{"article_title":"Computer-Use Agents: Methods, Evaluation, Human Agency, and Research Assets","catalog_sha256":"7f71a1683ab013e587a20086c3e357956981f517a24ae8bc8e81791108c66e4f","counts":{"abstract_and_project_only_works":1,"dataset_snapshots":1,"fresh_arxiv_bibliography_additions":14,"fulltext_section_read_works":87,"unique_cited_works":89},"read_boundary":"Selected sections, not entire-document audits or independently reproduced experiments. Works are deduplicated by retained canonical identity.","records":[{"canonical_id":"arxiv:2501.16150","citation_key":"arxiv_2501_16150","citation_url":"https://arxiv.org/abs/2501.16150v3","edition_label":"arXiv v3; first submitted 2025-01-27","ordered_authors":["Pascal J. Sager","Benjamin Meyer","Peng Yan","Rebekka von Wartburg-Kottler","Layan Etaiwi","Aref Enayati","Gabriel Nobel","Ahmed Abdulkadir","Benjamin F. Grewe","Thilo Stadelmann"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:08:02Z","sha256":"e9ca21daad357fee1442c6a83eecaa1c907255bdbd7fe4606c49d56281d94f44","url":"https://arxiv.org/abs/2501.16150v3"},{"read_sections":["1 Introduction","1.1 Relation to Other Surveys","2.1 Definitions","2.2 A Comprehensive Taxonomy","7 Agent Evaluation","8 Conclusions"],"retrieved_at":"2026-10-18T10:08:58Z","sha256":"2684d4124758b38973f20aa01aaee1040a03513f8cda2fe7cc8a8df1e9e200a1","url":"https://arxiv.org/html/2501.16150v3"}],"read_level":"selected_fulltext_sections_read","sections":["1 Introduction","1.1 Relation to Other Surveys","2.1 Definitions","2.2 A Comprehensive Taxonomy","7 Agent Evaluation","8 Conclusions"]}],"reference_number":1,"title":"A Comprehensive Survey of Agents for Computer Use: Foundations, Challenges, and Future Directions"},{"canonical_id":"arxiv:2508.04482","citation_key":"arxiv_2508_04482","citation_url":"https://arxiv.org/abs/2508.04482v1","edition_label":"arXiv v1; first submitted 2025-08-06","ordered_authors":["Xueyu Hu","Tao Xiong","Biao Yi","Zishu Wei","Ruixuan Xiao","Yurun Chen","Jiasheng Ye","Meiling Tao","Xiangxin Zhou","Ziyu Zhao","Yuhuai Li","Shengze Xu","Shenzhi Wang","Xinchen Xu","Shuofei Qiao","Zhaokai Wang","Kun Kuang","Tieyong Zeng","Liang Wang","Jiwei Li","Yuchen Eleanor Jiang","Wangchunshu Zhou","Guoyin Wang","Keting Yin","Zhou Zhao","Hongxia Yang","Fan Wu","Shengyu Zhang","Fei Wu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:08:28Z","sha256":"dd508a6ad544585c40f2c84fe79257408d5a6299d5b5f656e787362f48c26265","url":"https://arxiv.org/abs/2508.04482v1"},{"read_sections":["1 Introduction","2 Fundamentals including 2.1–2.2","5 Challenges and Future","6 Related Work","7 Conclusion"],"retrieved_at":"2026-10-18T10:09:26Z","sha256":"03855e424943d708cd568e89b55873a4e6b19421f7395e4d2cc202c9fbfdfa65","url":"https://arxiv.org/html/2508.04482v1"}],"read_level":"selected_fulltext_sections_read","sections":["1 Introduction","2 Fundamentals including 2.1–2.2","5 Challenges and Future","6 Related Work","7 Conclusion"]}],"reference_number":2,"title":"OS Agents: A Survey on MLLM-based Agents for General Computing Devices Use"},{"canonical_id":"arxiv:2411.18279","citation_key":"arxiv_2411_18279","citation_url":"https://arxiv.org/abs/2411.18279v12","edition_label":"arXiv v12; first submitted 2024-11-27","ordered_authors":["Chaoyun Zhang","Shilin He","Jiaxu Qian","Bowen Li","Liqun Li","Si Qin","Yu Kang","Minghua Ma","Guyue Liu","Qingwei Lin","Saravan Rajmohan","Dongmei Zhang","Qi Zhang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:07:52Z","sha256":"abf172a70f8d71624a484a27de7f18c74a4ab526df393f62ecaa8691502fe0dd","url":"https://arxiv.org/abs/2411.18279v12"},{"read_sections":["I Introduction and I-B Scope","IV-D GUI Agent vs API-Based Agent","V-A Architecture and Workflow","XI-C Safety and Reliability","XI-D Human-Agent Interaction"],"retrieved_at":"2026-10-18T10:08:43Z","sha256":"e451504bec33942bcc232dc0879c44e5502158d53ae3040d5d20d85160f90a33","url":"https://arxiv.org/html/2411.18279v12"}],"read_level":"selected_fulltext_sections_read","sections":["I Introduction and I-B Scope","IV-D GUI Agent vs API-Based Agent","V-A Architecture and Workflow","XI-C Safety and Reliability","XI-D Human-Agent Interaction"]}],"reference_number":3,"title":"Large Language Model-Brained GUI Agents: A Survey"},{"canonical_id":"arxiv:2210.03347","citation_key":"arxiv_2210_03347","citation_url":"https://arxiv.org/abs/2210.03347v2","edition_label":"arXiv v2; first submitted 2022-10-07","ordered_authors":["Kenton Lee","Mandar Joshi","Iulia Turc","Hexiang Hu","Fangyu Liu","Julian Eisenschlos","Urvashi Khandelwal","Peter Shaw","Ming-Wei Chang","Kristina Toutanova"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:10:20Z","sha256":"ab98a4f14402b4d2fbccfbfa052eb9a98f505e858f23f30eee97f943532fe688","url":"https://arxiv.org/abs/2210.03347v2"},{"read_sections":["Abstract","1 Introduction","2.2 Architecture","2.3 Pretraining","2.5 Finetuning","3.1 Benchmarks","4.2 UIs","6 Discussion"],"retrieved_at":"2026-10-18T10:10:25Z","sha256":"2b36f4959ea6714746a4db8b8fc13a19d15b1365fc20481f870cd53eeec14bf0","url":"https://arxiv.org/html/2210.03347v2"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","2.2 Architecture","2.3 Pretraining","2.5 Finetuning","3.1 Benchmarks","4.2 UIs","6 Discussion"]}],"reference_number":4,"title":"Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding"},{"canonical_id":"url:cccd967aa993404d74d0","citation_key":"cua_1b4a3a269ca9f1884a566476","citation_url":"https://doi.org/10.1145/1240624.1240767","edition_label":"2007","ordered_authors":["Greg Little","Tessa A. Lau","Allen Cypher","James Lin","Eben M. Haber","Eser Kandogan"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-17T09:53:23Z","sha256":"ae61a94cca68e058858244b6ff54ed9409973a5cecaf25c0e325b27adaf0e3dd","url":"https://research.ibm.com/publications/koala-capture-share-automate-personalize-business-processes-on-the-web"},{"read_sections":["PDF printed pp. 943–946: abstract, introduction, office-products use case, interpreter, personalization, wiki, conclusion"],"retrieved_at":"2026-10-17T09:55:21Z","sha256":"944b6f74a6cdb88fc253009f364dcc71cc42186b4907e04720da128c5292e357","url":"https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf"}],"read_level":"selected_fulltext_sections_read","sections":["PDF printed pp. 943–946: abstract, introduction, office-products use case, interpreter, personalization, wiki, conclusion"]}],"reference_number":5,"title":"Koala: Capture, Share, Automate, Personalize Business Processes on the Web"},{"canonical_id":"url:a4cdb2ee0d07d3980651","citation_key":"cua_510831381c08385eb74fb95e","citation_url":"https://doi.org/10.1145/1622176.1622213","edition_label":"2009-10","ordered_authors":["Tom Yeh","Tsung-Hsiang Chang","Robert C. Miller"],"read_level":"abstract_and_author_project_only","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["Complete MIT institutional abstract and issued-date/authors/published citation","Author project page: Sikuli Script and Sikuli Search"],"retrieved_at":"2026-10-17T09:53:21Z","sha256":"dbaaf21982522b34857be014b39d6c43fedcda20f808efe40d04958c77243a5a","url":"https://dspace.mit.edu/entities/publication/604a3f79-451d-4186-a94e-d12690ffe40a"},{"read_sections":["Complete MIT institutional abstract and issued-date/authors/published citation","Author project page: Sikuli Script and Sikuli Search"],"retrieved_at":"2026-10-18T10:11:51Z","sha256":"06c5dd8c68b6e1b4da97de0fabd575b94969dbfb7072626890e8a9718f489615","url":"https://people.csail.mit.edu/vgod/projects.shtml"}],"read_level":"abstract_and_author_project_only","sections":["Complete MIT institutional abstract and issued-date/authors/published citation","Author project page: Sikuli Script and Sikuli Search"]}],"reference_number":6,"title":"Sikuli: Using GUI screenshots for search and automation"},{"canonical_id":"url:392a602c660ffa91be46","citation_key":"cua_6b7f9d608ed89d1625610eb1","citation_url":"https://doi.org/10.1145/3025453.3025483","edition_label":"2017","ordered_authors":["Toby Jia-Jun Li","Amos Azaria","Brad A. Myers"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-17T09:53:22Z","sha256":"32ee05ff1a91ed4a22d4bc156a462872aa4e28f5c79abfbce5dbf3655f412033","url":"https://www.cs.cmu.edu/~NatProg/sugilite.html"},{"read_sections":["PDF printed pp. 6038–6039: abstract, introduction and contributions","Execution Handler and Technical Limitations/Semantics discussion on p. 6044","Conclusion on p. 6047","CMU project 2017 publication entry"],"retrieved_at":"2026-10-17T09:54:39Z","sha256":"cc23c19284f0a558c4b99b8b051abca8ed61e274e9e1f1fed1ae91487f4368ad","url":"https://toby.li/files/TobyLi-CHI2017-Sugilite.pdf"}],"read_level":"selected_fulltext_sections_read","sections":["PDF printed pp. 6038–6039: abstract, introduction and contributions","Execution Handler and Technical Limitations/Semantics discussion on p. 6044","Conclusion on p. 6047","CMU project 2017 publication entry"]}],"reference_number":7,"title":"SUGILITE: Creating Multimodal Smartphone Automation by Demonstration"},{"canonical_id":"url:b560b5d6abaab98483f6","citation_key":"cua_402ed654426712089390317c","citation_url":"https://doi.org/10.1145/3126594.3126651","edition_label":"2017-10-20","ordered_authors":["Biplab Deka","Zifeng Huang","Chad Franzen","Joshua Hibschman","Daniel Afergan","Yang Li","Jeffrey Nichols","Ranjitha Kumar"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["Publisher PDF cover date and printed p. 845 title/ordered authors/abstract","Introduction and Android App Datasets","Design Data Organization on p. 850","Future Work on p. 851"],"retrieved_at":"2026-10-17T10:01:32Z","sha256":"a1ec58c526b152c052c5c72f9a62725f1026b5f0e4f45c8fe85fce284295da84","url":"https://danafergan.com/publications/deka2017rico.pdf"},{"read_sections":["Publisher PDF cover date and printed p. 845 title/ordered authors/abstract","Introduction and Android App Datasets","Design Data Organization on p. 850","Future Work on p. 851"],"retrieved_at":"2026-10-17T10:01:32Z","sha256":"a1ec58c526b152c052c5c72f9a62725f1026b5f0e4f45c8fe85fce284295da84","url":"https://danafergan.com/publications/deka2017rico.pdf"}],"read_level":"selected_fulltext_sections_read","sections":["Publisher PDF cover date and printed p. 845 title/ordered authors/abstract","Introduction and Android App Datasets","Design Data Organization on p. 850","Future Work on p. 851"]}],"reference_number":8,"title":"Rico: A Mobile App Dataset for Building Data-Driven Design Applications"},{"canonical_id":"url:52a3261eb6a19b402e88","citation_key":"cua_ba263493b78b0263c7933138","citation_url":"https://proceedings.mlr.press/v70/shi17a.html","edition_label":"2017-07-17","ordered_authors":["Tianlin Shi","Andrej Karpathy","Linxi Fan","Jonathan Hernandez","Percy Liang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-17T09:53:21Z","sha256":"749d2a150e4c2bcde1b219a0ecb86c95546134a057892544e8495b5e45323c44","url":"https://proceedings.mlr.press/v70/shi17a.html"},{"read_sections":["Publisher metadata","PDF abstract and 1 Introduction","2.1 Web as an Environment","2.2 MiniWoB","2.3 FormWoB"],"retrieved_at":"2026-10-17T09:54:39Z","sha256":"087008efb9d056df5fd9fbc724413d6a3f20cb14c7f1597d00bacac53d8342d6","url":"https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"}],"read_level":"selected_fulltext_sections_read","sections":["Publisher metadata","PDF abstract and 1 Introduction","2.1 Web as an Environment","2.2 MiniWoB","2.3 FormWoB"]},{"chapter":"evaluation","pdf_pages":[1,2,3,4],"primary_sources":[{"retrieved_at":"2026-10-18T10:00:29.002410+00:00","sha256":"087008efb9d056df5fd9fbc724413d6a3f20cb14c7f1597d00bacac53d8342d6","url":"https://proceedings.mlr.press/v70/shi17a/shi17a.pdf","version":null}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Web as an Environment","2.2 Minimalistic Web Tasks: MiniWoB","2.3 Live Web Tasks: FormWoB","2.4 Crowdsourcing Web Tasks at Scale: QAWoB"]}],"reference_number":9,"title":"World of Bits: An Open-Domain Platform for Web-Based Agents"},{"canonical_id":"arxiv:1802.08802","citation_key":"arxiv_1802_08802","citation_url":"https://arxiv.org/abs/1802.08802v1","edition_label":"arXiv v1; first submitted 2018-02-24","ordered_authors":["Evan Zheran Liu","Kelvin Guu","Panupong Pasupat","Tianlin Shi","Percy Liang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:10:05Z","sha256":"df56095550e7d7560632209a4d86ce11e2e4e219df619c20239b739bd28484c7","url":"https://arxiv.org/abs/1802.08802v1"},{"read_sections":["Abstract","1 Introduction","2 Setup","3 Inducing workflows","6.1 Task setups","6.3.1 MiniWoB++"],"retrieved_at":"2026-10-18T10:10:08Z","sha256":"0bd8bfa1d6c028648e975a20df037c74e10995e02b93aece0ef18f9e05d48e8d","url":"https://arxiv.org/html/1802.08802v1"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","2 Setup","3 Inducing workflows","6.1 Task setups","6.3.1 MiniWoB++"]},{"chapter":"evaluation","pdf_pages":[6,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:00:31.908500+00:00","sha256":"7a7ce9d5f8f31c7f7c01a35271c5bdfdf2697dfe17e72ff090a9465d55f7f60d","url":"https://arxiv.org/pdf/1802.08802v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["6.1 Benchmarks: Environment and Evaluation metric","6.3.1 MiniWoB++ benchmark"]}],"reference_number":10,"title":"Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration"},{"canonical_id":"arxiv:2307.10088","citation_key":"arxiv_2307_10088","citation_url":"https://arxiv.org/abs/2307.10088v2","edition_label":"arXiv v2; first submitted 2023-07-19","ordered_authors":["Christopher Rawles","Alice Li","Daniel Rodriguez","Oriana Riva","Timothy Lillicrap"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:10:47Z","sha256":"bcc253aebf71fb693401e78ccb26e8ffa1eca42d51840a1708d53c1e2a62d3ad","url":"https://arxiv.org/abs/2307.10088v2"},{"read_sections":["Abstract","1 Introduction","3 Android in the Wild including 3.1–3.2","5.2 Evaluation methodology and metrics","6.1 Data Limitations","6.2 Ethical considerations"],"retrieved_at":"2026-10-18T10:10:51Z","sha256":"d326d463c42951c379dd541534e6329f4a032e8b815486503b691caf84a963e2","url":"https://arxiv.org/html/2307.10088v2"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","3 Android in the Wild including 3.1–3.2","5.2 Evaluation methodology and metrics","6.1 Data Limitations","6.2 Ethical considerations"]},{"chapter":"evaluation","pdf_pages":[4,5,8,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:36.415524+00:00","sha256":"3e4ace8fe18dedc903144894cb69a57f5e0dfbb75f7f23de52bff91f08b11e5a","url":"https://arxiv.org/pdf/2307.10088v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3 Dataset collection","3.1 Trajectories","3.2 Hindsight language relabeling","5.2 Evaluation methodology","5.3 Human comparison","6.1 Limitations"]}],"reference_number":11,"title":"Android in the Wild: A Large-Scale Dataset for Android Device Control"},{"canonical_id":"arxiv:2207.01206","citation_key":"arxiv_2207_01206","citation_url":"https://arxiv.org/abs/2207.01206v4","edition_label":"arXiv v4; first submitted 2022-07-04","ordered_authors":["Shunyu Yao","Howard Chen","John Yang","Karthik Narasimhan"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:10:13Z","sha256":"4df392cd073a841564fa30bc46bc1dfe9e0078b40bc2aebcc79982fcf682eda1","url":"https://arxiv.org/abs/2207.01206v4"},{"read_sections":["Abstract","1 Introduction","3.1 Task Formulation","3.2 Environment Implementation","Appendix E Potential Societal Impacts and Limitations"],"retrieved_at":"2026-10-18T10:10:17Z","sha256":"2c25234743e842bcc9686d61ef007f6ca26302bd5a7fc0321a97624425c75090","url":"https://arxiv.org/html/2207.01206v4"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","3.1 Task Formulation","3.2 Environment Implementation","Appendix E Potential Societal Impacts and Limitations"]}],"reference_number":12,"title":"WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents"},{"canonical_id":"arxiv:2306.06070","citation_key":"arxiv_2306_06070","citation_url":"https://arxiv.org/abs/2306.06070v3","edition_label":"arXiv v3; first submitted 2023-06-09","ordered_authors":["Xiang Deng","Yu Gu","Boyuan Zheng","Shijie Chen","Samuel Stevens","Boshi Wang","Huan Sun","Yu Su"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:10:30Z","sha256":"fe6248ffd11553e721f204cdf760e6b56362febcfe1e1cb26fcf1d5142ea2e8a","url":"https://arxiv.org/abs/2306.06070v3"},{"read_sections":["Abstract","1 Introduction","2.1 Task Definition","2.2 Data Collection","3 Method overview","6 Limitations and Potential Societal Impact"],"retrieved_at":"2026-10-18T10:10:34Z","sha256":"92f1e770ec28a36dd71a3381b67fc183efffb623d0803c5d7c11e298a6f2aeec","url":"https://arxiv.org/html/2306.06070v3"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","2.1 Task Definition","2.2 Data Collection","3 Method overview","6 Limitations and Potential Societal Impact"]},{"chapter":"methods","pdf_pages":[1,5,6,7,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:00:08.226086+00:00","sha256":"75d8af953769a53bf75913908e2bb6185c6b338c8d89b7304e7fd8cfb79934e8","url":"https://arxiv.org/pdf/2306.06070v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Candidate Generation","3.2 Action Prediction","4.1 Experimental Setup","4.2 Evaluation","6 Limitations"]},{"chapter":"evaluation","pdf_pages":[3,4,5,7,8],"primary_sources":[{"retrieved_at":"2026-10-18T10:00:52.831447+00:00","sha256":"75d8af953769a53bf75913908e2bb6185c6b338c8d89b7304e7fd8cfb79934e8","url":"https://arxiv.org/pdf/2306.06070v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Task Definition","2.2 Data Collection","2.3 Comparison","4.1 Experimental Setup","4.2 Data Preprocessing and Evaluation","4.3 Results"]}],"reference_number":13,"title":"Mind2Web: Towards a Generalist Agent for the Web"},{"canonical_id":"arxiv:2307.13854","citation_key":"arxiv_2307_13854","citation_url":"https://arxiv.org/abs/2307.13854v4","edition_label":"arXiv v4; first submitted 2023-07-25","ordered_authors":["Shuyan Zhou","Frank F. Xu","Hao Zhu","Xuhui Zhou","Robert Lo","Abishek Sridhar","Xianyi Cheng","Tianyue Ou","Yonatan Bisk","Daniel Fried","Uri Alon","Graham Neubig"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:10:38Z","sha256":"2bdd9cdee9419df180a60202c335104a779d1b7229918f038f0e59780f8b4506","url":"https://arxiv.org/abs/2307.13854v4"},{"read_sections":["Abstract","1 Introduction","2 Environment including 2.1–2.4","3 benchmark overview","3.2 Evaluation Annotation"],"retrieved_at":"2026-10-18T10:10:42Z","sha256":"15c4dcd5968f24bf21f1bdc6696be4180bdeed58f2703345ed7e11da89cedd2d","url":"https://arxiv.org/html/2307.13854v4"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","2 Environment including 2.1–2.4","3 benchmark overview","3.2 Evaluation Annotation"]},{"chapter":"evaluation","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:22.276000+00:00","sha256":"15c4dcd5968f24bf21f1bdc6696be4180bdeed58f2703345ed7e11da89cedd2d","url":"https://arxiv.org/html/2307.13854v4","version":4}],"read_level":"selected_fulltext_sections_read","sections":["HTML §2.2–2.4 (#S2.SS2–S2.SS4)","HTML §3.1 Intent Analysis (#S3.SS1.SSS0.Px1)","HTML §3.2 Evaluation Annotation (#S3.SS2)"]}],"reference_number":14,"title":"WebArena: A Realistic Web Environment for Building Autonomous Agents"},{"canonical_id":"arxiv:2404.07972","citation_key":"arxiv_2404_07972","citation_url":"https://arxiv.org/abs/2404.07972v2","edition_label":"arXiv v2; first submitted 2024-04-11","ordered_authors":["Tianbao Xie","Danyang Zhang","Jixuan Chen","Xiaochuan Li","Siheng Zhao","Ruisheng Cao","Toh Jing Hua","Zhoujun Cheng","Dongchan Shin","Fangyu Lei","Yitao Liu","Yiheng Xu","Shuyan Zhou","Silvio Savarese","Caiming Xiong","Victor Zhong","Tao Yu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:10:57Z","sha256":"7d2e5e48d63786bc29736a610e60a83b92ef07cd36424079496a60d7aedd685b","url":"https://arxiv.org/abs/2404.07972v2"},{"read_sections":["Abstract","1 Introduction","2.1 Task Definition","2.2 environment infrastructure and evaluation","2.3 Observation Space","3.1–3.2 operating systems and tasks","7 Conclusion and Future Work"],"retrieved_at":"2026-10-18T10:11:00Z","sha256":"fbd620dccd72032406ceb36d4d9e108292b1bc7fcbf4c3aee0c0b544601a05c2","url":"https://arxiv.org/html/2404.07972v2"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","2.1 Task Definition","2.2 environment infrastructure and evaluation","2.3 Observation Space","3.1–3.2 operating systems and tasks","7 Conclusion and Future Work"]},{"chapter":"evaluation","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:29.042609+00:00","sha256":"fbd620dccd72032406ceb36d4d9e108292b1bc7fcbf4c3aee0c0b544601a05c2","url":"https://arxiv.org/html/2404.07972v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["HTML §2.2 Infrastructure (#S2.SS2)","HTML §2.3–2.4 Interfaces","HTML §3.1–3.2 Benchmark and tasks","HTML §7 Safety discussion (#S7.SS0.SSS0.Px3)"]}],"reference_number":15,"title":"OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments"},{"canonical_id":"arxiv:2405.14573","citation_key":"arxiv_2405_14573","citation_url":"https://arxiv.org/abs/2405.14573v5","edition_label":"arXiv v5; first submitted 2024-05-23","ordered_authors":["Christopher Rawles","Sarah Clinckemaillie","Yifan Chang","Jonathan Waltz","Gabrielle Lau","Marybeth Fair","Alice Li","William Bishop","Wei Li","Folawiyo Campbell-Ajala","Daniel Toyama","Robert Berry","Divya Tyamagundlu","Timothy Lillicrap","Oriana Riva"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"retrieved_at":"2026-10-18T10:11:03Z","sha256":"84824419227fb39f3540bc7074b7b788aec0c241c6721b60e515ca2e41c60209","url":"https://arxiv.org/abs/2405.14573v5"},{"read_sections":["Abstract","1 Introduction","3.1–3.5 AndroidWorld environment","Appendix A Limitations"],"retrieved_at":"2026-10-18T10:11:07Z","sha256":"d3bdc04b692f57e3a933d67ab955fae790c1fe5deb7b5bcb71d80d24b8d2cdf6","url":"https://arxiv.org/html/2405.14573v5"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","3.1–3.5 AndroidWorld environment","Appendix A Limitations"]},{"chapter":"evaluation","pdf_pages":[5,6,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:24.587650+00:00","sha256":"5673a60129459d8b8c0efbe3219ffc2ff2b311a3560e50cc488fa03eefcc35d1","url":"https://arxiv.org/pdf/2405.14573v5","version":5}],"read_level":"selected_fulltext_sections_read","sections":["3.2 Observation/action","3.3 Reproducible parameterized tasks","3.4 State rewards","5 Robustness analysis"]}],"reference_number":16,"title":"AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents"},{"canonical_id":"arxiv:2411.04890","citation_key":"arxiv_2411_04890","citation_url":"https://arxiv.org/abs/2411.04890v2","edition_label":"arXiv v2; first submitted 2024-11-07","ordered_authors":["Shuai Wang","Weiwen Liu","Jingxuan Chen","Yuqi Zhou","Weinan Gan","Xingshan Zeng","Yuhan Che","Shuai Yu","Xinlong Hao","Kun Shao","Bin Wang","Chuhan Wu","Yasheng Wang","Ruiming Tang","Jianye Hao"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:07:44Z","sha256":"422c0e849422eb2c974eb8df661ffdc3b64af141b07a92274605a1baa8ce84ae","url":"https://arxiv.org/abs/2411.04890v2"},{"read_sections":["Abstract","1 Introduction","2 GUI Agent Data Resources","3.1 GUI Agent Framework","3.2 modality and learning taxonomy"],"retrieved_at":"2026-10-18T10:08:38Z","sha256":"8d5145cda2485311be2b6d937715efdc33df3d53bb0de995a6f411d1efd43da2","url":"https://arxiv.org/html/2411.04890v2"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","2 GUI Agent Data Resources","3.1 GUI Agent Framework","3.2 modality and learning taxonomy"]}],"reference_number":17,"title":"GUI Agents with Foundation Models: A Comprehensive Survey"},{"canonical_id":"arxiv:2503.23350","citation_key":"arxiv_2503_23350","citation_url":"https://arxiv.org/abs/2503.23350v4","edition_label":"arXiv v4; first submitted 2025-03-30","ordered_authors":["Liangbo Ning","Ziran Liang","Zhuohang Jiang","Haohao Qu","Yujuan Ding","Wenqi Fan","Xiao-yong Wei","Shanru Lin","Hui Liu","Philip S. Yu","Qing Li"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:08:06Z","sha256":"7952424ff757a731195e8633701da2befb7fe329b0765f2497e695168a597b17","url":"https://arxiv.org/abs/2503.23350v4"},{"read_sections":["Abstract","1 Introduction","2.3 AI Agents for Web Automation","5 Trustworthy WebAgents including 5.1–5.3"],"retrieved_at":"2026-10-18T10:09:04Z","sha256":"56200d817811c55e8d49edfe6da671c24dda0965efe891bb8109fbff258cc729","url":"https://arxiv.org/html/2503.23350v4"}],"read_level":"selected_fulltext_sections_read","sections":["Abstract","1 Introduction","2.3 AI Agents for Web Automation","5 Trustworthy WebAgents including 5.1–5.3"]}],"reference_number":18,"title":"A Survey of WebAgents: Towards Next-Generation AI Agents for Web Automation with Large Foundation Models"},{"canonical_id":"arxiv:2504.19838","citation_key":"arxiv_2504_19838","citation_url":"https://arxiv.org/abs/2504.19838v3","edition_label":"arXiv v3; first submitted 2025-04-28","ordered_authors":["Guangyi Liu","Pengxiang Zhao","Yaozhen Liang","Liang Liu","Yaxuan Guo","Han Xiao","Weifeng Lin","Yuxiang Chai","Yue Han","Shuai Ren","Hao Wang","Xiaoyu Liang","WenHao Wang","Tianze Wu","Zhengxi Lu","Siheng Chen","LiLinghao","Hao Wang","Guanjing Xiong","Yong Liu","Hongsheng Li"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:08:17Z","sha256":"5fc92d139141c12f25462b8b0ccb125d521e67c5b8b54873dfa16e633d732f6b","url":"https://arxiv.org/abs/2504.19838v3"},{"read_sections":["1 Introduction including scope paragraph","4 introductory framework/POMDP description","4.1.2 Phone State Perception","7 Challenges and Future Directions","8 Conclusion"],"retrieved_at":"2026-10-18T10:09:10Z","sha256":"a5d1d3f46469093b249db4a5af12a3ccb628b03348d43ef37f7d1254cbf6d476","url":"https://arxiv.org/html/2504.19838v3"}],"read_level":"selected_fulltext_sections_read","sections":["1 Introduction including scope paragraph","4 introductory framework/POMDP description","4.1.2 Phone State Perception","7 Challenges and Future Directions","8 Conclusion"]}],"reference_number":19,"title":"LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects"},{"canonical_id":"arxiv:2503.23434","citation_key":"arxiv_2503_23434","citation_url":"https://arxiv.org/abs/2503.23434v2","edition_label":"arXiv v2; first submitted 2025-03-30","ordered_authors":["Yucheng Shi","Wenhao Yu","Jingyuan Huang","Wenlin Yao","Wenhu Chen","Ninghao Liu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:08:10Z","sha256":"e2a4ec9fdc16c504818fdad8d6608c3c7668675b13529a7952650696fa607ced","url":"https://arxiv.org/abs/2503.23434v2"},{"read_sections":["1 Introduction","2 Foundations including 2.1–2.3","3 Perception Trust","6 Evaluation Methodologies","7 Conclusion","Limitations"],"retrieved_at":"2026-10-18T10:09:32Z","sha256":"f5d92f991e1560995e2cd1a2ff6ad6bf0e62c55ff8d653b22f64f479236b1dd0","url":"https://arxiv.org/html/2503.23434v2"}],"read_level":"selected_fulltext_sections_read","sections":["1 Introduction","2 Foundations including 2.1–2.3","3 Perception Trust","6 Evaluation Methodologies","7 Conclusion","Limitations"]}],"reference_number":20,"title":"Towards Trustworthy GUI Agents: A Survey"},{"canonical_id":"arxiv:2609.02309","citation_key":"arxiv_2609_02309","citation_url":"https://arxiv.org/abs/2609.02309v1","edition_label":"arXiv v1; first submitted 2026-09-02","ordered_authors":["Bizhe Bai","Jiakang Yuan","Hongming Wu","Xinyue Wang","Jie Ren","Siyao Chen","Yuchen Ya","Fan Bai","Pai Peng","Huafeng Qin","Tao Chen"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"context","pdf_pages":[],"primary_sources":[{"read_sections":["title","ordered authors","submission history","version","complete abstract"],"retrieved_at":"2026-10-18T10:08:35Z","sha256":"64080b2f1f34edcac11eef55855e43e0175df31b2bbb14427a0b3d57ad4e8fca","url":"https://arxiv.org/abs/2609.02309v1"},{"read_sections":["1 Introduction","2 Preliminaries including 2.1–2.4","6 Planner-Side and System Efficiency","7 Open Challenges","Appendix B Literature Expansion Protocol"],"retrieved_at":"2026-10-18T10:09:36Z","sha256":"09857589286579b14aeeee4a41653aa9e3b0f5b70f70e14f35cb2c1df4398a0a","url":"https://arxiv.org/html/2609.02309v1"}],"read_level":"selected_fulltext_sections_read","sections":["1 Introduction","2 Preliminaries including 2.1–2.4","6 Planner-Side and System Efficiency","7 Open Challenges","Appendix B Literature Expansion Protocol"]}],"reference_number":21,"title":"Efficient GUI Agents: A Systems Survey of Observation, Memory, Action, and Runtime Optimization"},{"canonical_id":"release:awesomecua:v2026.10.18","citation_key":"awesomecua2026","citation_url":"https://github.com/reacher-z/Awesome-CUA/releases/tag/v2026.10.18","edition_label":"v2026.10.18","ordered_authors":["AwesomeCUA contributors"],"read_level":"dataset_snapshot_integrity_inspection","readings":[{"chapter":"review design and research assets","pdf_pages":[],"primary_sources":[{"catalog_sha256":"7f71a1683ab013e587a20086c3e357956981f517a24ae8bc8e81791108c66e4f","source_commit":"87166fe8ad9c8e832f7a90ee84b2e93c6e0e40db","url":"https://github.com/reacher-z/Awesome-CUA/releases/tag/v2026.10.18"}],"read_level":"dataset_snapshot_integrity_inspection","sections":["Frozen catalog records, versioned release artifacts, source-record relationships and descriptive counts"]}],"reference_number":22,"title":"AwesomeCUA catalog: v2026.10.18"},{"canonical_id":"arxiv:2410.05243","citation_key":"arxiv_2410_05243","citation_url":"https://arxiv.org/abs/2410.05243v3","edition_label":"arXiv v3; first submitted 2024-10-07","ordered_authors":["Boyu Gou","Ruohan Wang","Boyuan Zheng","Yanan Xie","Cheng Chang","Yiheng Shu","Huan Sun","Yu Su"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:08:19.268422+00:00","sha256":"b166b8e09f604e64067d12ded4689e7582df106988b28f018651835843b04e68","url":"https://arxiv.org/pdf/2410.05243v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Overview","2.2 Data Construction","2.3 Model Design"]}],"reference_number":23,"title":"Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents"},{"canonical_id":"arxiv:2501.12326","citation_key":"arxiv_2501_12326","citation_url":"https://arxiv.org/abs/2501.12326v1","edition_label":"arXiv v1; first submitted 2025-01-21","ordered_authors":["Yujia Qin","Yining Ye","Junjie Fang","Haoming Wang","Shihao Liang","Shizuo Tian","Junda Zhang","Jiahao Li","Yunxin Li","Shijue Huang","Wanjun Zhong","Kuanye Li","Jiale Yang","Yu Miao","Woyu Lin","Longxiang Liu","Xu Jiang","Qianli Ma","Jingyu Li","Xiaojun Xiao","Kai Cai","Chuang Li","Yaowei Zheng","Chaolin Jin","Chen Li","Xiao Zhou","Minchao Wang","Haoli Chen","Zhaojian Li","Haihua Yang","Haifeng Liu","Feng Lin","Tao Peng","Xin Liu","Guang Shi"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,2,13,14,15,16,17,18,19,20,21,22],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:55.790354+00:00","sha256":"999d01bdd86eb39df1c322d7ab2a48f9f10d8f8a6cb2af913591612712b3964e","url":"https://arxiv.org/pdf/2501.12326v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["Front matter and abstract","4.1 Architecture Overview","4.2 Enhancing GUI Perception","4.3 Unified Action Modeling and Grounding","4.4 Infusing System-2 Reasoning","4.5 Learning from Prior Experience in Long-term Memory","4.6 Training"]}],"reference_number":24,"title":"UI-TARS: Pioneering Automated GUI Interaction with Native Agents"},{"canonical_id":"arxiv:2401.01614","citation_key":"arxiv_2401_01614","citation_url":"https://arxiv.org/abs/2401.01614v2","edition_label":"arXiv v2; first submitted 2024-01-03","ordered_authors":["Boyuan Zheng","Boyu Gou","Jihyung Kil","Huan Sun","Yu Su"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,2,3,4,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:00:23.715060+00:00","sha256":"ac281a3c5cdbda428e14605e9d96035fd35af82073ce46117c27c34792eea041","url":"https://arxiv.org/pdf/2401.01614v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Formulation","2.2 Action Generation","2.3 Action Grounding","3.2 Methods","4.3 Analysis"]}],"reference_number":25,"title":"GPT-4V(ision) is a Generalist Web Agent, if Grounded"},{"canonical_id":"arxiv:2401.13919","citation_key":"arxiv_2401_13919","citation_url":"https://arxiv.org/abs/2401.13919v4","edition_label":"arXiv v4; first submitted 2024-01-25","ordered_authors":["Hongliang He","Wenlin Yao","Kaixin Ma","Wenhao Yu","Yong Dai","Hongming Zhang","Zhenzhong Lan","Dong Yu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4],"primary_sources":[{"retrieved_at":"2026-10-18T10:03:33.620761+00:00","sha256":"476ed4a064311f93414f5d5d17d90f10d07d2922bc05e09ee89cea26b617a2cb","url":"https://arxiv.org/pdf/2401.13919v4","version":4}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Browsing Environment","3.2 Interaction Formulation","3.3 Observation Space","3.4 Action Space","4.1 Website Selection"]}],"reference_number":26,"title":"WebVoyager: Building an End-to-End Web Agent with Large Multimodal Models"},{"canonical_id":"arxiv:2312.13771","citation_key":"arxiv_2312_13771","citation_url":"https://arxiv.org/abs/2312.13771v3","edition_label":"arXiv v3; first submitted 2023-12-21","ordered_authors":["Chi Zhang","Zhao Yang","Jiaxuan Liu","Yanda Li","Yucheng Han","Xin Chen","Zebiao Huang","Bin Fu","Gang Yu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:04:11.632637+00:00","sha256":"800ec0437da9183b77fb6fe3db7b4f64e2ff23b88efc887f6809cd00db0af4c8","url":"https://arxiv.org/pdf/2312.13771v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Environment and Action Space","3.2 Exploration","3.3 Deployment"]}],"reference_number":27,"title":"AppAgent: Multimodal Agents as Smartphone Users"},{"canonical_id":"arxiv:2401.16158","citation_key":"arxiv_2401_16158","citation_url":"https://arxiv.org/abs/2401.16158v2","edition_label":"arXiv v2; first submitted 2024-01-29","ordered_authors":["Junyang Wang","Haiyang Xu","Jiabo Ye","Ming Yan","Weizhou Shen","Ji Zhang","Fei Huang","Jitao Sang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,2,4],"primary_sources":[{"retrieved_at":"2026-10-18T10:04:23.217337+00:00","sha256":"6b511d598fa2f9bea3692d3acf028ebfe46b36b83181baab60a0152955ebec0d","url":"https://arxiv.org/pdf/2401.16158v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Visual Perception","2.2 Instruction Execution","3 Experiments"]}],"reference_number":28,"title":"Mobile-Agent: Autonomous Multi-Modal Mobile Device Agent with Visual Perception"},{"canonical_id":"arxiv:2401.10935","citation_key":"arxiv_2401_10935","citation_url":"https://arxiv.org/abs/2401.10935v2","edition_label":"arXiv v2; first submitted 2024-01-17","ordered_authors":["Kanzhi Cheng","Qiushi Sun","Yougang Chu","Fangzhi Xu","Yantao Li","Jianbing Zhang","Zhiyong Wu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:02:26.938817+00:00","sha256":"b62bd796d8e6d45e13b41ed41a55d021a9df3329bf500d14efed129e8a1e04e2","url":"https://arxiv.org/pdf/2401.10935v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.1 GUI Grounding","3.2 Data Construction","3.3 Training Details","Limitations"]},{"chapter":"evaluation","pdf_pages":[4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:07:03.022590+00:00","sha256":"b62bd796d8e6d45e13b41ed41a55d021a9df3329bf500d14efed129e8a1e04e2","url":"https://arxiv.org/pdf/2401.10935v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.2 Training data","4 ScreenSpot","5.1 Grounding metric","5.2.1 MiniWoB experiment"]}],"reference_number":29,"title":"SeeClick: Harnessing GUI Grounding for Advanced Visual GUI Agents"},{"canonical_id":"arxiv:2312.08914","citation_key":"arxiv_2312_08914","citation_url":"https://arxiv.org/abs/2312.08914v3","edition_label":"arXiv v3; first submitted 2023-12-14","ordered_authors":["Wenyi Hong","Weihan Wang","Qingsong Lv","Jiazheng Xu","Wenmeng Yu","Junhui Ji","Yan Wang","Zihan Wang","Yuxuan Zhang","Juanzi Li","Bin Xu","Yuxiao Dong","Ming Ding","Jie Tang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:00:43.171632+00:00","sha256":"d9cb086d826eaee9f919e0cbf142193b77f476b0aef6260b34408abe142fe13b","url":"https://arxiv.org/pdf/2312.08914v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Architecture","2.2 High-Resolution Cross-Module","2.3 Pre-training","2.4 Multi-task Fine-tuning"]}],"reference_number":30,"title":"CogAgent: A Visual Language Model for GUI Agents"},{"canonical_id":"arxiv:2412.16256","citation_key":"arxiv_2412_16256","citation_url":"https://arxiv.org/abs/2412.16256v2","edition_label":"arXiv v2; first submitted 2024-12-20","ordered_authors":["Yuhao Yang","Yue Wang","Dongxu Li","Ziyang Luo","Bei Chen","Chao Huang","Junnan Li"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:49.808316+00:00","sha256":"169bc7694b456530ce8a6f0336b7efc33a32a21a731e1656caf4934e83f59334","url":"https://arxiv.org/pdf/2412.16256v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["Front matter and abstract","2.1 Large-scale Diverse GUI Data Synthesizing","2.1.2 Context-aware Data Extension from Trajectories","6 Limitations"]}],"reference_number":31,"title":"Aria-UI: Visual Grounding for GUI Instructions"},{"canonical_id":"arxiv:2410.23218","citation_key":"arxiv_2410_23218","citation_url":"https://arxiv.org/abs/2410.23218v1","edition_label":"arXiv v1; first submitted 2024-10-30","ordered_authors":["Zhiyong Wu","Zhenyu Wu","Fangzhi Xu","Yian Wang","Qiushi Sun","Chengyou Jia","Kanzhi Cheng","Zichen Ding","Liheng Chen","Paul Pu Liang","Yu Qiao"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:08:56.970087+00:00","sha256":"ed7ba8eb36d50e36b14e480a3cb89ac9f8fe825d28f763acd8210dcb7ade8bd0","url":"https://arxiv.org/pdf/2410.23218v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Task Formulation and Training","3.2 Grounding Data Collection","3.3 Unified Action Space","4.1 Evaluation Details"]}],"reference_number":32,"title":"OS-ATLAS: A Foundation Action Model for Generalist GUI Agents"},{"canonical_id":"arxiv:2411.17465","citation_key":"arxiv_2411_17465","citation_url":"https://arxiv.org/abs/2411.17465v1","edition_label":"arXiv v1; first submitted 2024-11-26","ordered_authors":["Kevin Qinghong Lin","Linjie Li","Difei Gao","Zhengyuan Yang","Shiwei Wu","Zechen Bai","Weixian Lei","Lijuan Wang","Mike Zheng Shou"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:07:06.161559+00:00","sha256":"8bc023be285e71dab992fb5aeffc2bcbcefe155a67b6ce5e6e7e3a3499bb5db5","url":"https://arxiv.org/pdf/2411.17465v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["2.1 UI-Guided Visual Token Selection","2.2 Interleaved VLA Streaming","2.3 GUI Instructional Tuning"]}],"reference_number":33,"title":"ShowUI: One Vision-Language-Action Model for GUI Visual Agent"},{"canonical_id":"arxiv:2402.07939","citation_key":"arxiv_2402_07939","citation_url":"https://arxiv.org/abs/2402.07939v5","edition_label":"arXiv v5; first submitted 2024-02-08","ordered_authors":["Chaoyun Zhang","Liqun Li","Shilin He","Xu Zhang","Bo Qiao","Si Qin","Minghua Ma","Yu Kang","Qingwei Lin","Saravan Rajmohan","Dongmei Zhang","Qi Zhang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,4,5,6,7,14],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:18.193134+00:00","sha256":"55680fbdd9e98455e4b8c93474e3f367f76d86d170b722433ea2e298fd0a8acf","url":"https://arxiv.org/pdf/2402.07939v5","version":5}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Agent Framework","3.4 Action Execution","5 Limitations"]}],"reference_number":34,"title":"UFO: A UI-Focused Agent for Windows OS Interaction"},{"canonical_id":"arxiv:2410.08164","citation_key":"arxiv_2410_08164","citation_url":"https://arxiv.org/abs/2410.08164v1","edition_label":"arXiv v1; first submitted 2024-10-10","ordered_authors":["Saaket Agashe","Jiuzhou Han","Shuyu Gan","Jiachen Yang","Ang Li","Xin Eric Wang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,4,5,6,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:49.111421+00:00","sha256":"ee8c9b38e20b5f747bd78828d96beab0b266eb99baaafe44c3936ff0d89cb92b","url":"https://arxiv.org/pdf/2410.08164v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Experience-Augmented Hierarchical Planning","3.2 Self-Supervised Memory Construction","3.3 Agent-Computer Interface","4.1 Experimental Setup"]}],"reference_number":35,"title":"Agent S: An Open Agentic Framework that Uses Computers Like a Human"},{"canonical_id":"arxiv:2412.04454","citation_key":"arxiv_2412_04454","citation_url":"https://arxiv.org/abs/2412.04454v2","edition_label":"arXiv v2; first submitted 2024-12-05","ordered_authors":["Yiheng Xu","Zekun Wang","Junli Wang","Dunjie Lu","Tianbao Xie","Amrita Saha","Doyen Sahoo","Tao Yu","Caiming Xiong"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:02.407376+00:00","sha256":"d9f0c31bdae82ae2be8ce1921ba5b23b3898cb61f14ad7755ff8290e7d877a6c","url":"https://arxiv.org/pdf/2412.04454v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["2.2 Agent Framework (continued)","2.3 AGUVIS DATA COLLECTION","2.4 Model Architecture","2.5 Training Paradigm","3.2 Offline GUI Agent Evaluation","3.3 Online GUI Agent Evaluation","4.2 Role of Inner Monologue"]}],"reference_number":36,"title":"Aguvis: Unified Pure Vision Agents for Autonomous GUI Interaction"},{"canonical_id":"arxiv:2504.00906","citation_key":"arxiv_2504_00906","citation_url":"https://arxiv.org/abs/2504.00906v1","edition_label":"arXiv v1; first submitted 2025-04-01","ordered_authors":["Saaket Agashe","Kyle Wong","Vincent Tu","Jiachen Yang","Ang Li","Xin Eric Wang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:44.157220+00:00","sha256":"b87310f7b40407c9413d99bbd6cd027bac381cae77ef4d116dc3070012aa3ae5","url":"https://arxiv.org/pdf/2504.00906v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3 Agent S2","3.1 Mixture of Grounding","3.2 Proactive Hierarchical Planning"]}],"reference_number":37,"title":"Agent S2: A Compositional Generalist-Specialist Framework for Computer Use Agents"},{"canonical_id":"arxiv:2504.14603","citation_key":"arxiv_2504_14603","citation_url":"https://arxiv.org/abs/2504.14603v2","edition_label":"arXiv v2; first submitted 2025-04-20","ordered_authors":["Chaoyun Zhang","He Huang","Chiming Ni","Jian Mu","Si Qin","Shilin He","Lu Wang","Fangkai Yang","Pu Zhao","Chao Du","Liqun Li","Yu Kang","Zhao Jiang","Suzhen Zheng","Rujia Wang","Jiaxu Qian","Minghua Ma","Jian-Guang Lou","Qingwei Lin","Saravan Rajmohan","Dongmei Zhang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6,7,8],"primary_sources":[{"retrieved_at":"2026-10-18T10:11:01.554138+00:00","sha256":"ceecfb496791b8b101275520066eb69c3289c7c727a9e69cc844219598acece5","url":"https://arxiv.org/pdf/2504.14603v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.1 UFO2 as a System Substrate","3.2 HostAgent","3.3 AppAgent","3.4 Hybrid Control Detection","3.5 Unified GUI–API Action Orchestrator"]}],"reference_number":38,"title":"UFO2: The Desktop AgentOS"},{"canonical_id":"arxiv:2509.02544","citation_key":"arxiv_2509_02544","citation_url":"https://arxiv.org/abs/2509.02544v2","edition_label":"arXiv v2; first submitted 2025-09-02","ordered_authors":["Haoming Wang","Haoyang Zou","Huatong Song","Jiazhan Feng","Junjie Fang","Junting Lu","Longxiang Liu","Qinyu Luo","Shihao Liang","Shijue Huang","Wanjun Zhong","Yining Ye","Yujia Qin","Yuwen Xiong","Yuxin Song","Zhiyong Wu","Aoyan Li","Bo Li","Chen Dun","Chong Liu","Daoguang Zan","Fuxing Leng","Hanbin Wang","Hao Yu","Haobin Chen","Hongyi Guo","Jing Su","Jingjia Huang","Kai Shen","Kaiyu Shi","Lin Yan","Peiyao Zhao","Pengfei Liu","Qinghao Ye","Renjie Zheng","Shulin Xin","Wayne Xin Zhao","Wen Heng","Wenhao Huang","Wenqian Wang","Xiaobo Qin","Yi Lin","Youbin Wu","Zehui Chen","Zihao Wang","Baoquan Zhong","Xinchun Zhang","Xujing Li","Yuanfan Li","Zhongkai Zhao","Chengquan Jiang","Faming Wu","Haotian Zhou","Jinlin Pang","Li Han","Qi Liu","Qianli Ma","Siyao Liu","Songhua Cai","Wenqi Fu","Xin Liu","Yaohui Wang","Zhi Zhang","Bo Zhou","Guoliang Li","Jiajun Shi","Jiale Yang","Jie Tang","Li Li","Qihua Han","Taoran Lu","Woyu Lin","Xiaokang Tong","Xinyao Li","Yichi Zhang","Yu Miao","Zhengxuan Jiang","Zili Li","Ziyuan Zhao","Chenxin Li","Dehua Ma","Feng Lin","Ge Zhang","Haihua Yang","Hangyu Guo","Hongda Zhu","Jiaheng Liu","Junda Du","Kai Cai","Kuanye Li","Lichen Yuan","Meilan Han","Minchao Wang","Shuyue Guo","Tianhao Cheng","Xiaobo Ma","Xiaojun Xiao","Xiaolong Huang","Xinjie Chen","Yidi Du","Yilin Chen","Yiwen Wang","Zhaojian Li","Zhenzhu Yang","Zhiyuan Zeng","Chaolin Jin","Chen Li","Hao Chen","Haoli Chen","Jian Chen","Qinghao Zhao","Guang Shi"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,2,4,5,7,9,10,11,12,13],"primary_sources":[{"retrieved_at":"2026-10-18T10:12:03.350197+00:00","sha256":"6a2df5206955e0eab044e1bc4e0ed9239439e0377f79d179044da41a0bd5c603","url":"https://arxiv.org/pdf/2509.02544v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Formulation","2.2 All-in-One GUI Sandbox","2.3 Data Flywheel","2.4.2 Interactive Annotation","2.5 Multi-turn Reinforcement Learning","2.6 Parameter Interpolation"]}],"reference_number":39,"title":"UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning"},{"canonical_id":"arxiv:2609.30186","citation_key":"arxiv_2609_30186","citation_url":"https://arxiv.org/abs/2609.30186v1","edition_label":"arXiv v1; first submitted 2026-09-24","ordered_authors":["Linghua Zhang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:49:32.541714+00:00","sha256":"ddc55d413582d0a17eec43bb37ffaeb348c4228b434694cc653853de2b32d33a","url":"https://arxiv.org/html/2609.30186v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["2 Problem Setup","3.1 Delegation State","3.2 Current Accessibility-Tree Candidates","3.3 Typed Local Decision","3.4 Budgets and Boundary","5 Experimental Design","8 Limitations"]}],"reference_number":40,"title":"Jev-Mobile: Jev as an Executor for Mobile GUI Agents"},{"canonical_id":"arxiv:2409.07429","citation_key":"arxiv_2409_07429","citation_url":"https://arxiv.org/abs/2409.07429v1","edition_label":"arXiv v1; first submitted 2024-09-11","ordered_authors":["Zora Zhiruo Wang","Jiayuan Mao","Daniel Fried","Graham Neubig"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:33.155315+00:00","sha256":"ddc6f6fedec48c88ae743b43a464f7eb4aecaab05acd18c29cfb657ab445ebe4","url":"https://arxiv.org/pdf/2409.07429v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["2.2 Workflow Representation","2.3 Inducing and Using Workflows","3.1 WebArena"]}],"reference_number":41,"title":"Agent Workflow Memory"},{"canonical_id":"arxiv:2605.18652","citation_key":"arxiv_2605_18652","citation_url":"https://arxiv.org/abs/2605.18652v1","edition_label":"arXiv v1; first submitted 2026-05-18","ordered_authors":["Ziyun Zeng","Hang Hua","Bocheng Zou","Mu Cai","Rogerio Feris","Jiebo Luo"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:13:01.881533+00:00","sha256":"bf6e672269ac55c35870ab2b93d652b472c30fd6ad9c98797b00127d58132305","url":"https://arxiv.org/pdf/2605.18652v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3 Data Curation","4.1 Framework","4.2 Memory System Design","4.3 Training MementoCore","4.4 Benchmarking","5.1 Experimental Setup"]}],"reference_number":42,"title":"MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI Agents"},{"canonical_id":"arxiv:2510.06042","citation_key":"arxiv_2510_06042","citation_url":"https://arxiv.org/abs/2510.06042v3","edition_label":"arXiv v3; first submitted 2025-10-07","ordered_authors":["Shang Ma","Xusheng Xiao","Yanfang Ye"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:14:29.861666+00:00","sha256":"5e2cb2f053983140904ec15c097d060ff18315c6c3c956bbcacbec26921709c5","url":"https://arxiv.org/pdf/2510.06042v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["3.3 UI Automation to Classical Planning","4.1 UTG Builder","4.2 Node Selector","4.3 Plan Generator","4.4 UI Explorer","5.1 Datasets"]}],"reference_number":43,"title":"Agent+P: Guiding UI Agents via Symbolic Planning"},{"canonical_id":"arxiv:2602.20502","citation_key":"arxiv_2602_20502","citation_url":"https://arxiv.org/abs/2602.20502v2","edition_label":"arXiv v2; first submitted 2026-02-24","ordered_authors":["Hongbin Zhong","Fazle Faisal","Luis França","Tanakorn Leesatapornwongsa","Adriana Szekeres","Kexin Rong","Suman Nath"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:12:47.197774+00:00","sha256":"ec8339a07ee3c83d4f416a27cf9e8bf254efa26968b7bc9fdf3e07ff09e38ab6","url":"https://arxiv.org/pdf/2602.20502v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3 State-Machine Memory","3.1 Application State Representation","3.2 Automated SMG Construction","4.1 Planner","4.2 Executor","4.3 Patcher (opening)"]}],"reference_number":44,"title":"ActionEngine: From Reactive to Programmatic Web Agents via State Machine Memory"},{"canonical_id":"arxiv:2412.09605","citation_key":"arxiv_2412_09605","citation_url":"https://arxiv.org/abs/2412.09605v2","edition_label":"arXiv v2; first submitted 2024-12-12","ordered_authors":["Yiheng Xu","Dunjie Lu","Zhennan Shen","Junli Wang","Zekun Wang","Yuchen Mao","Caiming Xiong","Tao Yu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:11:32.110232+00:00","sha256":"7bdb3828c245ef1ad1cf60ead2140e1bd5f6ff01f8a189f50914cb07f49ef603","url":"https://arxiv.org/pdf/2412.09605v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Tutorial Collection and Standardization","2.2.1 Trajectory Data Schema","2.2.2 Guided Replay with Tutorials","2.2.3 Evaluation of Trajectory"]}],"reference_number":45,"title":"AgentTrek: Agent Trajectory Synthesis via Guiding Replay with Web Tutorials"},{"canonical_id":"arxiv:2508.09123","citation_key":"arxiv_2508_09123","citation_url":"https://arxiv.org/abs/2508.09123v3","edition_label":"arXiv v3; first submitted 2025-08-12","ordered_authors":["Xinyuan Wang","Bowen Wang","Dunjie Lu","Junlin Yang","Tianbao Xie","Junli Wang","Jiaqi Deng","Xiaole Guo","Yiheng Xu","Chen Henry Wu","Zhennan Shen","Zhuokai Li","Ryan Li","Xiaochuan Li","Junda Chen","Boyuan Zheng","Peihang Li","Fangyu Lei","Ruisheng Cao","Yeqiao Fu","Dongchan Shin","Martin Shin","Jiarui Hu","Yuyan Wang","Jixuan Chen","Yuxiao Ye","Danyang Zhang","Dikang Du","Hao Hu","Huarong Chen","Zaida Zhou","Haotian Yao","Ziwei Chen","Qizheng Gu","Yipu Wang","Heng Wang","Diyi Yang","Victor Zhong","Flood Sung","Y. Charles","Zhilin Yang","Tao Yu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:11:08.501504+00:00","sha256":"fd1c1e7b4d467acf63fc743c3925c7ceb015e5b1625f62c22295c0b9e2d82a70","url":"https://arxiv.org/pdf/2508.09123v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["2 AgentNet Collection","2.2 AgentNet Tool","3.1 Synthesizing Reflective Long CoT Reasoning","3.2 Context Encoding and Test-Time Reasoning","3.3 Training Data Mixtures"]},{"chapter":"evaluation","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:31.930715+00:00","sha256":"bd90167d436738c1e50b6ab2fd89cac3d427ff59c31be9d731f72453ee3a6937","url":"https://arxiv.org/html/2508.09123v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["HTML §2.1–2.3 AgentNet Collection","HTML Table 2 footnote","HTML §3.1 Reflective CoT synthesis","HTML Appendix C.1–C.2"]}],"reference_number":46,"title":"OpenCUA: Open Foundations for Computer-Use Agents"},{"canonical_id":"arxiv:2502.11357","citation_key":"arxiv_2502_11357","citation_url":"https://arxiv.org/abs/2502.11357v4","edition_label":"arXiv v4; first submitted 2025-02-17","ordered_authors":["Vardaan Pahuja","Yadong Lu","Corby Rosset","Boyu Gou","Arindam Mitra","Spencer Whitehead","Yu Su","Ahmed Awadallah"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4],"primary_sources":[{"retrieved_at":"2026-10-18T10:11:37.955753+00:00","sha256":"038a14d81f9587ac869667d77f93ce33eac65d18e8037402680c091204e8cac8","url":"https://arxiv.org/pdf/2502.11357v4","version":4}],"read_level":"selected_fulltext_sections_read","sections":["3 Data Recipe","3.1 Website Selection","3.2 Data Generation Pipeline"]}],"reference_number":47,"title":"Explorer: Scaling Exploration-driven Web Trajectory Synthesis for Multimodal Web Agents"},{"canonical_id":"arxiv:2412.19723","citation_key":"arxiv_2412_19723","citation_url":"https://arxiv.org/abs/2412.19723v3","edition_label":"arXiv v3; first submitted 2024-12-27","ordered_authors":["Qiushi Sun","Kanzhi Cheng","Zichen Ding","Chuanyang Jin","Yian Wang","Fangzhi Xu","Zhenyu Wu","Chengyou Jia","Liheng Chen","Zhoumianze Liu","Ben Kao","Guohao Li","Junxian He","Yu Qiao","Zhiyong Wu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:11:42.264735+00:00","sha256":"65bb449a7e9d433f29b088e0ed5d52f22833572b7a67533b48cf07e88f14292d","url":"https://arxiv.org/pdf/2412.19723v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Interaction-Driven Functional Discovery","3.2 Reverse Task Synthesis","3.3 Trajectory Reward Model","4.1 Experimental Settings"]}],"reference_number":48,"title":"OS-Genesis: Automating GUI Agent Trajectory Construction via Reverse Task Synthesis"},{"canonical_id":"arxiv:2504.10458","citation_key":"arxiv_2504_10458","citation_url":"https://arxiv.org/abs/2504.10458v4","edition_label":"arXiv v4; first submitted 2025-04-14","ordered_authors":["Run Luo","Lu Wang","Wanwei He","Longze Chen","Jiaming Li","Xiaobo Xia"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:11:59.935447+00:00","sha256":"6e9d838b0c87ac3e21e266748924be44869abe8f1e818b7e1bb066f746d9bc9f","url":"https://arxiv.org/pdf/2504.10458v4","version":4}],"read_level":"selected_fulltext_sections_read","sections":["3 GUI-R1 Framework","3.1 Preliminaries","3.2 Verifiable Rewards in Unified Action Space","3.3 Training Data Curation"]}],"reference_number":49,"title":"GUI-R1 : A Generalist R1-Style Vision-Language Action Model For GUI Agents"},{"canonical_id":"arxiv:2506.07976","citation_key":"arxiv_2506_07976","citation_url":"https://arxiv.org/abs/2506.07976v2","edition_label":"arXiv v2; first submitted 2025-06-09","ordered_authors":["Junhong Shen","Hao Bai","Lunjun Zhang","Yifei Zhou","Amrith Setlur","Shengbang Tong","Diego Caples","Nan Jiang","Tong Zhang","Ameet Talwalkar","Aviral Kumar"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,7,8],"primary_sources":[{"retrieved_at":"2026-10-18T10:12:22.820764+00:00","sha256":"be76d2989976e3d59e20189821c3dfccc7b6008e2e29f1a2741b567fd5b01c65","url":"https://arxiv.org/pdf/2506.07976v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3 Problem Setup","4.1 Scaling Interaction by Acting Longer","4.2 Interaction vs Per-Step Compute","5.1 Fixed Long Horizons","5.2 Curriculum Over Interaction Horizon"]}],"reference_number":50,"title":"Thinking vs. Doing: Agents that Reason by Scaling Test-Time Interaction"},{"canonical_id":"arxiv:2411.02337","citation_key":"arxiv_2411_02337","citation_url":"https://arxiv.org/abs/2411.02337v3","edition_label":"arXiv v3; first submitted 2024-11-04","ordered_authors":["Zehan Qi","Xiao Liu","Iat Long Iong","Hanyu Lai","Xueqiao Sun","Wenyi Zhao","Yu Yang","Xinyue Yang","Jiadai Sun","Shuntian Yao","Tianjie Zhang","Wei Xu","Jie Tang","Yuxiao Dong"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:11:47.855035+00:00","sha256":"79a93b02cca5586d9e3e2742c5af53f683c632dc0eaed3b7bf87f53daaa2d462","url":"https://arxiv.org/pdf/2411.02337v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["2 Self-Evolving Online Curriculum RL","2.1 Self-evolving New Instruction","2.2 Reinforcement Learning for LLMs in Online Web Environments"]}],"reference_number":51,"title":"WebRL: Training LLM Web Agents via Self-Evolving Online Curriculum Reinforcement Learning"},{"canonical_id":"arxiv:2606.05597","citation_key":"arxiv_2606_05597","citation_url":"https://arxiv.org/abs/2606.05597v3","edition_label":"arXiv v3; first submitted 2026-06-04","ordered_authors":["Hao Bai","Rui Yang","Chenlu Ye","Spencer Whitehead","Aviral Kumar","Tong Zhang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:12:58.265627+00:00","sha256":"423c18c48b337d91ecd7fe548cd67a910ccf4cc373a4b400674e9f0b2bc4cba4","url":"https://arxiv.org/pdf/2606.05597v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["3.1 System","3.2 Algorithm","4.1 Setup","5.2 Ablations"]}],"reference_number":52,"title":"AsyncWebRL: Efficient Asynchronous Reinforcement Learning for Multi-Step Visual Web Agents"},{"canonical_id":"arxiv:2607.28227","citation_key":"arxiv_2607_28227","citation_url":"https://arxiv.org/abs/2607.28227v1","edition_label":"arXiv v1; first submitted 2026-07-30","ordered_authors":["Hanzhang Zhou","Panrong Tong","Xu Zhang","Quyu Kong","Chenglin Cai","Tianyu Xia","Gongjie Zhang","Jianan Zhang","Long Li","Long Chen","Lei Wang","Gaole Dai","Pengxiang Li","Liangyu Chen","Yue Wang","Steven Hoi"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,2,3,4,6,9,10,13,14,15,44],"primary_sources":[{"retrieved_at":"2026-10-18T10:13:06.605062+00:00","sha256":"705de6c54fa4eec0e5f81792cd323695e1d0274333886973cbad82d3c436fdb0","url":"https://arxiv.org/pdf/2607.28227v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["2.1 System Overview and Action Space","2.2.2 Real-Device Runtime","2.2.3 Hybrid GUI+CLI","2.4.1 SFT","2.4.2 Action RL","2.4.3 Online RL","7 Limitations"]}],"reference_number":53,"title":"Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents"},{"canonical_id":"arxiv:2605.17320","citation_key":"arxiv_2605_17320","citation_url":"https://arxiv.org/abs/2605.17320v1","edition_label":"arXiv v1; first submitted 2026-05-17","ordered_authors":["Yutong Huang","Vikranth Srivatsa","Alex Asch","Hansin Tushar Patwa","Yiying Zhang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,4,5,8],"primary_sources":[{"retrieved_at":"2026-10-18T10:12:53.041131+00:00","sha256":"cd6d543feb3ec804b35f91ae9dd897a563038f9230067abd5eaafd88976e46af","url":"https://arxiv.org/pdf/2605.17320v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3 TClone Overview","4 TClone Design","4.4 Network Subsystem","4.5 GUI Subsystem"]}],"reference_number":54,"title":"TClone: Low-Latency Forking of Live GUI Environments for Computer-Use Agents"},{"canonical_id":"arxiv:2606.20622","citation_key":"arxiv_2606_20622","citation_url":"https://arxiv.org/abs/2606.20622v1","edition_label":"arXiv v1; first submitted 2026-05-26","ordered_authors":["Daniel Beechey","Derek Yuen","Jianheng Liu","Dezhao Luo","Tiantian He","Weilin Luo","Jun Wang","Kun Shao"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6,7,8],"primary_sources":[{"retrieved_at":"2026-10-18T10:14:00.181958+00:00","sha256":"e4711466886372760daa83384cb4d129d54fdcfe320be6e91fb7b4971d1a61de","url":"https://arxiv.org/pdf/2606.20622v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["2 Agent-Environment Interface and Policy Optimisation","3 Scalable Asynchronous Mobile Environment","3.2 Persistent State Management","3.3 Task Lifecycle","4.2 Rollout and Optimisation","4.3 Anatomy of the Self-Evolving Agent"]}],"reference_number":55,"title":"Darwin Mobile Agent: A Roadmap for Self-Evolution"},{"canonical_id":"arxiv:2601.19306","citation_key":"arxiv_2601_19306","citation_url":"https://arxiv.org/abs/2601.19306v1","edition_label":"arXiv v1; first submitted 2026-01-27","ordered_authors":["Sijia Li","Xiaoyu Tan","Shahir Ali","Niels Schmidt","Gengchen Ma","Xihe Qiu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:14:20.444793+00:00","sha256":"ef3bbc8adcc147239ea910e0c59904d7a26ee1610a7a4d1613be2cb845098536","url":"https://arxiv.org/pdf/2601.19306v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Problem Formulation","3.2 Curiosity-Driven Information Gain Estimation","3.3 Retrieving and Embedding External Knowledge"]}],"reference_number":56,"title":"Curiosity Driven Knowledge Retrieval for Mobile Agents"},{"canonical_id":"arxiv:2605.05765","citation_key":"arxiv_2605_05765","citation_url":"https://arxiv.org/abs/2605.05765v2","edition_label":"arXiv v2; first submitted 2026-05-07","ordered_authors":["Xiaoming Ren","Ru Zhen","Chao Li","Yang Song","Qiuxia Hou","Yanhao Zhang","Peng Liu","Qi Qi","Quanlong Zheng","Qi Wu","Zhenyi Liao","Binqiang Pan","Haobo Ji","Haonan Lu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4,5,6,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:14:10.101459+00:00","sha256":"80375cd810dcda2e9ccca2c6fbee44730de8449972036c548bde312566b20673","url":"https://arxiv.org/pdf/2605.05765v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["2 Framework","3 Omni Perception","4 Omni Memory","5.1 Omni Action in App Ecosystem","5.2 Trajectory Cloned Execution"]}],"reference_number":57,"title":"X-OmniClaw Technical Report: A Unified Mobile Agent for Multimodal Understanding and Interaction"},{"canonical_id":"arxiv:2508.18040","citation_key":"arxiv_2508_18040","citation_url":"https://arxiv.org/abs/2508.18040v1","edition_label":"arXiv v1; first submitted 2025-08-25","ordered_authors":["Xin Wang","Zhiyao Cui","Hao Li","Ya Zeng","Chenxu Wang","Ruiqi Song","Yihang Chen","Kun Shao","Qiaosheng Zhang","Jinzhuo Liu","Siyue Ren","Shuyue Hu","Zhen Wang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,3,4],"primary_sources":[{"retrieved_at":"2026-10-18T10:18:36.598801+00:00","sha256":"3c3474b771af7bdd500e6aaa91b83d39ec0d455e59bb83df906e7a94700ad2b7","url":"https://arxiv.org/pdf/2508.18040v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["The Framework of PerPilot","Personalization Perception","Personalization Completion","PerInstruct Data Construction"]}],"reference_number":58,"title":"PerPilot: Personalizing VLM-based Mobile Agents via Memory and Exploration"},{"canonical_id":"arxiv:2509.02444","citation_key":"arxiv_2509_02444","citation_url":"https://arxiv.org/abs/2509.02444v2","edition_label":"arXiv v2; first submitted 2025-09-02","ordered_authors":["Jingru Fan","Yufan Dang","Jingyao Wu","Huatao Li","Runde Yang","Xiyuan Yang","Yuheng Wang","Chen Qian"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"methods","pdf_pages":[1,5,6,34,37,38,42,48,49,50],"primary_sources":[{"retrieved_at":"2026-10-18T10:14:33.619660+00:00","sha256":"6c2591dad8893f87cffdea211584a3733dc2137dafa9416c1885d1b0126fd540","url":"https://arxiv.org/pdf/2509.02444v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["4.2 OCR/OR Region Localization","4.2.2 Coordinate Calibration","4.3 Multi-Agent Voting","5.2 Task Decomposition","6.2 Experience Reuse","6.3 Function Calling"]}],"reference_number":59,"title":"AppCopilot: Toward General, Accurate, Long-Horizon, and Efficient Mobile Agent"},{"canonical_id":"arxiv:2412.05467","citation_key":"arxiv_2412_05467","citation_url":"https://arxiv.org/abs/2412.05467v4","edition_label":"arXiv v4; first submitted 2024-12-06","ordered_authors":["Thibault Le Sellier De Chezelles","Maxime Gasse","Alexandre Drouin","Massimo Caccia","Léo Boisvert","Megh Thakkar","Tom Marty","Rim Assouel","Sahar Omidi Shayegan","Lawrence Keunho Jang","Xing Han Lù","Ori Yoran","Dehan Kong","Frank F. Xu","Siva Reddy","Quentin Cappart","Graham Neubig","Ruslan Salakhutdinov","Nicolas Chapados","Alexandre Lacoste"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[5,7,8,10,11,17,26],"primary_sources":[{"retrieved_at":"2026-10-18T10:04:23.544564+00:00","sha256":"98d4d9b8088ca6bc5e0efd613389622b53ffedcb36cb285f093f7e6045499491","url":"https://arxiv.org/pdf/2412.05467v4","version":4}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Observation space","3.2 Action space","3.3 Task interface","4 Benchmark unification","5.1–5.4 Experiment management/reproducibility","7.1 Limitations","Appendix B"]}],"reference_number":60,"title":"The BrowserGym Ecosystem for Web Agent Research"},{"canonical_id":"arxiv:2401.13649","citation_key":"arxiv_2401_13649","citation_url":"https://arxiv.org/abs/2401.13649v2","edition_label":"arXiv v2; first submitted 2024-01-24","ordered_authors":["Jing Yu Koh","Robert Lo","Lawrence Jang","Vikram Duvvur","Ming Chong Lim","Po-Yu Huang","Graham Neubig","Shuyan Zhou","Ruslan Salakhutdinov","Daniel Fried"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:02:25.134588+00:00","sha256":"8674b4ae17290bb797e5988706f765c75a3310a4e9d14cbd32497e40235ff795","url":"https://arxiv.org/pdf/2401.13649v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3 Environment","3.1 Observation Space","3.2 Action Space","3.3 Evaluation","4.1 Web Environments","4.2 Tasks"]}],"reference_number":61,"title":"VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks"},{"canonical_id":"arxiv:2403.07718","citation_key":"arxiv_2403_07718","citation_url":"https://arxiv.org/abs/2403.07718v5","edition_label":"arXiv v5; first submitted 2024-03-12","ordered_authors":["Alexandre Drouin","Maxime Gasse","Massimo Caccia","Issam H. Laradji","Manuel Del Verme","Tom Marty","Léo Boisvert","Megh Thakkar","Quentin Cappart","David Vazquez","Nicolas Chapados","Alexandre Lacoste"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:03:18.698795+00:00","sha256":"686792a9085d998be3ac3de3e9c126fd0a9256b6ccb313aa006edb62f7645295","url":"https://arxiv.org/pdf/2403.07718v5","version":5}],"read_level":"selected_fulltext_sections_read","sections":["3 WorkArena","3.1 WorkArena Tasks","3.2 Challenges","3.3 Availability","4.1 Capabilities","4.2 An Ideal Experimental Framework","5 Experiments"]}],"reference_number":62,"title":"WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?"},{"canonical_id":"arxiv:2409.08264","citation_key":"arxiv_2409_08264","citation_url":"https://arxiv.org/abs/2409.08264v2","edition_label":"arXiv v2; first submitted 2024-09-12","ordered_authors":["Rogerio Bonatti","Dan Zhao","Francesco Bonacci","Dillon Dupont","Sara Abdali","Yinheng Li","Yadong Lu","Justin Wagle","Kazuhito Koishida","Arthur Bucker","Lawrence Jang","Zack Hui"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[4,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:14.005003+00:00","sha256":"fbd0821e72620273a47c4c5366bb4283cd88ed2d909ca77c96b58839e95e9b7e","url":"https://arxiv.org/pdf/2409.08264v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Task definition and reward","3.2 Task curation","3.3 Deployment"]}],"reference_number":63,"title":"Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale"},{"canonical_id":"arxiv:2407.05291","citation_key":"arxiv_2407_05291","citation_url":"https://arxiv.org/abs/2407.05291v2","edition_label":"arXiv v2; first submitted 2024-07-07","ordered_authors":["Léo Boisvert","Megh Thakkar","Maxime Gasse","Massimo Caccia","Thibault Le Sellier De Chezelles","Quentin Cappart","Nicolas Chapados","Alexandre Lacoste","Alexandre Drouin"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[4,6,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:04:05.429994+00:00","sha256":"9a3f424bfcdb2e5b2098c9fa00269a4dac371c11dada4bd4eec436bb2b61d479","url":"https://arxiv.org/pdf/2407.05291v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Difficulty levels","3.3 Salient features","4.1 Evaluation curriculum","4.2 Agent design"]}],"reference_number":64,"title":"WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks"},{"canonical_id":"arxiv:2407.15711","citation_key":"arxiv_2407_15711","citation_url":"https://arxiv.org/abs/2407.15711v2","edition_label":"arXiv v2; first submitted 2024-07-22","ordered_authors":["Ori Yoran","Samuel Joseph Amouyal","Chaitanya Malaviya","Ben Bogin","Ofir Press","Jonathan Berant"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,8,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:04:34.746850+00:00","sha256":"f85974c1441ac70de53675aa3c0bf3f2d719b963d296c1bc28733f203fcc027d","url":"https://arxiv.org/pdf/2407.15711v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Task criteria","3.2 Data collection","3.3 Statistics","3.4 Automatic evaluation","6.4 Time dependency","9 Limitations"]}],"reference_number":65,"title":"AssistantBench: Can Web Agents Solve Realistic and Time-Consuming Tasks?"},{"canonical_id":"arxiv:2609.24890","citation_key":"arxiv_2609_24890","citation_url":"https://arxiv.org/abs/2609.24890v1","edition_label":"arXiv v1; first submitted 2026-09-21","ordered_authors":["Zhilin Wang","Shaokun Zhang","Yifan Zhang","Hao Zhang","Jin Xu","Binfeng Xu","Jian Hu","Yunheng Zou","Karan Sapra","Andrew Tao","Jan Kautz","Yi Dong"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,5,11],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:35.635503+00:00","sha256":"71ebfedd965911d53840bccc05c8dedcf8812268529adba07ee68f8e3195d9d2","url":"https://arxiv.org/pdf/2609.24890v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["2 Benchmark overview","3 Data collection","4.1 Judge evaluation","Reproducibility statement"]}],"reference_number":66,"title":"OSWorld-Pro: Process-based Evaluation for Computer Use Agents"},{"canonical_id":"arxiv:2406.03679","citation_key":"arxiv_2406_03679","citation_url":"https://arxiv.org/abs/2406.03679v6","edition_label":"arXiv v6; first submitted 2024-06-06","ordered_authors":["Wei Li","William Bishop","Alice Li","Chris Rawles","Folawiyo Campbell-Ajala","Divya Tyamagundlu","Oriana Riva"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,5,6,7,8,10],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:46.442693+00:00","sha256":"d4a14ffbe97638159a2003a9b63eff4de21512c9799edcf7754d8705d757c4cf","url":"https://arxiv.org/pdf/2406.03679v6","version":6}],"read_level":"selected_fulltext_sections_read","sections":["Table 1","3.2 Statistics","3.3 Splits/Table 3","4.1 Agent implementation","4.2 Metrics","4.4 Scaling extrapolation","5 Limitations"]}],"reference_number":67,"title":"On the Effects of Data Scale on UI Control Agents"},{"canonical_id":"arxiv:2406.08451","citation_key":"arxiv_2406_08451","citation_url":"https://arxiv.org/abs/2406.08451v2","edition_label":"arXiv v2; first submitted 2024-06-12","ordered_authors":["Quanfeng Lu","Wenqi Shao","Zitao Liu","Lingxiao Du","Fanqing Meng","Boxuan Li","Botong Chen","Siyuan Huang","Kaipeng Zhang","Ping Luo"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,5,6,7,13],"primary_sources":[{"retrieved_at":"2026-10-18T10:06:29.054200+00:00","sha256":"3f95681c58642c8b899c3951f940e350b70fb50ece5004520fcd7794759aecb2","url":"https://arxiv.org/pdf/2406.08451v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Metadata","3.2 Collection","3.3 Statistics","5.1 Metrics","5.2 Experimental setup","Appendix 9.1 Splits"]}],"reference_number":68,"title":"GUIOdyssey: A Comprehensive Dataset for Cross-App GUI Navigation on Mobile Devices"},{"canonical_id":"arxiv:2504.07981","citation_key":"arxiv_2504_07981","citation_url":"https://arxiv.org/abs/2504.07981v1","edition_label":"arXiv v1; first submitted 2025-04-04","ordered_authors":["Kaixin Li","Ziyang Meng","Hongzhan Lin","Ziyang Luo","Yuchen Tian","Jing Ma","Zhiyong Huang","Tat-Seng Chua"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[2,4,5,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:07:31.910360+00:00","sha256":"4eec6bbf0bba837c66a0d69551cc1567200d7aea85c6ce72650706de637c10c0","url":"https://arxiv.org/pdf/2504.07981v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["Introduction benchmark summary","3.1–3.4 Collection and statistics","Limitations"]}],"reference_number":69,"title":"ScreenSpot-Pro: GUI Grounding for Professional High-Resolution Computer Use"},{"canonical_id":"arxiv:2506.14866","citation_key":"arxiv_2506_14866","citation_url":"https://arxiv.org/abs/2506.14866v2","edition_label":"arXiv v2; first submitted 2025-06-17","ordered_authors":["Thomas Kuntz","Agatha Duzan","Hao Zhao","Francesco Croce","Zico Kolter","Nicolas Flammarion","Maksym Andriushchenko"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[4,5,6,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:08:55.941603+00:00","sha256":"5a1133e729d882a7bc0a19fe3a08889072ecfa3789e7c5c2773aa3c6e18bdb6a","url":"https://arxiv.org/pdf/2506.14866v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Setup","3.2 Three harm categories","3.3 LLM judge","5 Discussion"]}],"reference_number":70,"title":"OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents"},{"canonical_id":"arxiv:2608.07775","citation_key":"arxiv_2608_07775","citation_url":"https://arxiv.org/abs/2608.07775v1","edition_label":"arXiv v1; first submitted 2026-08-07","ordered_authors":["Xiaoou Liu","Longchao Da","Hanyang Chen","Yuan Ling","Hua Wei"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[4,5,7,8],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:44.898646+00:00","sha256":"d4afe27efcce2a3b9401f9677304db3cdedfa66f537293bb742bd1abfb16ee59","url":"https://arxiv.org/pdf/2608.07775v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Perturbation formulation","3.2 Injection","3.3 Evaluation protocol","4.3 Error analysis"]}],"reference_number":71,"title":"AndroidReality: How Far Are Mobile Agents from the Real World?"},{"canonical_id":"arxiv:2610.06898","citation_key":"arxiv_2610_06898","citation_url":"https://arxiv.org/abs/2610.06898v1","edition_label":"arXiv v1; first submitted 2026-09-29","ordered_authors":["Li Hu","Kanghua Mo","Yingbin Jin","Qingqing Ye","Haibo Hu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,5,6,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:35.098754+00:00","sha256":"73f83c27e58e46086dbd5f32f108df2c5faa9ab2bd968d48062f39d94b2cc14b","url":"https://arxiv.org/pdf/2610.06898v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Benchmark components","3.2 Instance generation","3.3 Paired protocol/metrics","5 Limitations/Ethics"]}],"reference_number":72,"title":"DIBench: Benchmarking Decision Integrity of GUI-based Mobile Agents Under Deceptive Injections"},{"canonical_id":"arxiv:2506.16042","citation_key":"arxiv_2506_16042","citation_url":"https://arxiv.org/abs/2506.16042v2","edition_label":"arXiv v2; first submitted 2025-06-19","ordered_authors":["Reyna Abhyankar","Qi Qi","Yiying Zhang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[9,10],"primary_sources":[{"retrieved_at":"2026-10-18T10:08:35.906616+00:00","sha256":"ee48a997c1d55ba387cbe6a4d35297cf414675ff6f18b26dcac94fcd97abfcfc","url":"https://arxiv.org/pdf/2506.16042v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["4.1 Human trajectory construction","4.2 Action grouping","5.1 Weighted Efficiency Score"]}],"reference_number":73,"title":"OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agents"},{"canonical_id":"arxiv:2610.01491","citation_key":"arxiv_2610_01491","citation_url":"https://arxiv.org/abs/2610.01491v1","edition_label":"arXiv v1; first submitted 2026-10-01","ordered_authors":["Chengguang Gan","Zimeng He","Yoshihiro Tsujii","Ken-ichiro Kobayashi","Hiroki Itoh","Kotaro Funakoshi"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,5,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:38.289366+00:00","sha256":"a6eb4d550204825582dd0da428692968a443de99a6b8ff43fa5626d0c20de04f","url":"https://arxiv.org/pdf/2610.01491v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Study scope","4.1–4.3 Human verification","6 Limitations"]}],"reference_number":74,"title":"Auditing Web Agent Evaluation on WebArena-Lite: Human Review of Outcomes and Trajectories"},{"canonical_id":"arxiv:2510.18488","citation_key":"arxiv_2510_18488","citation_url":"https://arxiv.org/abs/2510.18488v1","edition_label":"arXiv v1; first submitted 2025-10-21","ordered_authors":["Ho Fai Leung","Xiaoyan Xi","Fei Zuo"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,5],"primary_sources":[{"retrieved_at":"2026-10-18T10:08:39.351780+00:00","sha256":"2b8d59200d47d1b3347f38ba6aa21c1f93a4ac724081178117a5bb9873a84797","url":"https://arxiv.org/pdf/2510.18488v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["2.1 Benchmark purification","3 Evaluation setup/metrics"]}],"reference_number":75,"title":"AndroidControl-Curated: Revealing the True Potential of GUI Agents through Benchmark Purification"},{"canonical_id":"arxiv:2410.17520","citation_key":"arxiv_2410_17520","citation_url":"https://arxiv.org/abs/2410.17520v3","edition_label":"arXiv v3; first submitted 2024-10-23","ordered_authors":["Juyong Lee","Dongyoon Hahm","June Suk Choi","W. Bradley Knox","Kimin Lee"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,8],"primary_sources":[{"retrieved_at":"2026-10-18T10:08:53.257108+00:00","sha256":"8f407be8a7e95dcedb68f82d6b4fe2e4da2a1d692f0e8d048a18853ab715ceb3","url":"https://arxiv.org/pdf/2410.17520v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["3.1–3.4 Benchmark design and evaluation","Limitations"]}],"reference_number":76,"title":"MobileSafetyBench: Evaluating Safety of Autonomous Agents in Mobile Device Control"},{"canonical_id":"arxiv:2504.18575","citation_key":"arxiv_2504_18575","citation_url":"https://arxiv.org/abs/2504.18575v3","edition_label":"arXiv v3; first submitted 2025-04-22","ordered_authors":["Ivan Evtimov","Arman Zharmagambetov","Aaron Grattafiori","Chuan Guo","Kamalika Chaudhuri"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,5,9],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:12.692007+00:00","sha256":"460fd2adff8d31a62f7743fc7f30e63f83ac3ed1c0eca2bd957fd2d5ba0b4290","url":"https://arxiv.org/pdf/2504.18575v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["2 Threat model","3.1–3.2 Benchmark construction","3.4 Main metrics","5 Limitations"]}],"reference_number":77,"title":"WASP: Benchmarking Web Agent Security Against Prompt Injection Attacks"},{"canonical_id":"arxiv:2406.12814","citation_key":"arxiv_2406_12814","citation_url":"https://arxiv.org/abs/2406.12814v3","edition_label":"arXiv v3; first submitted 2024-06-18","ordered_authors":["Chen Henry Wu","Rishi Shah","Jing Yu Koh","Ruslan Salakhutdinov","Daniel Fried","Aditi Raghunathan"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,5,6],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:27.478864+00:00","sha256":"b2956d9036dba351b756cbc5986a89b8180b6284ee91bcaa989ca6f274f869ff","url":"https://arxiv.org/pdf/2406.12814v3","version":3}],"read_level":"selected_fulltext_sections_read","sections":["3.1 Threat model","3.2 Agent graph","4.1 VWA-Adv curation","4.2 Attacker access"]}],"reference_number":78,"title":"Dissecting Adversarial Robustness of Multimodal LM Agents"},{"canonical_id":"arxiv:2505.21936","citation_key":"arxiv_2505_21936","citation_url":"https://arxiv.org/abs/2505.21936v5","edition_label":"arXiv v5; first submitted 2025-05-28","ordered_authors":["Zeyi Liao","Jaylen Jones","Linxi Jiang","Yuting Ning","Eric Fosler-Lussier","Yu Su","Zhiqiang Lin","Huan Sun"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[4,5,6,7,10,18],"primary_sources":[{"retrieved_at":"2026-10-18T10:09:08.748979+00:00","sha256":"f0622cc48b654aeeb2f2872162d7df21b3a32ee8475156b254544260e696b3b2","url":"https://arxiv.org/pdf/2505.21936v5","version":5}],"read_level":"selected_fulltext_sections_read","sections":["3 Hybrid sandbox","4.2 Threat model","5.1 Evaluation metrics","7 End-to-end evaluation","Appendix A limitations"]}],"reference_number":79,"title":"RedTeamCUA: Realistic Adversarial Testing of Computer-Use Agents in Hybrid Web-OS Environments"},{"canonical_id":"arxiv:2607.00333","citation_key":"arxiv_2607_00333","citation_url":"https://arxiv.org/abs/2607.00333v2","edition_label":"arXiv v2; first submitted 2026-07-01","ordered_authors":["Zidong Zhang","Zhentao Xie","Wenrui Diao","Jianliang Wu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,4,12,13],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:51.142293+00:00","sha256":"97bf1ad827e6c378b90ae67d58a77e19dc0fce1a4dc7125ed3c8687e6da5de8c","url":"https://arxiv.org/pdf/2607.00333v2","version":2}],"read_level":"selected_fulltext_sections_read","sections":["III Threat model","Table II permissions","VII Discussion/limitations"]}],"reference_number":80,"title":"(A)I Sees What You Don't: Exploiting New Attack Surfaces in Third-Party Mobile Agents"},{"canonical_id":"arxiv:2506.13205","citation_key":"arxiv_2506_13205","citation_url":"https://arxiv.org/abs/2506.13205v6","edition_label":"arXiv v6; first submitted 2025-06-16","ordered_authors":["Xuan Wang","Siyuan Liang","Zhe Liu","Yi Yu","Aishan Liu","Yuliang Lu","Xitong Gao","Ee-Chien Chang"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[2,5,7],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:59.891515+00:00","sha256":"c0f3e6d21428ae8f75c090006cd5d97f44b6d945cab072208408efbaf3b507da","url":"https://arxiv.org/pdf/2506.13205v6","version":6}],"read_level":"selected_fulltext_sections_read","sections":["Threat model","Evaluation metrics","Conclusion"]}],"reference_number":81,"title":"Poison Once, Control Anywhere: Clean-Text Visual Backdoors in VLM-based Mobile Agents"},{"canonical_id":"arxiv:2507.00841","citation_key":"arxiv_2507_00841","citation_url":"https://arxiv.org/abs/2507.00841v1","edition_label":"arXiv v1; first submitted 2025-07-01","ordered_authors":["Siyuan Liang","Tianmeng Fang","Zhe Liu","Aishan Liu","Yan Xiao","Jinyuan He","Ee-Chien Chang","Xiaochun Cao"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"evaluation","pdf_pages":[3,6,7,10],"primary_sources":[{"retrieved_at":"2026-10-18T10:10:56.912207+00:00","sha256":"a25da6700a238167d5e9915cf6fb3a4bcd910cefc300f4be6eed56271be26f00","url":"https://arxiv.org/pdf/2507.00841v1","version":1}],"read_level":"selected_fulltext_sections_read","sections":["II.B Defense objective","III Chain-level guard and GPTJudge","IV.B Evaluation setup","V.D Judge validity"]}],"reference_number":82,"title":"SafeMobile: Chain-level Jailbreak Detection and Automated Evaluation for Multimodal Mobile Agents"},{"canonical_id":"arxiv:2608.14132","citation_key":"arxiv_2608_14132","citation_url":"https://arxiv.org/abs/2608.14132v1","edition_label":"arXiv v1; first submitted 2026-08-14","ordered_authors":["Xiaokai Yan","Jingtao Ding","Yong Li","Zhiwen Yu"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"review design, human agency and research assets","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:22.211464+00:00","sha256":"8d6aa8484ff4a7cbaac589c2c220407ebf6ecd1e216015a248125abdb0f2ed2d","url":"https://arxiv.org/html/2608.14132v1"}],"read_level":"selected_fulltext_sections_read","sections":["3 Task definition","4.1 Raw data collection","4.2 Trajectory construction","7 Limitations and future directions"]}],"reference_number":83,"title":"Act2Intention: A Benchmark For Developing Active Mobile Agents Through Inferring User Intention from GUI Actions"},{"canonical_id":"arxiv:2507.18165","citation_key":"arxiv_2507_18165","citation_url":"https://arxiv.org/abs/2507.18165v1","edition_label":"arXiv v1; first submitted 2025-07-24","ordered_authors":["Yuheng Zhao","Xueli Shu","Liwen Fan","Lin Gao","Yu Zhang","Siming Chen"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"review design, human agency and research assets","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:30.529800+00:00","sha256":"7ca9569113479791a0482f4c196002a8459f1826eb6a3c9fab8a701b33360593","url":"https://arxiv.org/html/2507.18165v1"}],"read_level":"selected_fulltext_sections_read","sections":["5.3 User study","5.3.2 Observed issues","6.1 Design trade-off","6.2 Limitations"]}],"reference_number":84,"title":"ProactiveVA: Proactive Visual Analytics with LLM-Based UI Agent"},{"canonical_id":"arxiv:2608.09944","citation_key":"arxiv_2608_09944","citation_url":"https://arxiv.org/abs/2608.09944v1","edition_label":"arXiv v1; first submitted 2026-07-01","ordered_authors":["Santosh Patapati"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"review design, human agency and research assets","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:26.923383+00:00","sha256":"777197864eeaa34d180bf6b91f32408b7b3d8cc5c9faecc95eca276f5bd426de","url":"https://arxiv.org/html/2608.09944v1"}],"read_level":"selected_fulltext_sections_read","sections":["3.3 Interface states and task annotations","3.4 Evaluation","4.1 Models and experimental setup","5 Discussion"]}],"reference_number":85,"title":"Navigation Alone Is Not Enough: Evaluating Explanatory Assistive UI Agents"},{"canonical_id":"doi:10.1038/sdata.2016.18","citation_key":"wilkinson2016fair","citation_url":"https://www.nature.com/articles/sdata201618","edition_label":"2016","ordered_authors":["Wilkinson, Mark D.","Dumontier, Michel","Aalbersberg, IJsbrand Jan","Appleton, Gabrielle","Axton, Myles","Baak, Arie","Blomberg, Niklas","Boiten, Jan-Willem","da Silva Santos, Luiz Bonino","Bourne, Philip E.","Bouwman, Jildau","Brookes, Anthony J.","Clark, Tim","Crosas, Mercè","Dillo, Ingrid","Dumon, Olivier","Edmunds, Scott","Evelo, Chris T.","Finkers, Richard","Gonzalez-Beltran, Alejandra","Gray, Alasdair J.G.","Groth, Paul","Goble, Carole","Grethe, Jeffrey S.","Heringa, Jaap","’t Hoen, Peter A.C","Hooft, Rob","Kuhn, Tobias","Kok, Ruben","Kok, Joost","Lusher, Scott J.","Martone, Maryann E.","Mons, Albert","Packer, Abel L.","Persson, Bengt","Rocca-Serra, Philippe","Roos, Marco","van Schaik, Rene","Sansone, Susanna-Assunta","Schultes, Erik","Sengstag, Thierry","Slater, Ted","Strawn, George","Swertz, Morris A.","Thompson, Mark","van der Lei, Johan","van Mulligen, Erik","Velterop, Jan","Waagmeester, Andra","Wittenburg, Peter","Wolstencroft, Katherine","Zhao, Jun","Mons, Barend"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"review design, human agency and research assets","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:48.050920+00:00","sha256":"a1ffb0d1dd0fdc4fb3d98e543e9f2a33f5951d519cf2b49ccc41f60401d5efc3","url":"https://www.nature.com/articles/sdata201618"}],"read_level":"selected_fulltext_sections_read","sections":["The FAIR Guiding Principles in detail","Box 2","The Principles precede implementation"]}],"reference_number":86,"title":"The FAIR Guiding Principles for scientific data management and stewardship"},{"canonical_id":"doi:10.1038/s41597-022-01710-x","citation_key":"barker2022fair4rs","citation_url":"https://www.nature.com/articles/s41597-022-01710-x","edition_label":"2022","ordered_authors":["Barker, Michelle","Chue Hong, Neil P.","Katz, Daniel S.","Lamprecht, Anna-Lena","Martinez-Ortiz, Carlos","Psomopoulos, Fotis","Harrow, Jennifer","Castro, Leyla Jael","Gruenpeter, Morane","Martinez, Paula Andrea","Honeyman, Tom"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"review design, human agency and research assets","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:06:01.695367+00:00","sha256":"a6f9d6a33a1d1123f3fe94c69b299bf2a0eab0d903ada331059770109daad8ee","url":"https://www.nature.com/articles/s41597-022-01710-x"}],"read_level":"selected_fulltext_sections_read","sections":["Introduction","Results: FAIR4RS Principles"]}],"reference_number":87,"title":"Introducing the FAIR Principles for research software"},{"canonical_id":"arxiv:1803.09010","citation_key":"arxiv_1803_09010","citation_url":"https://arxiv.org/abs/1803.09010v8","edition_label":"arXiv v8; first submitted 2018-03-23","ordered_authors":["Timnit Gebru","Jamie Morgenstern","Briana Vecchione","Jennifer Wortman Vaughan","Hanna Wallach","Hal Daumé III","Kate Crawford"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"review design, human agency and research assets","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:34.384454+00:00","sha256":"c672395cf1c889a2e7512ca9b42814038075aff5082f71007d65337305f2ce5a","url":"https://arxiv.org/html/1803.09010v8"}],"read_level":"selected_fulltext_sections_read","sections":["3 Questions and workflow"]}],"reference_number":88,"title":"Datasheets for Datasets"},{"canonical_id":"arxiv:1810.03993","citation_key":"arxiv_1810_03993","citation_url":"https://arxiv.org/abs/1810.03993v2","edition_label":"arXiv v2; first submitted 2018-10-05","ordered_authors":["Margaret Mitchell","Simone Wu","Andrew Zaldivar","Parker Barnes","Lucy Vasserman","Ben Hutchinson","Elena Spitzer","Inioluwa Deborah Raji","Timnit Gebru"],"read_level":"selected_fulltext_sections_read","readings":[{"chapter":"review design, human agency and research assets","pdf_pages":[],"primary_sources":[{"retrieved_at":"2026-10-18T10:05:41.022829+00:00","sha256":"971e98662e58d700e887638adb4ab66fbc0ae4e119a4f4f8a09a1dd89919b20d","url":"https://arxiv.org/html/1810.03993v2"}],"read_level":"selected_fulltext_sections_read","sections":["4 Model card sections","6 Discussion and future work"]}],"reference_number":89,"title":"Model Cards for Model Reporting"}],"schema_version":1,"source_commit":"87166fe8ad9c8e832f7a90ee84b2e93c6e0e40db","working_manuscript_date":"2026-10-18"}