linux"],"id":"arxiv:2609.10451","kind":"paper","record_sha256":"4367545c329c2d93d7440cf2a55c1045a0fa49083a65f587ca0e81add10c246d","scope":"core","source_url":"https://arxiv.org/abs/2609.10451","title":"JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition","topics":["evaluation","cross-app","planning","long-horizon","state-tracking"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.12394","kind":"paper","record_sha256":"aab835068bb1070ef070294c46946bef00c9cf944aa58fdce457f50058cc5c61","scope":"core","source_url":"https://arxiv.org/abs/2609.12394","title":"BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents","topics":["mobile-navigation","reinforcement-learning","data-generation"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2609.13287","kind":"paper","record_sha256":"92fd95cfe51dbcc4d6e0780063e02592bbd1c0a3416d7cc685045c207115707c","scope":"core","source_url":"https://arxiv.org/abs/2609.13287","title":"LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents","topics":["grounding","imitation-learning","efficiency","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.13491","kind":"paper","record_sha256":"33d7ee69344064d0ebcf8b90854b989e35afb53de3d71619ab3c5b3321d628ef","scope":"core","source_url":"https://arxiv.org/abs/2609.13491","title":"Token Efficient Task Execution via Application Behavior Modeling for Web Agents","topics":["memory","efficiency","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.13548","kind":"paper","record_sha256":"eee8bb3744fab08a7198483e7ad78b8402f21961a16253b9512b63427bdb6d6f","scope":"core","source_url":"https://arxiv.org/abs/2609.13548","title":"AutoTailor: Automatic, User-Aligned Capability Selection and Adaptation for Web Agents","topics":["tool-integration","web-navigation","workflow","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.13873","kind":"paper","record_sha256":"2c1595cba4b4660b44fffcac07f9d2c37949348cd81f860ad01293ef70e4591d","scope":"core","source_url":"https://arxiv.org/abs/2609.13873","title":"PriMobiBench: Characterizing Visual Privacy Leakage in VLM-Driven Mobile GUI Agents","topics":["evaluation","safety-security","screen-understanding","mobile-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.15307","kind":"paper","record_sha256":"bfd0a0705605ecd6a8d37adf5ec6e1c07fcbc7af85b75bdec62d952f213dd768","scope":"core","source_url":"https://arxiv.org/abs/2609.15307","title":"Learning from Reliable Negatives: Confidence-Anchored Test-Time Adaptation for GUI Grounding","topics":["grounding","robustness","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.15457","kind":"paper","record_sha256":"641ac7f5b03149592dc8d6502d9d6cd9ba3f29859ca6b346c8b86834de0b87a5","scope":"core","source_url":"https://arxiv.org/abs/2609.15457","title":"AnchorGUI: Asymmetric Memory for Dual-Scale Learning in GUI Navigation","topics":["memory","interaction-history","state-tracking","self-reflection","efficiency"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.16251","kind":"paper","record_sha256":"a8443c68360123ee6f4efe393f7f14aaf4b15e13c102c41163242c6f0f5ea32a","scope":"core","source_url":"https://arxiv.org/abs/2609.16251","title":"CADWorld: Computer-Use Benchmark for Long-Horizon Computer-Aided Design","topics":["evaluation","long-horizon","workflow","science"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.16635","kind":"paper","record_sha256":"f0a011a1fde01ffbf10e56689cea046bc0d30fd10a1de9666d0a524de158cf72","scope":"core","source_url":"https://arxiv.org/abs/2609.16635","title":"EchoPath: Execution-Level Replayable Memory for GUI Agents","topics":["memory","workflow","automation","grounding","efficiency","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.16732","kind":"paper","record_sha256":"79be9c0aefb4ef33ac65d53e1fc4e50bc4733f131d924f24d0b42247e4c5479f","scope":"core","source_url":"https://arxiv.org/abs/2609.16732","title":"When Agents See Differently: Exposing UI Desynchronization Threats in Mobile Agents","topics":["safety-security","screen-understanding","evaluation","robustness"],"year":"2026"},{"environments":["android","desktop"],"id":"arxiv:2609.17653","kind":"paper","record_sha256":"d6355c78a8d2548ea27ee5aa0823dc302711027c2d136f1a10759ea46e817dac","scope":"core","source_url":"https://arxiv.org/abs/2609.17653","title":"Reflect, Revise, Reuse: Training-Free Skill Evolution for GUI Agents","topics":["memory","training-free","self-reflection","robustness","long-horizon","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.17885","kind":"paper","record_sha256":"2a457567e2fa9ebb563e1b3c74c6197d733336e4e2b5f18af53e2fbf0c0ec13d","scope":"core","source_url":"https://arxiv.org/abs/2609.17885","title":"ERPBench: A State-Grounded Evaluation Paradigm for Computer-Use Agents in Enterprise Software","topics":["evaluation","workflow","human-in-the-loop","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.18690","kind":"paper","record_sha256":"7540889767b79f6995de2bf8e7e94bd03249dca136e595f303147ef718eecde3","scope":"core","source_url":"https://arxiv.org/abs/2609.18690","title":"RankGround: Efficient High-Resolution GUI Grounding via Lightweight Reranker-Guided Crop Selection","topics":["grounding","efficiency","data-generation","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.19523","kind":"paper","record_sha256":"b9e78ffd41c7fb761a3bce37e3bcf749a510393cdbe85df5c06e1f0dac09f569","scope":"core","source_url":"https://arxiv.org/abs/2609.19523","title":"EconSkills: Studying Skill Transfer and Retrieval for Web Agents on Live Economic Data","topics":["memory","web-navigation","evaluation","workflow"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.19843","kind":"paper","record_sha256":"03f4c75eeef4867dfbcc6b28b6a3ea301aa7f2ecf064818c5c469a58a24d8991","scope":"core","source_url":"https://arxiv.org/abs/2609.19843","title":"A Dual-Process Perspective on Nudge Susceptibility in LLM-Based GUI Agents","topics":["safety-security","evaluation","ui-design","robustness"],"year":"2026"},{"environments":["desktop","linux","macos","windows","android","web"],"id":"arxiv:2609.22000","kind":"paper","record_sha256":"4842709e61f10b737af7cc1a228eac39fead534bd3375fda6e04954eb5a50799","scope":"core","source_url":"https://arxiv.org/abs/2609.22000","title":"RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents","topics":["evaluation","data-generation","tool-integration","workflow"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2609.22083","kind":"paper","record_sha256":"12a8c5f0d9abe5a3be876e34a62a0389d4a8c72c2618c32266a6a0741f7e1790","scope":"core","source_url":"https://arxiv.org/abs/2609.22083","title":"MintAct: A Unified Visual Agent for Digital Environments","topics":["grounding","reinforcement-learning","data-generation","tool-integration"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.22178","kind":"paper","record_sha256":"6f121d1aef75ad70d695847c6ef0eca5cb40fb3ea2cae11b86d08b9765d5418d","scope":"core","source_url":"https://arxiv.org/abs/2609.22178","title":"Beyond Task Completion: Training Capable and Safe Computer-Use Agents","topics":["safety-security","reinforcement-learning","data-generation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.24277","kind":"paper","record_sha256":"079b85c273d6c72939d168ef3c975f8e8a64064160d8c6048e62c83f3c9f5bb7","scope":"core","source_url":"https://arxiv.org/abs/2609.24277","title":"How Many Pixels Is a Digit Worth? Place-Aware Coordinate Entropy for GUI Agent Confidence Estimation","topics":["grounding","efficiency","evaluation","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.24890","kind":"paper","record_sha256":"753d60dbc30fd615ef40146777e8f890c06589b129c38de1da3eee6f92c1e64b","scope":"core","source_url":"https://arxiv.org/abs/2609.24890","title":"OSWorld-Pro: Process-based Evaluation for Computer Use Agents","topics":["evaluation","observability"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.25769","kind":"paper","record_sha256":"6897fba3420ff396efc7d8aca7b0234f820d4c1485becb92e6a09a1b918dd588","scope":"core","source_url":"https://arxiv.org/abs/2609.25769","title":"Towards Omni-dimensional GUI Agent Navigation with Masked Trajectory Prediction","topics":["imitation-learning","action-prediction","state-tracking","planning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.27273","kind":"paper","record_sha256":"68b5531b09e0f3199cf67415244f78f0744895962784a0849f201c925a409685","scope":"core","source_url":"https://arxiv.org/abs/2609.27273","title":"CAVEAT: Towards Robust Computer-Use Agents in Incentive-Misaligned Environments","topics":["evaluation","robustness","safety-security"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.27307","kind":"paper","record_sha256":"347bc203b3efbb94c4273c4efbba38b5944e9fb28ebfd769d5ba3cde87155245","scope":"core","source_url":"https://arxiv.org/abs/2609.27307","title":"Learn How to Act from Your Own Interactions: On-Policy Self-Distillation for GUI Agents","topics":["imitation-learning","memory","interaction-history","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.27353","kind":"paper","record_sha256":"30309af9e96899cbf9c108594f7fdb190377e7c4dbb9ebba69ceb06dbeefd20f","scope":"core","source_url":"https://arxiv.org/abs/2609.27353","title":"Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents","topics":["evaluation","action-prediction","imitation-learning","grounding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.27461","kind":"paper","record_sha256":"686ef6e3579a89e7ba877e706f7166649e3760ad26d5356e109ed36b8b3bf234","scope":"core","source_url":"https://arxiv.org/abs/2609.27461","title":"Invisible in Space, Visible in Time: Motion Vision CAPTCHA against GUI Agents","topics":["evaluation","screen-understanding","video","safety-security"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.27770","kind":"paper","record_sha256":"6bb46c807f4b82ec93c881fd1fd8f8661365b4873c290a4c21ef18f9651cf1a6","scope":"core","source_url":"https://arxiv.org/abs/2609.27770","title":"Improving LLM-based Autonomous Web Agents with Filtering","topics":["web-navigation","screen-understanding","efficiency","evaluation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.30186","kind":"paper","record_sha256":"433adbd8f7bd6bd69bbaf4dc9af7e568bf80e8c74a453bfc545870bcd3253e67","scope":"core","source_url":"https://arxiv.org/abs/2609.30186","title":"Jev-Mobile: Jev as an Executor for Mobile GUI Agents","topics":["mobile-navigation","efficiency","planning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.30604","kind":"paper","record_sha256":"49c820c4499f391be8a154589a6e72ee31706ca55a37b22116ac3a15ee303d25","scope":"core","source_url":"https://arxiv.org/abs/2609.30604","title":"The Hard Part Comes After Search: Benchmarking Web Agents on Synthesizing, Organizing, and Displaying Knowledge","topics":["evaluation","web-navigation","workflow","long-horizon"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.31957","kind":"paper","record_sha256":"7b1602406a69e319c806f0d5b7c2c8656112f48bb005b97e283351fe837170a2","scope":"core","source_url":"https://arxiv.org/abs/2609.31957","title":"CaptchaArena: A Large-Scale, Fine-Grained Dataset for Training Computer-Use Agents on Interactive CAPTCHAs","topics":["data-generation","grounding","reinforcement-learning"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2609.32036","kind":"paper","record_sha256":"79889a3b620718d9a7d95bcb7a054d803e18ec1ee5ab0ea318eac80bc00a2129","scope":"core","source_url":"https://arxiv.org/abs/2609.32036","title":"ScreenHaystack: Finding Blind Zones in GUI Grounding","topics":["grounding","evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.32130","kind":"paper","record_sha256":"eef9f73b3931e035f37e6750b3d718fcbc70403d4084b006bf6ef07dfbdf1670","scope":"core","source_url":"https://arxiv.org/abs/2609.32130","title":"REBASE: Device-Cloud Experience Coherence for GUI Agents Across App Updates","topics":["memory","robustness","efficiency","mobile-navigation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.32166","kind":"paper","record_sha256":"52583119fac08b80d9ea989057452e8a37a390a5f409b8a3cb3ef5123ee01be6","scope":"core","source_url":"https://arxiv.org/abs/2609.32166","title":"PastForward: Faster On-Device GUI Agents via Computational Experience Reuse","topics":["memory","efficiency","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.32351","kind":"paper","record_sha256":"ae1b6596d2bb9d29a4b2531a9849574efec0c3ba2bd2aa8b3252a509dd523691","scope":"core","source_url":"https://arxiv.org/abs/2609.32351","title":"From Trajectories to Grounded Preferences: Process Preference Synthesis via Interaction Element Graphs for Web PRMs","topics":["data-generation","reward-models","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.32385","kind":"paper","record_sha256":"b0c0454eb32c7fc6cfa98b0b3ae1fa235f115014f30dd6b0a2262bd3cbc64ee9","scope":"core","source_url":"https://arxiv.org/abs/2609.32385","title":"DashAct: A Progressive Diagnostic Benchmark for GUI Agents in Interactive Dashboard Analysis","topics":["evaluation","grounding","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.32600","kind":"paper","record_sha256":"420ed82aff008e5b07eed6ad8fb3e74116e98eed669a48a617a19ad124923a67","scope":"core","source_url":"https://arxiv.org/abs/2609.32600","title":"CUA-SWE: When Computer-Use Agents Meet Visual Software Engineering","topics":["evaluation","workflow","tool-integration"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.32679","kind":"paper","record_sha256":"c4cf315e41a87e6a18497ded8b3567b53e166bec8579f824a4a560454f1a019e","scope":"core","source_url":"https://arxiv.org/abs/2609.32679","title":"The GUI Is Not the State: Diagnosing State Aliasing in GUI World Models","topics":["world-models","state-tracking","evaluation"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2609.32750","kind":"paper","record_sha256":"702367857dc57c56f4021e8f580f173c1d481d7fd4edf3a434e7e08ea83c263b","scope":"core","source_url":"https://arxiv.org/abs/2609.32750","title":"CUA-Sandbox: Efficient Environments for Computer-Use Agent Reinforcement Learning","topics":["sandbox","reinforcement-learning","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.32915","kind":"paper","record_sha256":"a75aeffb173152ef8df7f385dd841cb19824d83b3e232d4f7992582480fe442c","scope":"core","source_url":"https://arxiv.org/abs/2609.32915","title":"AgentTell: Behavioural Side-Channel Leakage in Browser-Use Agents","topics":["safety-security","evaluation","memory"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.33304","kind":"paper","record_sha256":"e0c5ab429fc9cdb219d85115566ba849a9cfe63da0b80c3fff537358c3773156","scope":"core","source_url":"https://arxiv.org/abs/2609.33304","title":"Relevance Does Not Imply Applicability: Experience Activation for Personal GUI Agents","topics":["memory","planning","training-free"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.33646","kind":"paper","record_sha256":"8baa0ce9b0142acf4e625374bec802b646887ae6833d83446f70cb563b805b1d","scope":"core","source_url":"https://arxiv.org/abs/2609.33646","title":"Probe to Act: Elevating Browser-Use Agent via Active Visual Probing","topics":["grounding","memory","web-navigation"],"year":"2026"},{"environments":["android","web"],"id":"arxiv:2609.34113","kind":"paper","record_sha256":"284b1f45bdd8cbe95e60ea9200f7c737b7764a9fb258c2f08424393e4498f058","scope":"core","source_url":"https://arxiv.org/abs/2609.34113","title":"GUITAR: Structured Failure Diagnosis of GUI Agents via State Transitions","topics":["evaluation","observability","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.34139","kind":"paper","record_sha256":"74a85e3105d9edbd5e96b27f44f6db8916c7e8894ef758fd375e06b684f082a3","scope":"core","source_url":"https://arxiv.org/abs/2609.34139","title":"Same Tasks, Different Apps: Why Mobile GUI Agents Fail to Generalize?","topics":["evaluation","robustness","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.34321","kind":"paper","record_sha256":"5e2f753648e27e60b9ffd97d8794c327e236db963c454e7e73a4bb24d8a6a8d3","scope":"core","source_url":"https://arxiv.org/abs/2609.34321","title":"One Rollout Is All You Get: Fully Test-Time Adaptation for GUI Agents","topics":["memory","evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.34769","kind":"paper","record_sha256":"94882267abeb31971223619f3b834146fc9ff501a90772a1483a0a1315695e3e","scope":"adjacent","source_url":"https://arxiv.org/abs/2609.34769","title":"LongPuzzleBench: Evaluating GUI Agents on Long-Horizon Visual Puzzles","topics":["evaluation","planning","long-horizon","state-tracking"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.34974","kind":"paper","record_sha256":"400ff6e0da616efbaeec99063d84d6e8ac01c2423757ca8aac84b216ea31aebd","scope":"core","source_url":"https://arxiv.org/abs/2609.34974","title":"Before Acting, Change the State: Prospective State Intervention for Web Agents under Deceptive Interfaces","topics":["safety-security","planning","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.35026","kind":"paper","record_sha256":"1b0b645b890acd2b0bbfdef8979908f254cc7ca9f6c33b13b83b26b67ff9e38e","scope":"core","source_url":"https://arxiv.org/abs/2609.35026","title":"WebPageBench: Event-Level Verification and Controlled UI-Variant Generation for Web Agents","topics":["evaluation","robustness"],"year":"2026"},{"environments":["android"],"id":"arxiv:2609.35671","kind":"paper","record_sha256":"69202bda2f11c59e5b7fdb49de5063902192eb74f250d2a9d7c8747f7fb8362d","scope":"core","source_url":"https://arxiv.org/abs/2609.35671","title":"PhoneCLI: From App Interfaces to Callable Commands for Mobile Agents","topics":["automation","efficiency","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.35814","kind":"paper","record_sha256":"3582dfff8f34a6c1e0b6e3eaad72b9f3054cb3670e4cb40a86e4d2921f6e4311","scope":"core","source_url":"https://arxiv.org/abs/2609.35814","title":"Constructing Challenging Browser-Use Tasks by Controlled Environment Interventions","topics":["evaluation","robustness","data-generation","state-tracking"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.35904","kind":"paper","record_sha256":"d4e6af3711f1b5dbfb821281a5868debfd66d123c0bb4d03684e38f017a9975d","scope":"core","source_url":"https://arxiv.org/abs/2609.35904","title":"Structured Interaction, Visual Localization, and Robust Execution for Complex Web Tasks: A Technical Report on the WebRetriever Challenge","topics":["web-navigation","grounding","robustness","memory"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.36201","kind":"paper","record_sha256":"b60374384862256bc91a3110b49ab7d3ccf92e4cfde17daee0abe0eb8ae1fac9","scope":"core","source_url":"https://arxiv.org/abs/2609.36201","title":"SCOUT: Synergizing Reasoning and Tool-Use for Computer-Use Safety","topics":["safety-security","evaluation","tool-integration"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.36488","kind":"paper","record_sha256":"87dceb17792c5b4a3f832f9b1e376033ba62ce879f793babc88268b7d73f1a74","scope":"core","source_url":"https://arxiv.org/abs/2609.36488","title":"AdaptArena: Evaluating Test-Time Personalization of Web Agents","topics":["evaluation","memory","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.36923","kind":"paper","record_sha256":"8dd67cb9199e731875e92ed2fec19721bfb926206485cb7449dba2062ceed5af","scope":"core","source_url":"https://arxiv.org/abs/2609.36923","title":"PrecogUI: Proactive GUI Agents via Pre-cognitive Simulation and Experience Retrieval","topics":["memory","world-models","robustness","long-horizon"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.36927","kind":"paper","record_sha256":"04ca32cc8bb34da11944562ffcee4bf42673ffc8867f76ad5d1781d65ea7a1b4","scope":"core","source_url":"https://arxiv.org/abs/2609.36927","title":"Neuro-Symbolic Computer Use: Learning Reusable Policies for Reliable and Efficient Execution","topics":["automation","workflow","efficiency"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.36939","kind":"paper","record_sha256":"af9fc675268e26e79a6f9f7e34a0eba7b9c0a05b600a4a8ac16b3d8a04ab485e","scope":"core","source_url":"https://arxiv.org/abs/2609.36939","title":"SCA: Spatial Credit Assignment for Reinforcement Learning of GUI Agents","topics":["grounding","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2609.37053","kind":"paper","record_sha256":"af14f829f9a646238cf2a38dfc8840ce4e0d5ea9860a84421eda47823dbe226e","scope":"core","source_url":"https://arxiv.org/abs/2609.37053","title":"MatToolBench: Benchmarking Multimodal Agents in Real-World Materials Science Workflows","topics":["science","evaluation","cross-app"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.37176","kind":"paper","record_sha256":"dff1467587cd0b6efb6262cce77cdc3965715bf6f21159d7752ee09c927fcace","scope":"core","source_url":"https://arxiv.org/abs/2609.37176","title":"Absorbed in Inertia: Activation Analysis for Computer-Use Agents","topics":["robustness","observability","memory"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.37686","kind":"paper","record_sha256":"e6853c7c3bcd4fa47adcd16a523b58e4ff3de69142c3edc810ca365fafc39db9","scope":"core","source_url":"https://arxiv.org/abs/2609.37686","title":"EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments?","topics":["evaluation","science","cross-app"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2609.38008","kind":"paper","record_sha256":"3da7d9abb4eaebc094c15c1aaac1e8beed3cc166fd1a01b53920c0d713a707e5","scope":"core","source_url":"https://arxiv.org/abs/2609.38008","title":"HybridCUA: Learning to Orchestrate GUI and CLI for Computer-Use Agents","topics":["tool-integration","reinforcement-learning","data-generation"],"year":"2026"},{"environments":["web","desktop","cross-platform"],"id":"arxiv:2609.38184","kind":"paper","record_sha256":"f583fdf7dff64d8cace008dade8a3961dbfc2c08f008440803dbb6dccedbeeed","scope":"core","source_url":"https://arxiv.org/abs/2609.38184","title":"RealGUINoise: An Interactive Cross-Platform Benchmark for GUI Agent Robustness under Real-World Interface Noise","topics":["robustness","evaluation","safety-security"],"year":"2026"},{"environments":["web"],"id":"arxiv:2609.38778","kind":"paper","record_sha256":"7f0f9b2b00fdfdafdaf9fcd357fa031826f0e8f9559690d63498e444fb5c2567","scope":"core","source_url":"https://arxiv.org/abs/2609.38778","title":"Action Conditioned Bisimulation For GUI Agent Memory","topics":["memory","web-navigation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.38867","kind":"paper","record_sha256":"4bc6604e98970dc279c3ec79efb75be3f2d6fe2887dc3a16ea3f9c5bbe720495","scope":"core","source_url":"https://arxiv.org/abs/2609.38867","title":"Talk2Agent: Benchmarking Voice Interfaces for Text Agents","topics":["evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2609.39547","kind":"paper","record_sha256":"db0e46181ea157ff6f34397f8973fa5688a5b01a3aac68801720bd841822594e","scope":"core","source_url":"https://arxiv.org/abs/2609.39547","title":"Learning Reliable GUI Agents under Imperfect Priors","topics":["data-generation","memory","robustness"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.39903","kind":"paper","record_sha256":"2d4df19b75830a5c8f14103ddec46b1d784390dc3037ac54acaf00299ccac08b","scope":"core","source_url":"https://arxiv.org/abs/2609.39903","title":"OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software","topics":["evaluation","science"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.40253","kind":"paper","record_sha256":"ef0fc61ecc49685fff0d2c89385a370ab4a1ce5b19797f041d60a63a4ee90ee2","scope":"core","source_url":"https://arxiv.org/abs/2609.40253","title":"ComputerSD: Online Self-Distillation from Real-Time Feedback for Computer-Use Agents","topics":["reinforcement-learning","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2609.40284","kind":"paper","record_sha256":"b3757e4dd7143bb4d0e7f4187fe2317560b7b5d737e76a0275e0146ecdd570e4","scope":"core","source_url":"https://arxiv.org/abs/2609.40284","title":"cua-speedrun: Standardized Benchmarking of the Speed of Computer-Use Agents","topics":["evaluation","efficiency"],"year":"2026"},{"environments":["desktop","windows"],"id":"arxiv:2610.00948","kind":"paper","record_sha256":"cbbadf2213bde254b5da91fb2b3ed5704773f9b6647dbcd28271891f27cae273","scope":"core","source_url":"https://arxiv.org/abs/2610.00948","title":"GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution","topics":["automation","observability","robustness"],"year":"2026"},{"environments":["desktop","linux","windows","macos","web"],"id":"arxiv:2610.01215","kind":"paper","record_sha256":"747f049a23f78176fe37e5619386c68bc0381a5b7232ac5ad33564bae5e9c625","scope":"core","source_url":"https://arxiv.org/abs/2610.01215","title":"AutoGUIWorld: Image Generators as Visual World Models for GUI Agent","topics":["world-models","data-generation","grounding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.01491","kind":"paper","record_sha256":"12a69f2459f05a500b80c1b27ff8e4f8e527ac53056d16c569da8936a7f7a7ef","scope":"core","source_url":"https://arxiv.org/abs/2610.01491","title":"Auditing Web Agent Evaluation on WebArena-Lite: Human Review of Outcomes and Trajectories","topics":["evaluation","observability","memory"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.01787","kind":"paper","record_sha256":"374628489ce74adb74aec776be5c005d63e7697cf4f97e2f678dc1a2660af755","scope":"core","source_url":"https://arxiv.org/abs/2610.01787","title":"Not All Experience Belongs in the Weights: Component Routing for Self-Improving GUI Agents","topics":["memory","evaluation","efficiency"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2610.02320","kind":"paper","record_sha256":"1fe94c54f86fd7515fda151ffde96809bbccfafabf7b10a31279991833023acd","scope":"core","source_url":"https://arxiv.org/abs/2610.02320","title":"DeskForge: Dense Supervision from Desktop Environments for Computer-Use Agents","topics":["data-generation","grounding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.02932","kind":"paper","record_sha256":"dcafdf68bb65d9c44976cd22c5c4fb7b9e4a6a4f57a223ac0a8fd60dbb3a2964","scope":"core","source_url":"https://arxiv.org/abs/2610.02932","title":"When to Compile a Computer-Use Agent? Measuring Payback and Making Compilation Decisions for Token Efficiency","topics":["automation","efficiency","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.03036","kind":"paper","record_sha256":"944fb69679263d02432a0458a7ea70eeb738f3dec32faf27d193498e1a525602","scope":"core","source_url":"https://arxiv.org/abs/2610.03036","title":"WebFovea: When the Model Is Right but the Click Is Wrong -- Reliable Round Trips for Vision-Based Web Agents on Live Websites","topics":["web-navigation","robustness","observability"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.03089","kind":"paper","record_sha256":"d48185139177eed10c60740622732055468091035ade8c4b2ac0feadcbf28517","scope":"core","source_url":"https://arxiv.org/abs/2610.03089","title":"Securing Computer-Use Agents Against Branch Steering Attacks","topics":["safety-security","planning","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2610.03818","kind":"paper","record_sha256":"ce215215858c1e585eeb179c90718cc98f50609e89a12d95ef4bcedb320d23ec","scope":"core","source_url":"https://arxiv.org/abs/2610.03818","title":"Control OSWorld: An AI Control Environment for GUI Computer Use Agents","topics":["safety-security","evaluation"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2610.04116","kind":"paper","record_sha256":"293e7d643f84898576b81549c913c6714b8b31736c422f369fbe8610fff4ac63","scope":"adjacent","source_url":"https://arxiv.org/abs/2610.04116","title":"CUAWright: A Minimal Unified Interface for Digital Agents","topics":["tool-integration","efficiency"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2610.04800","kind":"paper","record_sha256":"3b741900cbaf735373a247082acb97b1fecba86bad3b1cff352265b5b466b037","scope":"core","source_url":"https://arxiv.org/abs/2610.04800","title":"MedImageOSWorld: Benchmarking GUI Agents for Medical Image Consoles","topics":["evaluation","science","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.05185","kind":"paper","record_sha256":"840ddff010982b4b3fcc018770087274289bb24fb3ee2b510dfef1cf479f7d50","scope":"core","source_url":"https://arxiv.org/abs/2610.05185","title":"Recurrent Latent Visual Search for GUI Grounding","topics":["grounding","screen-understanding","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.05437","kind":"paper","record_sha256":"aa5bcfdad5ac945ef0fcc2f9ffccd70bb97620634a21e2e04eab860abeb519cd","scope":"core","source_url":"https://arxiv.org/abs/2610.05437","title":"TeleTune: Evolving Agent Skills From Offline Telemetry","topics":["memory","workflow","efficiency"],"year":"2026"},{"environments":["android"],"id":"arxiv:2610.05861","kind":"paper","record_sha256":"48b3ce5782e18ec1dba9eecbdfa5846c9a3035b2ed50e35e3b55af5a869cf0ce","scope":"core","source_url":"https://arxiv.org/abs/2610.05861","title":"Imagine to Act: High-Fidelity Data Synthesis via Image Editing World Model for Scalable GUI Agent Training","topics":["world-models","data-generation","mobile-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.06829","kind":"paper","record_sha256":"2820f1402edf18acb05e286e07e8dc5675713de96220efad12a6e61957820816","scope":"core","source_url":"https://arxiv.org/abs/2610.06829","title":"CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling","topics":["reinforcement-learning","evaluation","web-navigation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.07118","kind":"paper","record_sha256":"40d3e0a766e06404b24a014d754368cdba71c79a3458e9db07445160fbd2f331","scope":"core","source_url":"https://arxiv.org/abs/2610.07118","title":"AMBER: Training Long-Horizon Web Agents through Append-Only Memory","topics":["memory","long-horizon","reinforcement-learning"],"year":"2026"},{"environments":["android"],"id":"arxiv:2610.07444","kind":"paper","record_sha256":"a1cf39bcf91d0dade35dbe45540140fa51b2764f710576a00658050d81e37ea8","scope":"core","source_url":"https://arxiv.org/abs/2610.07444","title":"Decoupling What from Where: How Should a Small GUI Grounding Model Receive the Action Type?","topics":["grounding","evaluation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.07972","kind":"paper","record_sha256":"9bc235588942e27fe52e4c008eda095420d7ce75f0a62e50c036aab150dae5ab","scope":"core","source_url":"https://arxiv.org/abs/2610.07972","title":"Can Agents Work for Everyone? Cross-User Reliability for Mobile GUI Agents in Personalized User Interfaces","topics":["evaluation","robustness","mobile-navigation","reinforcement-learning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.08773","kind":"paper","record_sha256":"fc4d2859c4744d2cc798d2bea27e4857a0cf8e1e17257d29f2d7b7ae26681065","scope":"core","source_url":"https://arxiv.org/abs/2610.08773","title":"AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model","topics":["safety-security","world-models","reinforcement-learning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.09240","kind":"paper","record_sha256":"d3e5672f35cb77892ba7242306ce3b9c8c0e5f238aa53fa42fc7a1e4323339bc","scope":"core","source_url":"https://arxiv.org/abs/2610.09240","title":"Adversarial Images Hijack Web Agents from Visual Grounding to Browser Execution","topics":["safety-security","grounding","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.09469","kind":"paper","record_sha256":"c96353f3d5000ddaf70e3742cd899c6ded2119ce668dada80724c9af2cdba32a","scope":"core","source_url":"https://arxiv.org/abs/2610.09469","title":"Secure-CUA: Controlling Untrusted Influence in Computer-Use Agents","topics":["safety-security","web-navigation"],"year":"2026"},{"environments":["cross-platform"],"id":"arxiv:2610.11050","kind":"paper","record_sha256":"6d08414415bade5d8748687df5e5faffe120ea4f6c20aeb42e056fce121347f1","scope":"core","source_url":"https://arxiv.org/abs/2610.11050","title":"AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks","topics":["evaluation","long-horizon","cross-app"],"year":"2026"},{"environments":["android"],"id":"arxiv:2610.11707","kind":"paper","record_sha256":"b0e24a1387d9817a7ddd9ea36a11f8d055a578676e79eab09cdf56dd78185215","scope":"core","source_url":"https://arxiv.org/abs/2610.11707","title":"DeltaReplay: Task-Relative Memory Reuse for Mobile GUI Agents","topics":["memory","mobile-navigation","planning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.11942","kind":"paper","record_sha256":"1728b3229b5cdf2f5985ee182c5247e2305f80bcb445266ded232afe5fdec8ca","scope":"core","source_url":"https://arxiv.org/abs/2610.11942","title":"Right Screen, Wrong Transition: World Models as Verifiers for GUI Agents","topics":["world-models","safety-security","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.12611","kind":"paper","record_sha256":"cbe687f0eb4b6d56ef8b29e38ae36ad45650019e9194d36591db189780b854e1","scope":"core","source_url":"https://arxiv.org/abs/2610.12611","title":"SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments","topics":["data-generation","evaluation","sandbox"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.12823","kind":"paper","record_sha256":"42dd71133fcbdabcef410e14f4f2515f96123d4730752ce185ef4d146c19ff2a","scope":"core","source_url":"https://arxiv.org/abs/2610.12823","title":"When Charts Deceive Web Agents: A Paired Benchmark for Misleading Visualization-Grounded Tasks","topics":["evaluation","robustness","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.12866","kind":"paper","record_sha256":"04955f0b9cd4f6eb36ef3d3415b25cc2c5914a646c937df37570505934687308","scope":"core","source_url":"https://arxiv.org/abs/2610.12866","title":"CaptchaAgent: Verifiable Visual Reasoning for CAPTCHA Solving","topics":["grounding","screen-understanding","evaluation"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2610.12984","kind":"paper","record_sha256":"70cf3ce76c30839af419a1e7fbb08d58d5eab2de86eedb365db54ee240e2cd14","scope":"core","source_url":"https://arxiv.org/abs/2610.12984","title":"Work2Skill: Capturing Human GUI Experience for Reusable Agent Skill Induction","topics":["recording","workflow","data-generation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.13268","kind":"paper","record_sha256":"f35985ec664ce929c60e55616ddb88b6329de922807f20a51f720dc70e5834a2","scope":"core","source_url":"https://arxiv.org/abs/2610.13268","title":"ConClick: Element-Wise Contrastive Learning Improves GUI Grounding","topics":["grounding","data-generation","robustness"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.13416","kind":"paper","record_sha256":"08dfd34b8d3591925d6c4c3b3a07497e68a7155a7617813b6a481081f0249118","scope":"core","source_url":"https://arxiv.org/abs/2610.13416","title":"Web$hell: A Unified Interface for Agentic Web Access","topics":["web-navigation","tool-integration","memory","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.13703","kind":"paper","record_sha256":"f6cf66e1c1b6148e0644e3827cb010514834874468ca1d8a1a715940c2777b0c","scope":"core","source_url":"https://arxiv.org/abs/2610.13703","title":"A Baseline Assessment for the Agentic Web Technology Stack","topics":["evaluation","web-navigation"],"year":"2026"},{"environments":["android","web"],"id":"arxiv:2610.14157","kind":"paper","record_sha256":"0092343d2e1980fa19087b6c5d4124d80c76a7c87e0f07a86906462b30d97439","scope":"core","source_url":"https://arxiv.org/abs/2610.14157","title":"OpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native Tools","topics":["mobile-navigation","tool-integration","data-generation","reinforcement-learning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.14228","kind":"paper","record_sha256":"3319aad671b2abec7b5d8d9649e78235241e870029cff13d1e248a6e24964f83","scope":"core","source_url":"https://arxiv.org/abs/2610.14228","title":"Latent World Action Model for GUI Agents","topics":["world-models","planning","efficiency"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.15138","kind":"paper","record_sha256":"84690570c17617432fe8db07d64447d4fccd6426b2aef54ecd75c0d217ee2522","scope":"core","source_url":"https://arxiv.org/abs/2610.15138","title":"The Agent Clicks Approve: Injected Page Text Makes Browser Agents Authorize Access the Task Never Named","topics":["safety-security","evaluation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.16234","kind":"paper","record_sha256":"ac9505631e9fd147f355fdad80a6af0f03a9b6c13a6bc2b5fbe40238037c1fae","scope":"core","source_url":"https://arxiv.org/abs/2610.16234","title":"Seg2Click: Learning Pixel-Level Action Masks for GUI Visual Grounding","topics":["grounding","screen-understanding"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.16319","kind":"paper","record_sha256":"353f6b6c8588db2444f9be8df66828d1007763fb2d0f6d85ebfda4c99abf7b37","scope":"core","source_url":"https://arxiv.org/abs/2610.16319","title":"HyWe: Unifying Search and Operation for Real-World Hybrid Web Tasks","topics":["web-navigation","planning","data-generation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.16720","kind":"paper","record_sha256":"a9c7b4ff7db5994861c0d0181bedd85728499be94b947d681dd83ef71e52ce30","scope":"core","source_url":"https://arxiv.org/abs/2610.16720","title":"Learn from Move: the Next Step for GUI Agents","topics":["evaluation","grounding","reinforcement-learning"],"year":"2026"},{"environments":["android"],"id":"arxiv:2610.16778","kind":"paper","record_sha256":"87e8aee502a11fff19f7e41ab56a239f9c23ab5b15ddf512b57ffb1e4ab45f29","scope":"core","source_url":"https://arxiv.org/abs/2610.16778","title":"PrivacyGuard: Source-Linked Sanitization and Locally Resolvable PII References for Cloud GUI Agents","topics":["safety-security","mobile-navigation","evaluation"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.17457","kind":"paper","record_sha256":"a84fd0d21d72ca85605c114cc22c8044911ae9e96138213325a0bbbe798068af","scope":"core","source_url":"https://arxiv.org/abs/2610.17457","title":"BrowseSafe++: Benchmarking Prompt Injection Detection Across Web Contexts, Languages, and Browser Views","topics":["safety-security","evaluation","web-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.17710","kind":"paper","record_sha256":"9ec24b476c71c66ae544612e0703db9555bb4478656bbe4951924c9f33760a31","scope":"core","source_url":"https://arxiv.org/abs/2610.17710","title":"Adaptive-Group Process Rewards for GUI Agents","topics":["reinforcement-learning","long-horizon","planning"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.17792","kind":"paper","record_sha256":"a55ea737c724e6ef168e818393c5ece2a105eda5dfd50b6ccf6d030c15359b6a","scope":"core","source_url":"https://arxiv.org/abs/2610.17792","title":"MHVF: Training-Free GUI Grounding via Verification-Guided Multi-Hypothesis Search","topics":["grounding","screen-understanding","robustness"],"year":"2026"},{"environments":["web","desktop"],"id":"arxiv:2610.17895","kind":"paper","record_sha256":"e2dba726b933a302f1baeaf7436d28206cd40cbe0cca7861f83530e26ab2bc63","scope":"core","source_url":"https://arxiv.org/abs/2610.17895","title":"Interaction-as-a-Judge","topics":["evaluation","data-generation","automation"],"year":"2026"},{"environments":["android"],"id":"arxiv:2610.17928","kind":"paper","record_sha256":"16724f2f84295c4e691cd14ac046abe3bdb3fd96a922dca1a78099b6d19df721","scope":"core","source_url":"https://arxiv.org/abs/2610.17928","title":"Learning How to Remember: Memory Modality Selection for GUI Agents","topics":["memory","mobile-navigation"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.18508","kind":"paper","record_sha256":"9817023e583f5d02bf636f4ffdfd798cf166010cb35540dd2c527874f399cf29","scope":"core","source_url":"https://arxiv.org/abs/2610.18508","title":"Towards Safe Autonomy for Computer Use","topics":["safety-security","screen-understanding"],"year":"2026"},{"environments":["desktop","web"],"id":"arxiv:2610.18784","kind":"paper","record_sha256":"846063969bef68d29d43c8ffe19dc630fd373454d52435074007edf9fcf2ed1d","scope":"adjacent","source_url":"https://arxiv.org/abs/2610.18784","title":"Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents","topics":["memory","long-horizon","efficiency"],"year":"2026"},{"environments":["desktop","web","cross-platform"],"id":"arxiv:2610.19100","kind":"paper","record_sha256":"b584716ffc914fe421b09670ad190dd27c92997bc46ed1e915f7a98d01b0099e","scope":"core","source_url":"https://arxiv.org/abs/2610.19100","title":"GroundGUI: GUI Grounding by High-Quality Data and Distillation-to-Reinforcement Learning","topics":["grounding","data-generation","reinforcement-learning"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.19128","kind":"paper","record_sha256":"dbd63ff8da39b5e4a86ee20634f734f5747c92b6f70d7904839c8c15e4c401d7","scope":"core","source_url":"https://arxiv.org/abs/2610.19128","title":"CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation Pipeline","topics":["evaluation","robustness","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.19834","kind":"paper","record_sha256":"2c338ed927abf171af9fea4029bce79dca536436b6ed990cbb3de32ff76f40ef","scope":"core","source_url":"https://arxiv.org/abs/2610.19834","title":"Teaching Agents in their Own Style: Blackbox Teacher Distillation for Web Agents","topics":["data-generation","web-navigation","efficiency"],"year":"2026"},{"environments":["android"],"id":"arxiv:2610.19900","kind":"paper","record_sha256":"f77ececf1e48102dc69ee36330d1b670cd9c8b4786cc28462c75ee3c774d0e34","scope":"core","source_url":"https://arxiv.org/abs/2610.19900","title":"How GUI Agents Fail: GraceBench for Execution Stability Diagnosis","topics":["evaluation","mobile-navigation","robustness"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.20100","kind":"paper","record_sha256":"cbee41fa5964b21c6c0e648118c78cc5679cfa0260e1471c4502ccf9c666f777","scope":"core","source_url":"https://arxiv.org/abs/2610.20100","title":"Task-State Governance: A Virtual Machine for Verifiable and Recoverable GUI Agent Execution","topics":["human-in-the-loop","safety-security","workflow"],"year":"2026"},{"environments":["web"],"id":"arxiv:2610.20107","kind":"paper","record_sha256":"b02ecda79446b00742ef27ab2893e0e9c1ea707ae701a1345e0045897eb58e44","scope":"core","source_url":"https://arxiv.org/abs/2610.20107","title":"Why Generate a Click? Plug-in Decision Heads for Web Agents","topics":["efficiency","grounding","web-navigation"],"year":"2026"},{"environments":["web","android"],"id":"arxiv:2610.20390","kind":"paper","record_sha256":"80498aa029306c30caa146eda1bc9da2035d592d5c55fa036aa2ceae8129c435","scope":"core","source_url":"https://arxiv.org/abs/2610.20390","title":"STP-GUI: Spatially-Aware Token Pruning for Efficient Web-GUI Agents","topics":["efficiency","grounding","screen-understanding"],"year":"2026"},{"environments":["unspecified"],"id":"arxiv:2610.20584","kind":"paper","record_sha256":"fe884fead09a04832c2871bab3047637fe45de4ba3ade71fb73223a3aef1b2ef","scope":"adjacent","source_url":"https://arxiv.org/abs/2610.20584","title":"Player-1: Learning to Play from Synthetic Experience","topics":["data-generation","reinforcement-learning","evaluation"],"year":"2026"},{"environments":["cross-platform","desktop","web"],"id":"arxiv:2610.20939","kind":"paper","record_sha256":"c8de030460fa026deab4d80e7e45614930d5e9506b5ed00f653dd98e80a4f951","scope":"core","source_url":"https://arxiv.org/abs/2610.20939","title":"MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer Environments","topics":["sandbox","evaluation","reinforcement-learning"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2610.22200","kind":"paper","record_sha256":"198ddcea51a71e3c7dd7c819fb4e0e72be0f660dadde96beec9b68cd1552515b","scope":"core","source_url":"https://arxiv.org/abs/2610.22200","title":"OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use","topics":["tool-integration","planning"],"year":"2026"},{"environments":["desktop","linux"],"id":"arxiv:2610.22234","kind":"paper","record_sha256":"e4b7db0bfaec9fa809f7019ef0a32abeb23869bfe3d5624b7e65d4fc46bb05e0","scope":"core","source_url":"https://arxiv.org/abs/2610.22234","title":"OS-Helix: Unifying GUI Interaction and Code Execution in Native Computer-Using Agents","topics":["tool-integration","reinforcement-learning"],"year":"2026"},{"environments":["desktop"],"id":"arxiv:2610.22280","kind":"paper","record_sha256":"df30c62303ae1501c99ebf1bf4eb23dfeb8a4a37af4df8b67728244ad946a5df","scope":"core","source_url":"https://arxiv.org/abs/2610.22280","title":"LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents","topics":["efficiency","planning"],"year":"2026"},{"environments":["web"],"id":"url:347925d0f440d6396080","kind":"paper","record_sha256":"01a77c7512ce863a9931b100645ec00f5044b615b3474b58fe3e5a44f3e9a95a","scope":"core","source_url":"https://doi.org/10.1145/2983990.2984020","title":"Ringer: Web Automation by Demonstration","topics":["web-navigation","recording","robustness"],"year":"2016"},{"environments":["android"],"id":"url:392a602c660ffa91be46","kind":"paper","record_sha256":"0cfa98dba34fd4680eae70536df7630c50c325dcb4740da19cd4bfd45c764ce9","scope":"core","source_url":"https://doi.org/10.1145/3025453.3025483","title":"SUGILITE: Creating Multimodal Smartphone Automation by Demonstration","topics":["mobile-navigation","recording","human-in-the-loop","cross-app"],"year":"2017"},{"environments":["web"],"id":"url:52a3261eb6a19b402e88","kind":"paper","record_sha256":"c8b28333b0a0f9b95bd44f78e0c71099972c085161fd4ef8ef88dba60513f91e","scope":"core","source_url":"https://proceedings.mlr.press/v70/shi17a.html","title":"World of Bits: An Open-Domain Platform for Web-Based Agents","topics":["web-navigation","reinforcement-learning","evaluation","data-generation"],"year":"2017"},{"environments":["unspecified"],"id":"url:6262c757080f8838e62c","kind":"paper","record_sha256":"7a0e25df3c31170d528bbf5083c959cdd138b7ca53379919bd710c0e97f4302c","scope":"core","source_url":"https://doi.org/10.1145/1753326.1753555","title":"GUI Testing Using Computer Vision","topics":["evaluation","grounding","automation"],"year":"2010"},{"environments":["web"],"id":"url:628b5251757bb3888e89","kind":"paper","record_sha256":"027cbc2eaa97616fd73c40b8f56879880b6e27abb2234d06e1630730ec0f30be","scope":"core","source_url":"https://aaai.org/Papers/AAAI/2007/AAAI07-240.pdf","title":"PLOW: A Collaborative Task Learning Agent","topics":["web-navigation","human-in-the-loop","planning"],"year":"2007"},{"environments":["android"],"id":"url:901ca855640457dfe2f1","kind":"paper","record_sha256":"f08bad499bd2fe1ac17d2631b24a0ef42b167413991f38e06e9fd37934b4b5e0","scope":"core","source_url":"https://doi.org/10.1145/3210240.3210339","title":"Kite: Building Conversational Bots from Mobile Apps","topics":["mobile-navigation","workflow","human-in-the-loop"],"year":"2018"},{"environments":["web"],"id":"url:97c0b1896525632434e2","kind":"paper","record_sha256":"654e5feeb40aa424c08b26730a3b40bac5ff46b166e08f93808717d5dc4245e7","scope":"core","source_url":"https://doi.org/10.1145/3242587.3242661","title":"Rousillon: Scraping Distributed Hierarchical Web Data","topics":["web-navigation","recording","automation"],"year":"2018"},{"environments":["unspecified"],"id":"url:a4cdb2ee0d07d3980651","kind":"paper","record_sha256":"b611a5aaff2aedd8225da5177aa66db9de402324f973e7e34ce2d55cf728e100","scope":"core","source_url":"https://doi.org/10.1145/1622176.1622213","title":"Sikuli: Using GUI screenshots for search and automation","topics":["grounding","automation","screen-understanding"],"year":"2009"},{"environments":["android"],"id":"url:b560b5d6abaab98483f6","kind":"paper","record_sha256":"53785664754fa029cb7086983828b7613148a76964318f17c342e307bc7417fe","scope":"core","source_url":"https://doi.org/10.1145/3126594.3126651","title":"Rico: A Mobile App Dataset for Building Data-Driven Design Applications","topics":["data-generation","screen-understanding","ui-design"],"year":"2017"},{"environments":["android"],"id":"url:ba2c664c1b3c6f1b172f","kind":"paper","record_sha256":"8a550e9a45778b7b359277cd569934dc295899089a4374693848405a9422a24f","scope":"core","source_url":"https://ieeexplore.ieee.org/document/8506506","title":"APPINITE: A Multi-Modal Interface for Specifying Data Descriptions in Programming by Demonstration Using Natural Language Instructions","topics":["grounding","human-in-the-loop","mobile-navigation"],"year":"2018"},{"environments":["android"],"id":"url:beb8f1fd6b945ffc9d93","kind":"paper","record_sha256":"e26ac97c8b5016bec0ef47fa2ef9fd50980f1044d2f9395514fafcefe2f391d1","scope":"core","source_url":"https://openaccess.thecvf.com/content/CVPR2026/html/Tang_GUI-SAGE_Enhancing_GUI_Automation_with_Self-Explanatory_Learning_CVPR_2026_paper.html","title":"GUI-SAGE: Enhancing GUI Automation with Self-Explanatory Learning","topics":["reinforcement-learning","data-generation","evaluation","robustness"],"year":"2026"},{"environments":["desktop"],"id":"url:bf0c11018e49dff6aeec","kind":"paper","record_sha256":"01b367ee77e10cb5175b7d0fd07cfa8cb312704cd94e87bf0a6018fb72004e5a","scope":"core","source_url":"https://acypher.com/Publications/CHI91/EagerCHI.html","title":"Eager: Programming Repetitive Tasks by Example","topics":["automation","human-in-the-loop","interaction-history"],"year":"1991"},{"environments":["web"],"id":"url:cccd967aa993404d74d0","kind":"paper","record_sha256":"1c2bd6da124f938b166e6fa744e1884261656262fa15ce161d0e51e979aa01b4","scope":"core","source_url":"https://doi.org/10.1145/1240624.1240767","title":"Koala: Capture, Share, Automate, Personalize Business Processes on the Web","topics":["web-navigation","workflow","recording"],"year":"2007"}],"reviewed_resources":1572,"reviewed_surveys":15,"snapshot_as_of_utc":"2026-10-18T04:40:40Z","source_commit":"87166fe8ad9c8e832f7a90ee84b2e93c6e0e40db","source_repository":"https://github.com/reacher-z/Awesome-CUA","year_policy":"First-submission year for arXiv records; supplied publication year for publisher-only records. Partial dates retain their precision."}