{"bibliographic_cautions":[{"id":"arxiv:2503.23350","issue":"Unfilled ACM placeholder metadata in generated full-text HTML (2018 Woodstock and XXXXX DOI).","resolution":"Use verified exact-version abstract-page identity/submission history. Do not infer venue/DOI from template strings."},{"id":"arxiv:2501.16150","issue":"Primary citation_publication_date is 2026 while first submission is 27 January 2025.","resolution":"Keep first submission and v3 revision separately; primary submission-history text verifies the first date."},{"id":"url:cccd967aa993404d74d0","issue":"IBM metadata has 2007/10/22; paper conference interval is April–May 2007 and frozen observation retains year precision.","resolution":"Use year 2007 only in this historical narrative. Do not silently change publication observation or manufacture a conference-derived day."},{"id":"url:b560b5d6abaab98483f6","issue":"Original PDF dataset quantities and later project-release quantities are not identical.","resolution":"No merged numerical size claim; preserve paper/resource distinction."},{"id":"arxiv:2405.14573","issue":"First-submission and inspected revision years differ.","resolution":"Cite v5 source, retain first-submission and revision dates explicitly in the evidence record."}],"citation_key_convention":"scripts/export_citations.py citation_key: arxiv IDs use arxiv_ plus numeric ID with underscore; non-arxiv use cua_ plus first 24 SHA256 hex characters of canonical ID; retained record edition is cited.","citation_map":{"arxiv_1802_08802":"arxiv:1802.08802","arxiv_2207_01206":"arxiv:2207.01206","arxiv_2210_03347":"arxiv:2210.03347","arxiv_2306_06070":"arxiv:2306.06070","arxiv_2307_10088":"arxiv:2307.10088","arxiv_2307_13854":"arxiv:2307.13854","arxiv_2404_07972":"arxiv:2404.07972","arxiv_2405_14573":"arxiv:2405.14573","arxiv_2411_04890":"arxiv:2411.04890","arxiv_2411_18279":"arxiv:2411.18279","arxiv_2501_16150":"arxiv:2501.16150","arxiv_2503_23350":"arxiv:2503.23350","arxiv_2503_23434":"arxiv:2503.23434","arxiv_2504_19838":"arxiv:2504.19838","arxiv_2508_04482":"arxiv:2508.04482","arxiv_2609_02309":"arxiv:2609.02309","cua_1b4a3a269ca9f1884a566476":"url:cccd967aa993404d74d0","cua_402ed654426712089390317c":"url:b560b5d6abaab98483f6","cua_510831381c08385eb74fb95e":"url:a4cdb2ee0d07d3980651","cua_6b7f9d608ed89d1625610eb1":"url:392a602c660ffa91be46","cua_ba263493b78b0263c7933138":"url:52a3261eb6a19b402e88"},"cited_canonical_ids":["arxiv:1802.08802","arxiv:2207.01206","arxiv:2210.03347","arxiv:2306.06070","arxiv:2307.10088","arxiv:2307.13854","arxiv:2404.07972","arxiv:2405.14573","arxiv:2411.04890","arxiv:2411.18279","arxiv:2501.16150","arxiv:2503.23350","arxiv:2503.23434","arxiv:2504.19838","arxiv:2508.04482","arxiv:2609.02309","url:cccd967aa993404d74d0","url:b560b5d6abaab98483f6","url:a4cdb2ee0d07d3980651","url:392a602c660ffa91be46","url:52a3261eb6a19b402e88"],"formal_data_changed":false,"format":"semantic HTML fragment","inventory_anchor":{"active_records":3048,"catalog_bytes_reverified":true,"catalog_sha256":"7f71a1683ab013e587a20086c3e357956981f517a24ae8bc8e81791108c66e4f","count_meaning":"Frozen discovery inventory, not a count of papers whose complete texts were read for this manuscript.","curated_papers_and_surveys":1123,"source_commit":"87166fe8ad9c8e832f7a90ee84b2e93c6e0e40db"},"language":"en","output":{"html":"research/survey/sections/context.html","html_sha256":"0ccf28e374d1511857cd3b8f34c6910706000493efa2155cc1c015c99a8bc7ed"},"reading_counts":{"additional_foundational_abstract_and_author_project_only":1,"deep_prior_survey_comparisons":8,"foundational_fulltext_section_inspections":12,"full_document_exhaustive_audits":0,"survey_primary_metadata_and_abstract_checks":15,"unique_cited_papers":21,"unique_fulltext_section_inspections":20,"unique_papers_in_context_evidence":28},"receipt_reading_status_note":"Acquisition receipts preserve their not_yet_read status at download time. The retained_sources entries, source-map.json and read-decisions.json record the completed current reading pass without rewriting acquisition history.","records":[{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:1802.08802","catalog_first_date":"2018-02-24","citation_key":"arxiv_1802_08802","finding":"Derives abstract workflow constraints from demonstrations to guide sparse-reward exploration, then trains a neural policy from successful episodes; introduces MiniWoB++ task extensions.","limits":["Its action space is restricted to DOM-element clicks and typing strings from the goal.","The reported original MiniWoB experiment filters tasks to that action space; no cross-benchmark ranking is made."],"metadata_abstract_extracted_sha256":"75d19fa97ce85a93bb9665a8ae06607bc0b90fde1c8376cdd60e99b55e90729c","ordered_authors":["Evan Zheran Liu","Kelvin Guu","Panupong Pasupat","Tianlin Shi","Percy Liang"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 24 Feb 2018]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["1802.08802"],"citation_author":["Liu, Evan Zheran","Guu, Kelvin","Pasupat, Panupong","Shi, Tianlin","Liang, Percy"],"citation_date":["2018/02/24"],"citation_online_date":["2018/02/24"],"citation_pdf_url":["https://arxiv.org/pdf/1802.08802"],"citation_title":["Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"75d19fa97ce85a93bb9665a8ae06607bc0b90fde1c8376cdd60e99b55e90729c","submitted_at":"2018-02-24T05:32:47Z","updated_at":"2018-02-24T05:32:47Z","version":1,"versioned_url":"https://arxiv.org/abs/1802.08802v1"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","2 Setup","3 Inducing workflows","6.1 Task setups","6.3.1 MiniWoB++"],"source_receipt_keys":["arxiv-1802.08802.metadata.html","arxiv-1802.08802.full.html"],"supported_context_claims":["Demonstrations can constrain exploration without being a literal policy to replay.","MiniWoB++ is a distinct follow-up contribution to MiniWoB."],"title":"Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2207.01206","catalog_first_date":"2022-07-04","citation_key":"arxiv_2207_01206","finding":"A simulated shopping environment uses real product information, compositional natural-language requirements, search/choose actions and product-based programmatic rewards.","limits":["HTML mode was used for human interaction, while reported models trained/evaluated in simple mode.","In sim-to-real experiments, consequential form/option/buy actions were kept inside the interface for reward computation.","Product and instruction coverage is not representative of every language, country or shopping task."],"metadata_abstract_extracted_sha256":"a414e32694cde7db74ff94b9d3f1b1b186d49f7cedd3e7d1b33f0bfe146c20ea","ordered_authors":["Shunyu Yao","Howard Chen","John Yang","Karthik Narasimhan"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 4 Jul 2022 (<a href=\"https://arxiv.org/abs/2207.01206v1\">v1</a>), last revised 8 Feb 2023 (this version, v4)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2207.01206"],"citation_author":["Yao, Shunyu","Chen, Howard","Yang, John","Narasimhan, Karthik"],"citation_date":["2022/07/04"],"citation_online_date":["2023/02/08"],"citation_pdf_url":["https://arxiv.org/pdf/2207.01206"],"citation_title":["WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"6de4ab2a89875f1ccd99bbaa075c234d82979e845f27fe9530bdad0539c49570","submitted_at":"2022-07-04T05:30:22Z","updated_at":"2023-02-08T01:39:30Z","version":4,"versioned_url":"https://arxiv.org/abs/2207.01206v4"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","3.1 Task Formulation","3.2 Environment Implementation","Appendix E Potential Societal Impacts and Limitations"],"source_receipt_keys":["arxiv-2207.01206.metadata.html","arxiv-2207.01206.full.html"],"supported_context_claims":["It is a simulated web environment, not unrestricted live shopping.","Model evaluation used simple mode; transfer limited real-site behavior to navigation."],"title":"WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2210.03347","catalog_first_date":"2022-10-07","citation_key":"arxiv_2210_03347","finding":"Trains a pixel-to-text model using masked webpage screenshots and simplified HTML targets; downstream UI tasks include referring-expression selection and screen/widget captioning.","limits":["The cited UI tasks do not demonstrate autonomous multistep software execution.","The paper acknowledges resolution bottlenecks and that OCR or metadata can remain appropriate."],"metadata_abstract_extracted_sha256":"7ebb243c4690aba5f1a92b0d1e050399560998a7f304ea483dae2f451cf51ec0","ordered_authors":["Kenton Lee","Mandar Joshi","Iulia Turc","Hexiang Hu","Fangyu Liu","Julian Eisenschlos","Urvashi Khandelwal","Peter Shaw","Ming-Wei Chang","Kristina Toutanova"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 7 Oct 2022 (<a href=\"https://arxiv.org/abs/2210.03347v1\">v1</a>), last revised 15 Jun 2023 (this version, v2)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2210.03347"],"citation_author":["Lee, Kenton","Joshi, Mandar","Turc, Iulia","Hu, Hexiang","Liu, Fangyu","Eisenschlos, Julian","Khandelwal, Urvashi","Shaw, Peter","Chang, Ming-Wei","Toutanova, Kristina"],"citation_date":["2022/10/07"],"citation_online_date":["2023/06/15"],"citation_pdf_url":["https://arxiv.org/pdf/2210.03347"],"citation_title":["Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"7ebb243c4690aba5f1a92b0d1e050399560998a7f304ea483dae2f451cf51ec0","submitted_at":"2022-10-07T06:42:06Z","updated_at":"2023-06-15T21:34:23Z","version":2,"versioned_url":"https://arxiv.org/abs/2210.03347v2"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","2.2 Architecture","2.3 Pretraining","2.5 Finetuning","3.1 Benchmarks","4.2 UIs","6 Discussion"],"source_receipt_keys":["arxiv-2210.03347.metadata.html","arxiv-2210.03347.full.html"],"supported_context_claims":["UI understanding is directly relevant to agents without itself being an interactive-control result."],"title":"Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2306.06070","catalog_first_date":"2023-06-09","citation_key":"arxiv_2306_06070","finding":"Collects high-level tasks, demonstrated actions and webpage snapshots from real websites to study generalization across tasks, sites and domains.","limits":["Evaluation uses cached offline environments and may reject uncached but valid alternatives.","The initial MindAct approach uses textual context; broad multimodal and conversational interactions are future directions.","Website/language/annotator coverage is limited."],"metadata_abstract_extracted_sha256":"3ab9d60bf8314d6b9bb05c93ccd56098c2d5a8498a89da21e848db136c5cc2e1","ordered_authors":["Xiang Deng","Yu Gu","Boyuan Zheng","Shijie Chen","Samuel Stevens","Boshi Wang","Huan Sun","Yu Su"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 9 Jun 2023 (<a href=\"https://arxiv.org/abs/2306.06070v1\">v1</a>), last revised 9 Dec 2023 (this version, v3)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2306.06070"],"citation_author":["Deng, Xiang","Gu, Yu","Zheng, Boyuan","Chen, Shijie","Stevens, Samuel","Wang, Boshi","Sun, Huan","Su, Yu"],"citation_date":["2023/06/09"],"citation_online_date":["2023/12/09"],"citation_pdf_url":["https://arxiv.org/pdf/2306.06070"],"citation_title":["Mind2Web: Towards a Generalist Agent for the Web"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"de055dfa452130f3576839c99c007053c62a0c96c9bfbcdd79ecad1a9f74f830","submitted_at":"2023-06-09T17:44:31Z","updated_at":"2023-12-09T05:57:46Z","version":3,"versioned_url":"https://arxiv.org/abs/2306.06070v3"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","2.1 Task Definition","2.2 Data Collection","3 Method overview","6 Limitations and Potential Societal Impact"],"source_receipt_keys":["arxiv-2306.06070.metadata.html","arxiv-2306.06070.full.html"],"supported_context_claims":["Real-website collection does not imply that reported evaluation is unrestricted live interaction."],"title":"Mind2Web: Towards a Generalist Agent for the Web"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2307.10088","catalog_first_date":"2023-07-19","citation_key":"arxiv_2307_10088","finding":"Collects Android task demonstrations with screenshots, coordinate gestures, typed text and navigation actions across versions/device configurations; defines action-matching evaluation.","limits":["Raters use mouse/keyboard with emulators, not natural touch interaction from ordinary users.","Prompts are English, form factor is phones, and UI drift is only partially covered.","No claim that demonstrations alone establish recovery from agent-generated errors."],"metadata_abstract_extracted_sha256":"6bf3d898c91a16ccaad1b90a44ccf0ea26cdcc16e7054587ee4fde21b5cb3c27","ordered_authors":["Christopher Rawles","Alice Li","Daniel Rodriguez","Oriana Riva","Timothy Lillicrap"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 19 Jul 2023 (<a href=\"https://arxiv.org/abs/2307.10088v1\">v1</a>), last revised 27 Oct 2023 (this version, v2)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2307.10088"],"citation_author":["Rawles, Christopher","Li, Alice","Rodriguez, Daniel","Riva, Oriana","Lillicrap, Timothy"],"citation_date":["2023/07/19"],"citation_online_date":["2023/10/27"],"citation_pdf_url":["https://arxiv.org/pdf/2307.10088"],"citation_title":["Android in the Wild: A Large-Scale Dataset for Android Device Control"]},"publication_metadata":{"authors":["Christopher Rawles","Alice Li","Daniel Rodriguez","Oriana Riva","Timothy Lillicrap"],"date_basis":"Explicit publisher citation_publication_date: 2023-12-15. Preserve supplied precision; do not substitute conference event dates.","date_parts":[2023,12,15],"doi":"10.52202/075280-2609","event_dates":null,"evidence_urls":["https://proceedings.neurips.cc/paper_files/paper/2023/hash/bbbb6308b402fe909c39dd29950c32e0-Abstract-Datasets_and_Benchmarks.html"],"identity_basis":"Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.","issue":null,"notes":"Publisher proceedings edition; retained separately from the arXiv snapshot.","pages":"59708-59728","pdf_url":"https://proceedings.neurips.cc/paper_files/paper/2023/file/bbbb6308b402fe909c39dd29950c32e0-Paper-Datasets_and_Benchmarks.pdf","publisher":null,"series":null,"title":"AndroidInTheWild: A Large-Scale Dataset For Android Device Control","type":"conference-paper","url":"https://proceedings.neurips.cc/paper_files/paper/2023/hash/bbbb6308b402fe909c39dd29950c32e0-Abstract-Datasets_and_Benchmarks.html","venue":"Advances in Neural Information Processing Systems","volume":"36"},"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"def4616f7cb13fe839e500c4b3171eafbe37b155b5d14301137806d6a5cde939","submitted_at":"2023-07-19T15:57:24Z","updated_at":"2023-10-27T14:24:31Z","version":2,"versioned_url":"https://arxiv.org/abs/2307.10088v2"},"reviewed_at":"2026-10-18T10:22:08Z","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"],"source_receipt_keys":["arxiv-2307.10088.metadata.html","arxiv-2307.10088.full.html"],"supported_context_claims":["Behavioral demonstrations and live success are different evidence types.","The paper explicitly contrasts offline matching and human online validation."],"title":"Android in the Wild: A Large-Scale Dataset for Android Device Control"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2307.13854","catalog_first_date":"2023-07-25","citation_key":"arxiv_2307_13854","finding":"Provides standalone functional websites, browser actions and task-specific success checks so evaluation can inspect achieved state rather than only a demonstrated action string.","limits":["The websites are deliberately controlled self-hosted instances, not the live public services.","Different observation/action configurations should not be silently equated."],"metadata_abstract_extracted_sha256":"0e9eca0165573ba53b663acbb1c7962fd2d64bc6633f10397f9cad229772a363","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 25 Jul 2023 (<a href=\"https://arxiv.org/abs/2307.13854v1\">v1</a>), last revised 16 Apr 2024 (this version, v4)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2307.13854"],"citation_author":["Zhou, Shuyan","Xu, Frank F.","Zhu, Hao","Zhou, Xuhui","Lo, Robert","Sridhar, Abishek","Cheng, Xianyi","Ou, Tianyue","Bisk, Yonatan","Fried, Daniel","Alon, Uri","Neubig, Graham"],"citation_date":["2023/07/25"],"citation_online_date":["2024/04/16"],"citation_pdf_url":["https://arxiv.org/pdf/2307.13854"],"citation_title":["WebArena: A Realistic Web Environment for Building Autonomous Agents"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"0e9eca0165573ba53b663acbb1c7962fd2d64bc6633f10397f9cad229772a363","submitted_at":"2023-07-25T22:59:32Z","updated_at":"2024-04-16T15:13:18Z","version":4,"versioned_url":"https://arxiv.org/abs/2307.13854v4"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","2 Environment including 2.1–2.4","3 benchmark overview","3.2 Evaluation Annotation"],"source_receipt_keys":["arxiv-2307.13854.metadata.html","arxiv-2307.13854.full.html"],"supported_context_claims":["Interactive evaluation need not depend on mutable public websites.","Observation and target-selection options are explicit components of the environment."],"title":"WebArena: A Realistic Web Environment for Building Autonomous Agents"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2404.07972","catalog_first_date":"2024-04-11","citation_key":"arxiv_2404_07972","finding":"Provides virtual-machine computer environments, task initialization and execution-based checks for desktop software, files and multi-application workflows.","limits":["The main v2 task suite is on Ubuntu, with adapted Windows tasks; infrastructure support is broader.","The conclusion notes that task correctness checks may miss collateral harmful actions.","No scores are compared across heterogeneous benchmark contracts."],"metadata_abstract_extracted_sha256":"10e297624ea0fbee20c5e5f8e6d10fa30d4b172a71f6dffff5e533e7e62b9f29","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 11 Apr 2024 (<a href=\"https://arxiv.org/abs/2404.07972v1\">v1</a>), last revised 30 May 2024 (this version, v2)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2404.07972"],"citation_author":["Xie, Tianbao","Zhang, Danyang","Chen, Jixuan","Li, Xiaochuan","Zhao, Siheng","Cao, Ruisheng","Hua, Toh Jing","Cheng, Zhoujun","Shin, Dongchan","Lei, Fangyu","Liu, Yitao","Xu, Yiheng","Zhou, Shuyan","Savarese, Silvio","Xiong, Caiming","Zhong, Victor","Yu, Tao"],"citation_date":["2024/04/11"],"citation_online_date":["2024/05/30"],"citation_pdf_url":["https://arxiv.org/pdf/2404.07972"],"citation_title":["OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments"]},"publication_metadata":{"authors":["Tianbao Xie","Danyang Zhang","Jixuan Chen","Xiaochuan Li","Siheng Zhao","Ruisheng Cao","Toh J. Hua","Zhoujun Cheng","Dongchan Shin","Fangyu Lei","Yitao Liu","Yiheng Xu","Shuyan Zhou","Silvio Savarese","Caiming Xiong","Victor Zhong","Tao Yu"],"date_basis":"Explicit publisher citation_publication_date: 2024-12-16. Preserve supplied precision; do not substitute conference event dates.","date_parts":[2024,12,16],"doi":"10.52202/079017-1650","event_dates":null,"evidence_urls":["https://proceedings.neurips.cc/paper_files/paper/2024/hash/5d413e48f84dc61244b6be550f1cd8f5-Abstract-Datasets_and_Benchmarks_Track.html"],"identity_basis":"Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.","issue":null,"notes":"Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.","pages":"52040-52094","pdf_url":"https://proceedings.neurips.cc/paper_files/paper/2024/file/5d413e48f84dc61244b6be550f1cd8f5-Paper-Datasets_and_Benchmarks_Track.pdf","publisher":null,"series":null,"title":"OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments","type":"conference-paper","url":"https://proceedings.neurips.cc/paper_files/paper/2024/hash/5d413e48f84dc61244b6be550f1cd8f5-Abstract-Datasets_and_Benchmarks_Track.html","venue":"Advances in Neural Information Processing Systems","volume":"37"},"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"a347bfcb4de03a60410392b52d522e5b7ad5dbabb474d60fc1ec097e74c4008d","submitted_at":"2024-04-11T17:56:05Z","updated_at":"2024-05-30T08:55:12Z","version":2,"versioned_url":"https://arxiv.org/abs/2404.07972v2"},"reviewed_at":"2026-10-18T10:22:08Z","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"],"source_receipt_keys":["arxiv-2404.07972.metadata.html","arxiv-2404.07972.full.html"],"supported_context_claims":["Infrastructure support and the platform coverage of a particular benchmark/experiment must be distinguished."],"title":"OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2405.14573","catalog_first_date":"2024-05-23","citation_key":"arxiv_2405_14573","finding":"Uses parameterized Android tasks and controlled initialization, success checks and cleanup based on system/application state; exposes screenshots/UI trees and device actions.","limits":["The described setup fixes a Pixel 6 emulator, Android 13, app versions and initial clock.","It focuses on open-source and built-in apps and avoids apps requiring login for reproducibility.","A first-submission year of 2024 is separate from the inspected v5 revision in 2025."],"metadata_abstract_extracted_sha256":"05a1ecb7b7c1cb1ff916f8135b9d2399bf26e8eb11914971a61db643f7e87532","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 23 May 2024 (<a href=\"https://arxiv.org/abs/2405.14573v1\">v1</a>), last revised 6 Apr 2025 (this version, v5)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2405.14573"],"citation_author":["Rawles, Christopher","Clinckemaillie, Sarah","Chang, Yifan","Waltz, Jonathan","Lau, Gabrielle","Fair, Marybeth","Li, Alice","Bishop, William","Li, Wei","Campbell-Ajala, Folawiyo","Toyama, Daniel","Berry, Robert","Tyamagundlu, Divya","Lillicrap, Timothy","Riva, Oriana"],"citation_date":["2024/05/23"],"citation_online_date":["2025/04/06"],"citation_pdf_url":["https://arxiv.org/pdf/2405.14573"],"citation_title":["AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents"]},"publication_metadata":{"authors":["Chris 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"],"date_basis":"Explicit publisher citation_publication_date: 2025-05-01. Preserve supplied precision; do not substitute conference event dates.","date_parts":[2025,5,1],"doi":null,"event_dates":null,"evidence_urls":["https://proceedings.iclr.cc/paper_files/paper/2025/hash/01a83bc2f2732a58e6aa731e659e7101-Abstract-Conference.html"],"identity_basis":"Same work: normalized title, ordered author correspondence and the specific contribution in the complete publisher abstract were reviewed against the retained arXiv identity.","issue":null,"notes":"Publisher proceedings edition; retained separately from the arXiv snapshot. Publisher spelling, initials, name annotations or author lists differ from the retained preprint.","pages":"406-441","pdf_url":"https://proceedings.iclr.cc/paper_files/paper/2025/file/01a83bc2f2732a58e6aa731e659e7101-Paper-Conference.pdf","publisher":null,"series":null,"title":"AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents","type":"conference-paper","url":"https://proceedings.iclr.cc/paper_files/paper/2025/hash/01a83bc2f2732a58e6aa731e659e7101-Abstract-Conference.html","venue":"International Conference on Learning Representations","volume":"2025"},"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"a33b8b0a03c0dc7ac3f699c34193a526d20791bfedfdde2a2fb89604959e08c2","submitted_at":"2024-05-23T13:48:54Z","updated_at":"2025-04-06T20:37:50Z","version":5,"versioned_url":"https://arxiv.org/abs/2405.14573v5"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","3.1–3.5 AndroidWorld environment","Appendix A Limitations"],"source_receipt_keys":["arxiv-2405.14573.metadata.html","arxiv-2405.14573.full.html"],"supported_context_claims":["Parameterized tasks and application state validation are concrete reproducibility choices."],"title":"AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2411.02006","catalog_first_date":"2024-11-04","citation_key":"arxiv_2411_02006","finding":"Mobile-agent survey; abstract organizes evaluation environments and prompting/training approaches with multimodal adaptability.","limits":["No full-text claim is based on this inspection."],"metadata_abstract_extracted_sha256":"ca08b725fb3281382c7eaeab8903f5c5133f9cf19303e549fc4f7f50ca82f403","ordered_authors":["Biao Wu","Yanda Li","Zhiwei Zhang","Yunchao Wei","Meng Fang","Ling Chen"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 4 Nov 2024 (<a href=\"https://arxiv.org/abs/2411.02006v1\">v1</a>), last revised 14 Sep 2025 (this version, v3)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2411.02006"],"citation_author":["Wu, Biao","Li, Yanda","Zhang, Zhiwei","Wei, Yunchao","Fang, Meng","Chen, Ling"],"citation_date":["2024/11/04"],"citation_online_date":["2025/09/14"],"citation_pdf_url":["https://arxiv.org/pdf/2411.02006"],"citation_title":["Foundations and Recent Trends in Multimodal Mobile Agents: A Survey"]},"publication_metadata":null,"read_status":"primary_metadata_and_complete_abstract_only","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"de860a29f6c54317586c2218688436898f1d7a8b61207b1d5f4a65b05b3a4e6e","submitted_at":"2024-11-04T11:50:58Z","updated_at":"2025-09-14T09:13:39Z","version":3,"versioned_url":"https://arxiv.org/abs/2411.02006v3"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["primary abstract-page title, ordered authors, dates, version and full abstract"],"source_receipt_keys":["arxiv-2411.02006.metadata.html"],"supported_context_claims":["Inventory screening only; not a deep comparison row."],"title":"Foundations and Recent Trends in Multimodal Mobile Agents: A Survey"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2411.04890","catalog_first_date":"2024-11-07","citation_key":"arxiv_2411_04890","finding":"Organizes foundation-model GUI research through data resources, agent framework and applications; distinguishes static and dynamic environments and discusses action matching versus task outcome checks.","limits":["Used to characterize the survey taxonomy, not to independently validate the experiments it summarizes.","No industrial-product capability claim is adopted from its application section."],"metadata_abstract_extracted_sha256":"57670445f9b0041bfb505a195d01a9ded6234e01e44c9aa43a6aee5ebd57dd93","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 7 Nov 2024 (<a href=\"https://arxiv.org/abs/2411.04890v1\">v1</a>), last revised 13 Feb 2025 (this version, v2)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2411.04890"],"citation_author":["Wang, Shuai","Liu, Weiwen","Chen, Jingxuan","Zhou, Yuqi","Gan, Weinan","Zeng, Xingshan","Che, Yuhan","Yu, Shuai","Hao, Xinlong","Shao, Kun","Wang, Bin","Wu, Chuhan","Wang, Yasheng","Tang, Ruiming","Hao, Jianye"],"citation_date":["2024/11/07"],"citation_online_date":["2025/02/13"],"citation_pdf_url":["https://arxiv.org/pdf/2411.04890"],"citation_title":["GUI Agents with Foundation Models: A Comprehensive Survey"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"57670445f9b0041bfb505a195d01a9ded6234e01e44c9aa43a6aee5ebd57dd93","submitted_at":"2024-11-07T17:28:10Z","updated_at":"2025-02-13T10:09:37Z","version":2,"versioned_url":"https://arxiv.org/abs/2411.04890v2"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","2 GUI Agent Data Resources","3.1 GUI Agent Framework","3.2 modality and learning taxonomy"],"source_receipt_keys":["arxiv-2411.04890.metadata.html","arxiv-2411.04890.full.html"],"supported_context_claims":["Broad GUI surveys already address datasets, evaluation and architecture.","Its generic framework separates perception, task planning, decisions, memory retrieval and execution."],"title":"GUI Agents with Foundation Models: A Comprehensive Survey"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2411.10943","catalog_first_date":"2024-11-17","citation_key":"arxiv_2411_10943","finding":"Generalist virtual agents across digital platforms; abstract gives broad environment/task/capability taxonomy. Current survey pass does not reuse a past full-text review as a new read.","limits":["No full-text claim is based on this inspection."],"metadata_abstract_extracted_sha256":"daa9b7092616e94a1d0124d36d8b75d0401b9647ec762c963524616a92def9a5","ordered_authors":["Minghe Gao","Wendong Bu","Bingchen Miao","Yang Wu","Yunfei Li","Juncheng Li","Siliang Tang","Qi Wu","Yueting Zhuang","Meng Wang"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 17 Nov 2024]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2411.10943"],"citation_author":["Gao, Minghe","Bu, Wendong","Miao, Bingchen","Wu, Yang","Li, Yunfei","Li, Juncheng","Tang, Siliang","Wu, Qi","Zhuang, Yueting","Wang, Meng"],"citation_date":["2024/11/17"],"citation_online_date":["2024/11/17"],"citation_pdf_url":["https://arxiv.org/pdf/2411.10943"],"citation_title":["Generalist Virtual Agents: A Survey on Autonomous Agents Across Digital Platforms"]},"publication_metadata":null,"read_status":"primary_metadata_and_complete_abstract_only","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"daa9b7092616e94a1d0124d36d8b75d0401b9647ec762c963524616a92def9a5","submitted_at":"2024-11-17T02:44:56Z","updated_at":"2024-11-17T02:44:56Z","version":1,"versioned_url":"https://arxiv.org/abs/2411.10943v1"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["primary abstract-page title, ordered authors, dates, version and full abstract"],"source_receipt_keys":["arxiv-2411.10943.metadata.html"],"supported_context_claims":["Inventory screening only; not a deep comparison row."],"title":"Generalist Virtual Agents: A Survey on Autonomous Agents Across Digital Platforms"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2411.18279","catalog_first_date":"2024-11-27","citation_key":"arxiv_2411_18279","finding":"A broad, construction-oriented survey with explicit research questions for history, components, frameworks, data, models, evaluation and applications; discusses hybrid GUI/API agents and user intervention in a shared interface.","limits":["The email scenario in XI-D is explicitly fabricated; it is not treated as an experiment.","No claim that all sections or hundreds of cited works were audited."],"metadata_abstract_extracted_sha256":"2f692aa06152820ae60359fc2cc52aaed2ca8a97fe1efb9b8037de973956467a","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 27 Nov 2024 (<a href=\"https://arxiv.org/abs/2411.18279v1\">v1</a>), last revised 6 May 2025 (this version, v12)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2411.18279"],"citation_author":["Zhang, Chaoyun","He, Shilin","Qian, Jiaxu","Li, Bowen","Li, Liqun","Qin, Si","Kang, Yu","Ma, Minghua","Liu, Guyue","Lin, Qingwei","Rajmohan, Saravan","Zhang, Dongmei","Zhang, Qi"],"citation_date":["2024/11/27"],"citation_online_date":["2025/05/06"],"citation_pdf_url":["https://arxiv.org/pdf/2411.18279"],"citation_title":["Large Language Model-Brained GUI Agents: A Survey"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"2f692aa06152820ae60359fc2cc52aaed2ca8a97fe1efb9b8037de973956467a","submitted_at":"2024-11-27T12:13:39Z","updated_at":"2025-05-06T15:08:00Z","version":12,"versioned_url":"https://arxiv.org/abs/2411.18279v12"},"reviewed_at":"2026-10-18T10:22:08Z","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"],"source_receipt_keys":["arxiv-2411.18279.metadata.html","arxiv-2411.18279.full.html"],"supported_context_claims":["Hybrid GUI/API execution is an explicit design topic in prior work.","Human-agent collaboration includes clarification, interruption and confirmation."],"title":"Large Language Model-Brained GUI Agents: A Survey"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2412.13501","catalog_first_date":"2024-12-18","citation_key":"arxiv_2412_13501","finding":"GUI-agent survey; abstract organizes benchmarks, metrics, architecture and training around perception/reasoning/planning/action.","limits":["No full-text claim is based on this inspection."],"metadata_abstract_extracted_sha256":"ae8f83d0490e70db1ae9c55c071935fb1f59e1358335c1bd3f50835611f4360c","ordered_authors":["Dang Nguyen","Jian Chen","Yu Wang","Gang Wu","Namyong Park","Zhengmian Hu","Hanjia Lyu","Junda Wu","Ryan Aponte","Yu Xia","Xintong Li","Jing Shi","Hongjie Chen","Viet Dac Lai","Zhouhang Xie","Sungchul Kim","Ruiyi Zhang","Tong Yu","Mehrab Tanjim","Nesreen K. Ahmed","Puneet Mathur","Seunghyun Yoon","Lina Yao","Branislav Kveton","Jihyung Kil","Thien Huu Nguyen","Trung Bui","Tianyi Zhou","Ryan A. Rossi","Franck Dernoncourt"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 18 Dec 2024 (<a href=\"https://arxiv.org/abs/2412.13501v1\">v1</a>), last revised 26 Sep 2025 (this version, v3)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2412.13501"],"citation_author":["Nguyen, Dang","Chen, Jian","Wang, Yu","Wu, Gang","Park, Namyong","Hu, Zhengmian","Lyu, Hanjia","Wu, Junda","Aponte, Ryan","Xia, Yu","Li, Xintong","Shi, Jing","Chen, Hongjie","Lai, Viet Dac","Xie, Zhouhang","Kim, Sungchul","Zhang, Ruiyi","Yu, Tong","Tanjim, Mehrab","Ahmed, Nesreen K.","Mathur, Puneet","Yoon, Seunghyun","Yao, Lina","Kveton, Branislav","Kil, Jihyung","Nguyen, Thien Huu","Bui, Trung","Zhou, Tianyi","Rossi, Ryan A.","Dernoncourt, Franck"],"citation_date":["2024/12/18"],"citation_online_date":["2025/09/26"],"citation_pdf_url":["https://arxiv.org/pdf/2412.13501"],"citation_title":["GUI Agents: A Survey"]},"publication_metadata":null,"read_status":"primary_metadata_and_complete_abstract_only","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"ae8f83d0490e70db1ae9c55c071935fb1f59e1358335c1bd3f50835611f4360c","submitted_at":"2024-12-18T04:48:28Z","updated_at":"2025-09-26T21:01:59Z","version":3,"versioned_url":"https://arxiv.org/abs/2412.13501v3"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["primary abstract-page title, ordered authors, dates, version and full abstract"],"source_receipt_keys":["arxiv-2412.13501.metadata.html"],"supported_context_claims":["Inventory screening only; not a deep comparison row."],"title":"GUI Agents: A Survey"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2501.16150","catalog_first_date":"2025-01-27","citation_key":"arxiv_2501_16150","finding":"Uses domain, interaction and agent perspectives across classical and foundation-model computer-use systems; defines partial observations and executable actions and discusses incompatible metrics and deployment assumptions.","limits":["Its own conclusions decline a comprehensive cross-agent performance comparison because tasks and metrics differ.","The current context does not reuse its 87-agent/33-dataset count as the count of this survey."],"metadata_abstract_extracted_sha256":"b36fe818cb176c434b953fe8d2c97b67176b58cc6f8426b629231875de3c0896","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"],"primary_identity_check":{"date_note":"citation_publication_date is year 2026; first-submission date comes from the explicit submission-history field, not that publication-year metadata.","exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 27 Jan 2025 (<a href=\"https://arxiv.org/abs/2501.16150v1\">v1</a>), last revised 27 Mar 2026 (this version, v3)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2501.16150"],"citation_author":["Sager, Pascal J.","Meyer, Benjamin","Yan, Peng","von Wartburg-Kottler, Rebekka","Etaiwi, Layan","Enayati, Aref","Nobel, Gabriel","Abdulkadir, Ahmed","Grewe, Benjamin F.","Stadelmann, Thilo"],"citation_doi":["10.1613/jair.1.19490"],"citation_journal_title":["Journal of Artificial Intelligence Research"],"citation_online_date":["2026/03/27"],"citation_pdf_url":["https://arxiv.org/pdf/2501.16150"],"citation_publication_date":["2026"],"citation_title":["A Comprehensive Survey of Agents for Computer Use: Foundations, Challenges, and Future Directions"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"b36fe818cb176c434b953fe8d2c97b67176b58cc6f8426b629231875de3c0896","submitted_at":"2025-01-27T15:44:02Z","updated_at":"2026-03-27T07:41:10Z","version":3,"versioned_url":"https://arxiv.org/abs/2501.16150v3"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["1 Introduction","1.1 Relation to Other Surveys","2.1 Definitions","2.2 A Comprehensive Taxonomy","7 Agent Evaluation","8 Conclusions"],"source_receipt_keys":["arxiv-2501.16150.metadata.html","arxiv-2501.16150.full.html"],"supported_context_claims":["Computer-use scope can be defined without fixing the learning technology.","Environment, observation/action interface and policy are useful distinct units of analysis."],"title":"A Comprehensive Survey of Agents for Computer Use: Foundations, Challenges, and Future Directions"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2503.23350","catalog_first_date":"2025-03-30","citation_key":"arxiv_2503_23350","finding":"Centers web agents on architecture, training and trustworthiness, defining execution against webpage observations and reviewing safety/robustness, privacy and generalization.","limits":["HTML contains unfilled ACM template DOI and Woodstock 2018 venue/date text; those strings are not publication evidence.","No unsupported claim that the survey excludes other platforms or all asset provenance."],"metadata_abstract_extracted_sha256":"4b214bd8e5a34c0e8ca16a3be4fcf638a59a17ca524276dbef3c6b75f08398bb","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 30 Mar 2025 (<a href=\"https://arxiv.org/abs/2503.23350v1\">v1</a>), last revised 5 Aug 2025 (this version, v4)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2503.23350"],"citation_author":["Ning, Liangbo","Liang, Ziran","Jiang, Zhuohang","Qu, Haohao","Ding, Yujuan","Fan, Wenqi","Wei, Xiao-yong","Lin, Shanru","Liu, Hui","Yu, Philip S.","Li, Qing"],"citation_date":["2025/03/30"],"citation_online_date":["2025/08/05"],"citation_pdf_url":["https://arxiv.org/pdf/2503.23350"],"citation_title":["A Survey of WebAgents: Towards Next-Generation AI Agents for Web Automation with Large Foundation Models"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"4b214bd8e5a34c0e8ca16a3be4fcf638a59a17ca524276dbef3c6b75f08398bb","submitted_at":"2025-03-30T08:15:44Z","updated_at":"2025-08-05T11:47:41Z","version":4,"versioned_url":"https://arxiv.org/abs/2503.23350v4"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["Abstract","1 Introduction","2.3 AI Agents for Web Automation","5 Trustworthy WebAgents including 5.1–5.3"],"source_receipt_keys":["arxiv-2503.23350.metadata.html","arxiv-2503.23350.full.html"],"supported_context_claims":["The review deliberately concentrates on web automation.","Its trustworthiness discussion includes more than adversarial robustness alone."],"title":"A Survey of WebAgents: Towards Next-Generation AI Agents for Web Automation with Large Foundation Models"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2503.23434","catalog_first_date":"2025-03-30","citation_key":"arxiv_2503_23434","finding":"Reframes trustworthiness through alignment among perception, reasoning and interaction and through failures propagated by action–observation loops; discusses trust-aware evaluation and security/utility trade-offs.","limits":["Version 2 is a February 2026 revision of a March 2025 first submission; earlier search snippets are not used as v2 evidence.","Attack percentages and individual defense effectiveness claims are not carried into the context section."],"metadata_abstract_extracted_sha256":"5a6811c6d0c3b30d4e6af2f7b0beef8aa8f1ad99ceb325182305e571fe5c398f","ordered_authors":["Yucheng Shi","Wenhao Yu","Jingyuan Huang","Wenlin Yao","Wenhu Chen","Ninghao Liu"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 30 Mar 2025 (<a href=\"https://arxiv.org/abs/2503.23434v1\">v1</a>), last revised 24 Feb 2026 (this version, v2)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2503.23434"],"citation_author":["Shi, Yucheng","Yu, Wenhao","Huang, Jingyuan","Yao, Wenlin","Chen, Wenhu","Liu, Ninghao"],"citation_date":["2025/03/30"],"citation_online_date":["2026/02/24"],"citation_pdf_url":["https://arxiv.org/pdf/2503.23434"],"citation_title":["Towards Trustworthy GUI Agents: A Survey"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"5a6811c6d0c3b30d4e6af2f7b0beef8aa8f1ad99ceb325182305e571fe5c398f","submitted_at":"2025-03-30T13:26:00Z","updated_at":"2026-02-24T05:24:23Z","version":2,"versioned_url":"https://arxiv.org/abs/2503.23434v2"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["1 Introduction","2 Foundations including 2.1–2.3","3 Perception Trust","6 Evaluation Methodologies","7 Conclusion","Limitations"],"source_receipt_keys":["arxiv-2503.23434.metadata.html","arxiv-2503.23434.full.html"],"supported_context_claims":["Successful task completion and trustworthy behavior need distinct evidence.","Some qualitative trade-off and future-direction claims are explicitly qualified in the survey."],"title":"Towards Trustworthy GUI Agents: A Survey"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2504.13865","catalog_first_date":"2025-03-27","citation_key":"arxiv_2504_13865","finding":"GUI-agent survey; abstract organizes perception, exploration, planning and interaction, with evaluation and safety.","limits":["No full-text claim is based on this inspection."],"metadata_abstract_extracted_sha256":"216d0605205cfd7f17ffd7d7f4c9752577ca427f370df42faf4a2250b80bcc39","ordered_authors":["Fei Tang","Haolei Xu","Hang Zhang","Siqi Chen","Xingyu Wu","Yongliang Shen","Wenqi Zhang","Guiyang Hou","Zeqi Tan","Yuchen Yan","Kaitao Song","Jian Shao","Weiming Lu","Jun Xiao","Yueting Zhuang"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 27 Mar 2025 (<a href=\"https://arxiv.org/abs/2504.13865v1\">v1</a>), last revised 4 Jun 2025 (this version, v2)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2504.13865"],"citation_author":["Tang, Fei","Xu, Haolei","Zhang, Hang","Chen, Siqi","Wu, Xingyu","Shen, Yongliang","Zhang, Wenqi","Hou, Guiyang","Tan, Zeqi","Yan, Yuchen","Song, Kaitao","Shao, Jian","Lu, Weiming","Xiao, Jun","Zhuang, Yueting"],"citation_date":["2025/03/27"],"citation_online_date":["2025/06/04"],"citation_pdf_url":["https://arxiv.org/pdf/2504.13865"],"citation_title":["A Survey on (M)LLM-Based GUI Agents"]},"publication_metadata":null,"read_status":"primary_metadata_and_complete_abstract_only","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"216d0605205cfd7f17ffd7d7f4c9752577ca427f370df42faf4a2250b80bcc39","submitted_at":"2025-03-27T17:58:31Z","updated_at":"2025-06-04T17:29:13Z","version":2,"versioned_url":"https://arxiv.org/abs/2504.13865v2"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["primary abstract-page title, ordered authors, dates, version and full abstract"],"source_receipt_keys":["arxiv-2504.13865.metadata.html"],"supported_context_claims":["Inventory screening only; not a deep comparison row."],"title":"A Survey on (M)LLM-Based GUI Agents"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2504.19838","catalog_first_date":"2025-04-28","citation_key":"arxiv_2504_19838","finding":"Phone-centered taxonomy covers framework choices, prompting/training and datasets/evaluation while explicitly acknowledging cross-platform foundations; discusses on-device resource limits and personalization.","limits":["The comparison reports the documented organizing focus, not the performance rankings in section 6.3.","Location information in 4.1.2 is described as a potential input, not an implemented demonstrated capability."],"metadata_abstract_extracted_sha256":"d65d8aa2e7d77ea52219118de1a8e594206408e4e1af5c6e3964931ec3528282","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 28 Apr 2025 (<a href=\"https://arxiv.org/abs/2504.19838v1\">v1</a>), last revised 17 Nov 2025 (this version, v3)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2504.19838"],"citation_author":["Liu, Guangyi","Zhao, Pengxiang","Liang, Yaozhen","Liu, Liang","Guo, Yaxuan","Xiao, Han","Lin, Weifeng","Chai, Yuxiang","Han, Yue","Ren, Shuai","Wang, Hao","Liang, Xiaoyu","Wang, WenHao","Wu, Tianze","Lu, Zhengxi","Chen, Siheng","LiLinghao","Wang, Hao","Xiong, Guanjing","Liu, Yong","Li, Hongsheng"],"citation_date":["2025/04/28"],"citation_online_date":["2025/11/17"],"citation_pdf_url":["https://arxiv.org/pdf/2504.19838"],"citation_title":["LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"9103197fa1274fac79ae9e4472f5fbdc9a210548afb024f1d4259c6bb0131bf2","submitted_at":"2025-04-28T14:39:25Z","updated_at":"2025-11-17T15:51:58Z","version":3,"versioned_url":"https://arxiv.org/abs/2504.19838v3"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["1 Introduction including scope paragraph","4 introductory framework/POMDP description","4.1.2 Phone State Perception","7 Challenges and Future Directions","8 Conclusion"],"source_receipt_keys":["arxiv-2504.19838.metadata.html","arxiv-2504.19838.full.html"],"supported_context_claims":["Phone interactions and deployment constraints motivate a platform-focused survey.","Cross-platform papers are not automatically outside this survey’s stated scope."],"title":"LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2504.20464","catalog_first_date":"2025-04-29","citation_key":"arxiv_2504_20464","finding":"GUI-agent survey with reinforcement-learning emphasis; abstract also covers prompting and supervised fine-tuning.","limits":["No full-text claim is based on this inspection."],"metadata_abstract_extracted_sha256":"5040073e7c278734df219d71d94e236468468621f5971ef3ccbc97b94e1792b9","ordered_authors":["Jiahao Li","Kaer Huang"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 29 Apr 2025 (<a href=\"https://arxiv.org/abs/2504.20464v1\">v1</a>), last revised 13 May 2025 (this version, v2)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2504.20464"],"citation_author":["Li, Jiahao","Huang, Kaer"],"citation_date":["2025/04/29"],"citation_online_date":["2025/05/13"],"citation_pdf_url":["https://arxiv.org/pdf/2504.20464"],"citation_title":["A Survey on GUI Agents with Foundation Models Enhanced by Reinforcement Learning"]},"publication_metadata":null,"read_status":"primary_metadata_and_complete_abstract_only","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"5040073e7c278734df219d71d94e236468468621f5971ef3ccbc97b94e1792b9","submitted_at":"2025-04-29T06:55:15Z","updated_at":"2025-05-13T01:59:29Z","version":2,"versioned_url":"https://arxiv.org/abs/2504.20464v2"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["primary abstract-page title, ordered authors, dates, version and full abstract"],"source_receipt_keys":["arxiv-2504.20464.metadata.html"],"supported_context_claims":["Inventory screening only; not a deep comparison row."],"title":"A Survey on GUI Agents with Foundation Models Enhanced by Reinforcement Learning"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2505.10924","catalog_first_date":"2025-05-16","citation_key":"arxiv_2505_10924","finding":"Safety/security systematization of computer-use agents; abstract covers threats, defenses and evaluation.","limits":["No full-text claim is based on this inspection."],"metadata_abstract_extracted_sha256":"e8a0890cd57809ff5e3bede6dc2b851b42fad9307a5bfa0ddd744cc2ac70d0c2","ordered_authors":["Ada Chen","Yongjiang Wu","Junyuan Zhang","Jingyu Xiao","Shu Yang","Jen-tse Huang","Kun Wang","Wenxuan Wang","Shuai Wang"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 16 May 2025 (<a href=\"https://arxiv.org/abs/2505.10924v1\">v1</a>), last revised 29 Apr 2026 (this version, v4)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2505.10924"],"citation_author":["Chen, Ada","Wu, Yongjiang","Zhang, Junyuan","Xiao, Jingyu","Yang, Shu","Huang, Jen-tse","Wang, Kun","Wang, Wenxuan","Wang, Shuai"],"citation_date":["2025/05/16"],"citation_online_date":["2026/04/29"],"citation_pdf_url":["https://arxiv.org/pdf/2505.10924"],"citation_title":["A Survey on the Safety and Security Threats of Computer-Using Agents: JARVIS or Ultron?"]},"publication_metadata":null,"read_status":"primary_metadata_and_complete_abstract_only","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"e8a0890cd57809ff5e3bede6dc2b851b42fad9307a5bfa0ddd744cc2ac70d0c2","submitted_at":"2025-05-16T06:56:42Z","updated_at":"2026-04-29T05:56:22Z","version":4,"versioned_url":"https://arxiv.org/abs/2505.10924v4"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["primary abstract-page title, ordered authors, dates, version and full abstract"],"source_receipt_keys":["arxiv-2505.10924.metadata.html"],"supported_context_claims":["Inventory screening only; not a deep comparison row."],"title":"A Survey on the Safety and Security Threats of Computer-Using Agents: JARVIS or Ultron?"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2508.04482","catalog_first_date":"2025-08-06","citation_key":"arxiv_2508_04482","finding":"Organizes OS agents around environments, observation/action spaces, understanding/planning/grounding, foundation-model construction and agent frameworks; spans desktop, mobile and web.","limits":["The title is not evidence of a separate benchmark or deployed operating system.","The survey’s illustrative commercial claims are not independently verified or adopted here."],"metadata_abstract_extracted_sha256":"f14fdd895e52277a3695eeb83e5095472051d139aec63f22f9c63d344c051dad","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"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 6 Aug 2025]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2508.04482"],"citation_author":["Hu, Xueyu","Xiong, Tao","Yi, Biao","Wei, Zishu","Xiao, Ruixuan","Chen, Yurun","Ye, Jiasheng","Tao, Meiling","Zhou, Xiangxin","Zhao, Ziyu","Li, Yuhuai","Xu, Shengze","Wang, Shenzhi","Xu, Xinchen","Qiao, Shuofei","Wang, Zhaokai","Kuang, Kun","Zeng, Tieyong","Wang, Liang","Li, Jiwei","Jiang, Yuchen Eleanor","Zhou, Wangchunshu","Wang, Guoyin","Yin, Keting","Zhao, Zhou","Yang, Hongxia","Wu, Fan","Zhang, Shengyu","Wu, Fei"],"citation_date":["2025/08/06"],"citation_online_date":["2025/08/06"],"citation_pdf_url":["https://arxiv.org/pdf/2508.04482"],"citation_title":["OS Agents: A Survey on MLLM-based Agents for General Computing Devices Use"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"f14fdd895e52277a3695eeb83e5095472051d139aec63f22f9c63d344c051dad","submitted_at":"2025-08-06T14:33:45Z","updated_at":"2025-08-06T14:33:45Z","version":1,"versioned_url":"https://arxiv.org/abs/2508.04482v1"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["1 Introduction","2 Fundamentals including 2.1–2.2","5 Challenges and Future","6 Related Work","7 Conclusion"],"source_receipt_keys":["arxiv-2508.04482.metadata.html","arxiv-2508.04482.full.html"],"supported_context_claims":["An OS-agent perspective is broader than desktop-only automation.","Safety, privacy, personalization and self-evolution are explicit prior-survey topics."],"title":"OS Agents: A Survey on MLLM-based Agents for General Computing Devices Use"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2602.12430","catalog_first_date":"2026-02-12","citation_key":"arxiv_2602_12430","finding":"Broader agent-skills survey; abstract explicitly discusses the CUA stack among deployment topics, but is not a GUI-only survey.","limits":["No full-text claim is based on this inspection."],"metadata_abstract_extracted_sha256":"939c73ef9563bce0630161b5d2870c1fde6cd23a4fdeda0a6596d55fb430606d","ordered_authors":["Renjun Xu","Yang Yan"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 12 Feb 2026 (<a href=\"https://arxiv.org/abs/2602.12430v1\">v1</a>), last revised 2 Jun 2026 (this version, v4)]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2602.12430"],"citation_author":["Xu, Renjun","Yan, Yang"],"citation_date":["2026/02/12"],"citation_online_date":["2026/06/02"],"citation_pdf_url":["https://arxiv.org/pdf/2602.12430"],"citation_title":["Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward"]},"publication_metadata":null,"read_status":"primary_metadata_and_complete_abstract_only","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"f0edc70265a8d46a7c22a9c6384de8aa19f79dbb6987f8c23f4a260647a6a3f0","submitted_at":"2026-02-12T21:33:25Z","updated_at":"2026-06-02T16:14:54Z","version":4,"versioned_url":"https://arxiv.org/abs/2602.12430v4"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["primary abstract-page title, ordered authors, dates, version and full abstract"],"source_receipt_keys":["arxiv-2602.12430.metadata.html"],"supported_context_claims":["Inventory screening only; not a deep comparison row."],"title":"Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward"},{"abstract_hash_note":"Hash of extracted primary-HTML text; links and formula rendering can differ from normalized Atom. It is not asserted to equal the retained Atom abstract hash.","canonical_id":"arxiv:2609.02309","catalog_first_date":"2026-09-02","citation_key":"arxiv_2609_02309","finding":"Defines efficiency across observations, context/memory, actions and planner/runtime and emphasizes that added parsing, retrieval, verification and orchestration may offset local savings.","limits":["Its literature-expansion protocol describes that paper, not the method or counts of this manuscript.","No reported numerical efficiency gains are independently reproduced or ranked."],"metadata_abstract_extracted_sha256":"ec439fd2cfdf7319d77ef5ab58e50143bceb3f8205ccffa0d83555f10dece42a","ordered_authors":["Bizhe Bai","Jiakang Yuan","Hongming Wu","Xinyue Wang","Jie Ren","Siyao Chen","Yuchen Ya","Fan Bai","Pai Peng","Huafeng Qin","Tao Chen"],"primary_identity_check":{"exact_requested_version_in_primary_page":true,"ordered_author_match_after_citation_name_order_conversion":true,"primary_submission_history_html":"[Submitted on 2 Sep 2026]","revision_date_match":true,"submitted_date_match":true,"title_match_after_whitespace_case_normalization":true},"primary_metadata_literal":{"citation_arxiv_id":["2609.02309"],"citation_author":["Bai, Bizhe","Yuan, Jiakang","Wu, Hongming","Wang, Xinyue","Ren, Jie","Chen, Siyao","Ya, Yuchen","Bai, Fan","Peng, Pai","Qin, Huafeng","Chen, Tao"],"citation_date":["2026/09/02"],"citation_online_date":["2026/09/02"],"citation_pdf_url":["https://arxiv.org/pdf/2609.02309"],"citation_title":["Efficient GUI Agents: A Systems Survey of Observation, Memory, Action, and Runtime Optimization"]},"publication_metadata":null,"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":{"abstract_hash_method":"sha256-utf8-whitespace-collapsed-v1","abstract_sha256":"ec439fd2cfdf7319d77ef5ab58e50143bceb3f8205ccffa0d83555f10dece42a","submitted_at":"2026-09-02T08:54:58Z","updated_at":"2026-09-02T08:54:58Z","version":1,"versioned_url":"https://arxiv.org/abs/2609.02309v1"},"reviewed_at":"2026-10-18T10:22:08Z","sections":["1 Introduction","2 Preliminaries including 2.1–2.4","6 Planner-Side and System Efficiency","7 Open Challenges","Appendix B Literature Expansion Protocol"],"source_receipt_keys":["arxiv-2609.02309.metadata.html","arxiv-2609.02309.full.html"],"supported_context_claims":["End-to-end cost accounting is a distinct documented survey emphasis.","Its search protocol is targeted expansion and citation chaining."],"title":"Efficient GUI Agents: A Systems Survey of Observation, Memory, Action, and Runtime Optimization"},{"canonical_id":"url:392a602c660ffa91be46","catalog_first_date":null,"citation_key":"cua_6b7f9d608ed89d1625610eb1","finding":"SUGILITE combines speech instructions, user demonstrations and Android accessibility hierarchies to generalize task procedures and checks context during replay.","limits":["Generalization treats identifying features as strings and requires exact matches in the described implementation.","The implementation relies on accessible UI structure; the paper acknowledges technical exceptions.","No general-purpose language-model capability is claimed."],"ordered_authors":["Toby Jia-Jun Li","Amos Azaria","Brad A. Myers"],"primary_identity_check":{"date_precision_preserved":[2017],"notes":"Only established year is used in historical prose; conference intervals are not converted into publication dates.","status":"manually_checked_in_listed_primary_fields_and_PDF_title_pages"},"publication_metadata":{"authors":["Toby Jia-Jun Li","Amos Azaria","Brad A. Myers"],"date_basis":"Supplied year precision: 2017. Publications: SUGILITE, CHI 2017 Only publication year is asserted; event dates are kept separate.","date_parts":[2017],"doi":"10.1145/3025453.3025483","event_dates":null,"evidence_urls":["https://www.cs.cmu.edu/~NatProg/sugilite.html","https://toby.li/files/TobyLi-CHI2017-Sugilite.pdf"],"identity_basis":"No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.","issue":null,"notes":"Related to later APPINITE and PUMICE work but distinguished here by its 2017 three-author publication and base demonstration-automation contribution. Publication month/day are not independently identified by the retained sources.","pages":"6038-6049","pdf_url":"https://toby.li/files/TobyLi-CHI2017-Sugilite.pdf","publisher":null,"series":null,"title":"SUGILITE: Creating Multimodal Smartphone Automation by Demonstration","type":"conference-paper","url":"https://doi.org/10.1145/3025453.3025483","venue":"Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems","volume":null},"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":null,"reviewed_at":"2026-10-18T10:22:08Z","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"],"source_receipt_keys":["url-392a602c660ffa91be46.metadata.html","url-392a602c660ffa91be46.full.pdf"],"supported_context_claims":["Mobile demonstration-based automation can include cross-application procedures and explicit error handling."],"title":"SUGILITE: Creating Multimodal Smartphone Automation by Demonstration"},{"canonical_id":"url:52a3261eb6a19b402e88","catalog_first_date":"2017-07-17","citation_key":"cua_ba263493b78b0263c7933138","finding":"World of Bits exposes pixels/DOM and low-level input, combines controlled MiniWoB tasks with cached website approximations, and uses captured HTTP traffic for repeatable interaction.","limits":["Cached FormWoB terminates with zero reward on unrecorded requests; it is not unlimited live-web execution.","Publisher title/authors/date are checked separately from PDF nickname variants and conference dates."],"ordered_authors":["Tianlin Shi","Andrej Karpathy","Linxi Fan","Jonathan Hernandez","Percy Liang"],"primary_identity_check":{"date_precision_preserved":[2017,7,17],"notes":"Only established year is used in historical prose; conference intervals are not converted into publication dates.","status":"manually_checked_in_listed_primary_fields_and_PDF_title_pages"},"publication_metadata":{"authors":["Tianlin Shi","Andrej Karpathy","Linxi Fan","Jonathan Hernandez","Percy Liang"],"date_basis":"Publisher citation_publication_date metadata explicitly states 2017/07/17; this is separate from the conference interval.","date_parts":[2017,7,17],"doi":null,"event_dates":"06--11 Aug (conference interval reported in the publisher BibTeX month field)","evidence_urls":["https://proceedings.mlr.press/v70/shi17a.html","https://proceedings.mlr.press/v70/shi17a/shi17a.pdf"],"identity_basis":"No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.","issue":null,"notes":"Publisher citation uses Tianlin Shi and Linxi Fan; the PDF also prints the nicknames Tim and Jim. The ordered publisher author names are retained. DOI not supplied in the retained publisher citation. Publisher metadata supplies the publication date separately from the conference interval.","pages":"3135-3144","pdf_url":"https://proceedings.mlr.press/v70/shi17a/shi17a.pdf","publisher":"PMLR","series":"Proceedings of Machine Learning Research","title":"World of Bits: An Open-Domain Platform for Web-Based Agents","type":"conference-paper","url":"https://proceedings.mlr.press/v70/shi17a.html","venue":"Proceedings of the 34th International Conference on Machine Learning","volume":"70"},"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":null,"reviewed_at":"2026-10-18T10:22:08Z","sections":["Publisher metadata","PDF abstract and 1 Introduction","2.1 Web as an Environment","2.2 MiniWoB","2.3 FormWoB"],"source_receipt_keys":["url-52a3261eb6a19b402e88.metadata.html","url-52a3261eb6a19b402e88.full.pdf"],"supported_context_claims":["Reproducible software interaction was a learning-environment concern before LLM agents.","MiniWoB belongs to this original platform paper, while MiniWoB++ is a later extension."],"title":"World of Bits: An Open-Domain Platform for Web-Based Agents"},{"canonical_id":"url:a4cdb2ee0d07d3980651","catalog_first_date":null,"citation_key":"cua_510831381c08385eb74fb95e","finding":"The primary abstract describes screenshot queries for help and screenshot-pattern scripts for mouse/keyboard automation; the author page links the exact UIST 2009 paper.","limits":["NOT a full-text inspection: the author-linked PDF returned HTTP 500 and was not used.","No experimental effect size, implementation detail or additional platform conclusion is inferred."],"ordered_authors":["Tom Yeh","Tsung-Hsiang Chang","Robert C. Miller"],"primary_identity_check":{"date_precision_preserved":[2009,10],"notes":"Only established year is used in historical prose; conference intervals are not converted into publication dates.","status":"manually_checked_in_listed_primary_fields_and_PDF_title_pages"},"publication_metadata":{"authors":["Tom Yeh","Tsung-Hsiang Chang","Robert C. Miller"],"date_basis":"Supplied month precision: 2009-10. Date Issued: October 2009","date_parts":[2009,10],"doi":"10.1145/1622176.1622213","event_dates":null,"evidence_urls":["https://dspace.mit.edu/entities/publication/604a3f79-451d-4186-a94e-d12690ffe40a"],"identity_basis":"No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.","issue":null,"notes":"Institutional abstract and published-version citation supply the retained evidence; the institution identifies an author final manuscript record. Exact publication day is not supplied. Operating-system-specific experiments are not asserted.","pages":"183-192","pdf_url":null,"publisher":null,"series":null,"title":"Sikuli: Using GUI screenshots for search and automation","type":"conference-paper","url":"https://doi.org/10.1145/1622176.1622213","venue":"Proceedings of the 22nd Annual ACM Symposium on User Interface Software and Technology (UIST '09)","volume":null},"read_status":"abstract_and_author_project_only","retained_arxiv_snapshot":null,"reviewed_at":"2026-10-18T10:22:08Z","sections":["Complete MIT institutional abstract and issued-date/authors/published citation","Author project page: Sikuli Script and Sikuli Search"],"source_receipt_keys":["url-a4cdb2ee0d07d3980651.metadata.html","sikuli-author-project.html","sikuli-author-full.pdf"],"supported_context_claims":["The narrow historical description of screenshot-based search and scripting."],"title":"Sikuli: Using GUI screenshots for search and automation"},{"canonical_id":"url:b560b5d6abaab98483f6","catalog_first_date":"2017-10-20","citation_key":"cua_402ed654426712089390317c","finding":"Rico combines crowdsourcing and automated Android exploration and exposes screenshots, view hierarchies, interaction traces and layout representations for design applications.","limits":["Original-paper and project-release counts differ in retained catalog evidence; no merged size is used.","The paper explicitly notes that app data becomes outdated without renewed crawling."],"ordered_authors":["Biplab Deka","Zifeng Huang","Chad Franzen","Joshua Hibschman","Daniel Afergan","Yang Li","Jeffrey Nichols","Ranjitha Kumar"],"primary_identity_check":{"date_precision_preserved":[2017,10,20],"notes":"Only established year is used in historical prose; conference intervals are not converted into publication dates.","status":"manually_checked_in_listed_primary_fields_and_PDF_title_pages"},"publication_metadata":{"authors":["Biplab Deka","Zifeng Huang","Chad Franzen","Joshua Hibschman","Daniel Afergan","Yang Li","Jeffrey Nichols","Ranjitha Kumar"],"date_basis":"Supplied day precision: 2017-10-20. PDF cover, Published: 20 October 2017 The 18 January 2026 download date and October 22-25 conference dates are not publication dates.","date_parts":[2017,10,20],"doi":"10.1145/3126594.3126651","event_dates":null,"evidence_urls":["https://danafergan.com/publications/deka2017rico.pdf","https://dev.interactionmining.org/archive/rico"],"identity_basis":"No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.","issue":null,"notes":"The paper is a distinct resource from the already cataloged Rico dataset (url:0b134b082a7bd507fc6a). The publisher PDF describes 9.7k apps and 72k screens; the project page describes 9.3k apps and 66k screens. The annotation intentionally supplies no merged size claim. No volume number is supplied for this proceedings article.","pages":"845-854","pdf_url":"https://danafergan.com/publications/deka2017rico.pdf","publisher":null,"series":null,"title":"Rico: A Mobile App Dataset for Building Data-Driven Design Applications","type":"conference-paper","url":"https://doi.org/10.1145/3126594.3126651","venue":"UIST '17: Proceedings of the 30th Annual ACM Symposium on User Interface Software and Technology","volume":null},"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":null,"reviewed_at":"2026-10-18T10:22:08Z","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"],"source_receipt_keys":["url-b560b5d6abaab98483f6.metadata.pdf","url-b560b5d6abaab98483f6.full.pdf"],"supported_context_claims":["An interface corpus is an enabling resource rather than proof of autonomous task performance.","The collected data and later distributions should retain their own identities."],"title":"Rico: A Mobile App Dataset for Building Data-Driven Design Applications"},{"canonical_id":"url:cccd967aa993404d74d0","catalog_first_date":null,"citation_key":"cua_1b4a3a269ca9f1884a566476","finding":"Records browser demonstrations as readable instructions, interprets them against page elements, personalizes script values and shares procedures in a wiki; the example permits correction and manual steps.","limits":["The paper presents initial experience; extensive real-world user studies are listed as future work.","Koala is not merged with every later CoScripter paper.","Only year 2007 is used; the institutional metadata date and conference interval remain separate."],"ordered_authors":["Greg Little","Tessa A. Lau","Allen Cypher","James Lin","Eben M. Haber","Eser Kandogan"],"primary_identity_check":{"date_precision_preserved":[2007],"notes":"Only established year is used in historical prose; conference intervals are not converted into publication dates.","status":"manually_checked_in_listed_primary_fields_and_PDF_title_pages"},"publication_metadata":{"authors":["Greg Little","Tessa A. Lau","Allen Cypher","James Lin","Eben M. Haber","Eser Kandogan"],"date_basis":"Supplied year precision: 2007. CHI 2007 proceedings heading and copyright year The IBM page has an unlabeled Date of 22 Oct 2007, while the paper gives April/May event dates. Their roles do not establish a reliable day-level publication date.","date_parts":[2007],"doi":"10.1145/1240624.1240767","event_dates":null,"evidence_urls":["https://research.ibm.com/publications/koala-capture-share-automate-personalize-business-processes-on-the-web","https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf"],"identity_basis":"No same-work record located in the full active catalog. Similarity screening was not used as automatic identity approval. The raw archive and exclusions are additionally checked during normalization.","issue":null,"notes":"This proposal is the specific Koala paper. It does not automatically merge every later CoScripter publication or project record. Exact publication date is unresolved; only the clearly established year is proposed.","pages":"943-946","pdf_url":"https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf","publisher":null,"series":null,"title":"Koala: Capture, Share, Automate, Personalize Business Processes on the Web","type":"conference-paper","url":"https://doi.org/10.1145/1240624.1240767","venue":"CHI 2007 Proceedings","volume":null},"read_status":"selected_fulltext_sections_read","retained_arxiv_snapshot":null,"reviewed_at":"2026-10-18T10:22:08Z","sections":["PDF printed pp. 943–946: abstract, introduction, office-products use case, interpreter, personalization, wiki, conclusion"],"source_receipt_keys":["url-cccd967aa993404d74d0.metadata.html","url-cccd967aa993404d74d0.full.pdf"],"supported_context_claims":["Editable shared workflows and mixed-initiative execution predate foundation models."],"title":"Koala: Capture, Share, Automate, Personalize Business Processes on the Web"}],"retained_sources":[{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":41977,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2411.02006v3","id":"arxiv:2411.02006","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2411.02006.metadata.html","request_attempts":1,"retrieved_at":"2026-10-17T09:13:27Z","reuse_verified_at":"2026-10-18T10:07:41Z","sha256":"bda47fcb119c581a798dbc791b4929af91832d5053ac073a02dab8fd4e10ace3","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2411.02006v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":42586,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2411.04890v2","id":"arxiv:2411.04890","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2411.04890.metadata.html","retrieved_at":"2026-10-18T10:07:44Z","sha256":"422c0e849422eb2c974eb8df661ffdc3b64af141b07a92274605a1baa8ce84ae","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2411.04890v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":42602,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2411.10943v1","id":"arxiv:2411.10943","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2411.10943.metadata.html","retrieved_at":"2026-10-18T10:07:48Z","sha256":"c4eddda9dd616bc09ad0fa9b884bed7ba96a06014ac0dfae21ce876363156113","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2411.10943v1"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":47068,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2411.18279v12","id":"arxiv:2411.18279","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2411.18279.metadata.html","retrieved_at":"2026-10-18T10:07:52Z","sha256":"abf172a70f8d71624a484a27de7f18c74a4ab526df393f62ecaa8691502fe0dd","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2411.18279v12"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":44814,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2412.13501v3","id":"arxiv:2412.13501","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2412.13501.metadata.html","retrieved_at":"2026-10-18T10:07:56Z","sha256":"bb3afe587b0ef55b9974fd3e8944f4d5edb878f17e634093045e3b8bde5f138d","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2412.13501v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":47565,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2501.16150v3","id":"arxiv:2501.16150","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2501.16150.metadata.html","retrieved_at":"2026-10-18T10:08:02Z","sha256":"e9ca21daad357fee1442c6a83eecaa1c907255bdbd7fe4606c49d56281d94f44","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2501.16150v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":44715,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2503.23350v4","id":"arxiv:2503.23350","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2503.23350.metadata.html","retrieved_at":"2026-10-18T10:08:06Z","sha256":"7952424ff757a731195e8633701da2befb7fe329b0765f2497e695168a597b17","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2503.23350v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":42427,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2503.23434v2","id":"arxiv:2503.23434","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2503.23434.metadata.html","retrieved_at":"2026-10-18T10:08:10Z","sha256":"e2a4ec9fdc16c504818fdad8d6608c3c7668675b13529a7952650696fa607ced","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2503.23434v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":45506,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2504.13865v2","id":"arxiv:2504.13865","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2504.13865.metadata.html","retrieved_at":"2026-10-18T10:08:13Z","sha256":"6d983774e185115d6cb8e8ca83cdf68b7d480fc2eb9c8b357b95644e3517ddd9","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2504.13865v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":45684,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2504.19838v3","id":"arxiv:2504.19838","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2504.19838.metadata.html","retrieved_at":"2026-10-18T10:08:17Z","sha256":"5fc92d139141c12f25462b8b0ccb125d521e67c5b8b54873dfa16e633d732f6b","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2504.19838v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":41096,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2504.20464v2","id":"arxiv:2504.20464","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2504.20464.metadata.html","retrieved_at":"2026-10-18T10:08:21Z","sha256":"ef141b387db5f75ddbd9a653ff51c09c6fee1aeb45617118c8237bd136494418","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2504.20464v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":44060,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2505.10924v4","id":"arxiv:2505.10924","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2505.10924.metadata.html","retrieved_at":"2026-10-18T10:08:24Z","sha256":"c4ea2ae29157d7afc502782ef60e17db2a6acf08e2011445f0f3e887d761d27e","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2505.10924v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":46987,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2508.04482v1","id":"arxiv:2508.04482","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2508.04482.metadata.html","retrieved_at":"2026-10-18T10:08:28Z","sha256":"dd508a6ad544585c40f2c84fe79257408d5a6299d5b5f656e787362f48c26265","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2508.04482v1"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":44641,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2602.12430v4","id":"arxiv:2602.12430","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2602.12430.metadata.html","retrieved_at":"2026-10-18T10:08:31Z","sha256":"c5bc7f9f0930296e6d97633ce79d05265f80b642c7363f204e861a31967ffe8f","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2602.12430v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":43134,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2609.02309v1","id":"arxiv:2609.02309","method":"https_get","ok":true,"purpose":"survey_metadata","read_sections":["title","ordered authors","submission history","version","complete abstract"],"reading_status":"primary_metadata_and_complete_abstract_read","receipt_key":"arxiv-2609.02309.metadata.html","retrieved_at":"2026-10-18T10:08:35Z","sha256":"64080b2f1f34edcac11eef55855e43e0175df31b2bbb14427a0b3d57ad4e8fca","status":200,"supports":["identity","dates","abstract_scope"],"url":"https://arxiv.org/abs/2609.02309v1"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":134193,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2411.04890v2","id":"arxiv:2411.04890","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","read_sections":["Abstract","1 Introduction","2 GUI Agent Data Resources","3.1 GUI Agent Framework","3.2 modality and learning taxonomy"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2411.04890.full.html","retrieved_at":"2026-10-18T10:08:38Z","sha256":"8d5145cda2485311be2b6d937715efdc33df3d53bb0de995a6f411d1efd43da2","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2411.04890v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":2106269,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2411.18279v12","id":"arxiv:2411.18279","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","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"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2411.18279.full.html","retrieved_at":"2026-10-18T10:08:43Z","sha256":"e451504bec33942bcc232dc0879c44e5502158d53ae3040d5d20d85160f90a33","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2411.18279v12"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":1214626,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2501.16150v3","id":"arxiv:2501.16150","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","read_sections":["1 Introduction","1.1 Relation to Other Surveys","2.1 Definitions","2.2 A Comprehensive Taxonomy","7 Agent Evaluation","8 Conclusions"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2501.16150.full.html","retrieved_at":"2026-10-18T10:08:58Z","sha256":"2684d4124758b38973f20aa01aaee1040a03513f8cda2fe7cc8a8df1e9e200a1","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2501.16150v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":360892,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2503.23350v4","id":"arxiv:2503.23350","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","read_sections":["Abstract","1 Introduction","2.3 AI Agents for Web Automation","5 Trustworthy WebAgents including 5.1–5.3"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2503.23350.full.html","retrieved_at":"2026-10-18T10:09:04Z","sha256":"56200d817811c55e8d49edfe6da671c24dda0965efe891bb8109fbff258cc729","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2503.23350v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":1246332,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2504.19838v3","id":"arxiv:2504.19838","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","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"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2504.19838.full.html","retrieved_at":"2026-10-18T10:09:10Z","sha256":"a5d1d3f46469093b249db4a5af12a3ccb628b03348d43ef37f7d1254cbf6d476","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2504.19838v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":376575,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2508.04482v1","id":"arxiv:2508.04482","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","read_sections":["1 Introduction","2 Fundamentals including 2.1–2.2","5 Challenges and Future","6 Related Work","7 Conclusion"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2508.04482.full.html","retrieved_at":"2026-10-18T10:09:26Z","sha256":"03855e424943d708cd568e89b55873a4e6b19421f7395e4d2cc202c9fbfdfa65","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2508.04482v1"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":218320,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2503.23434v2","id":"arxiv:2503.23434","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","read_sections":["1 Introduction","2 Foundations including 2.1–2.3","3 Perception Trust","6 Evaluation Methodologies","7 Conclusion","Limitations"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2503.23434.full.html","retrieved_at":"2026-10-18T10:09:32Z","sha256":"f5d92f991e1560995e2cd1a2ff6ad6bf0e62c55ff8d653b22f64f479236b1dd0","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2503.23434v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":527500,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2609.02309v1","id":"arxiv:2609.02309","method":"https_get","ok":true,"purpose":"deep_prior_survey_fulltext","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"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2609.02309.full.html","retrieved_at":"2026-10-18T10:09:36Z","sha256":"09857589286579b14aeeee4a41653aa9e3b0f5b70f70e14f35cb2c1df4398a0a","status":200,"supports":["context_claims","prior_survey_focus"],"url":"https://arxiv.org/html/2609.02309v1"},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":40020,"content_type":"text/html; charset=utf-8","final_url":"https://research.ibm.com/publications/koala-capture-share-automate-personalize-business-processes-on-the-web","id":"url:cccd967aa993404d74d0","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"url-cccd967aa993404d74d0.metadata.html","retrieved_at":"2026-10-17T09:53:23Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"ae61a94cca68e058858244b6ff54ed9409973a5cecaf25c0e325b27adaf0e3dd","source_id":"koala-ibm","status":200,"supports":["identity","dates"],"url":"https://research.ibm.com/publications/koala-capture-share-automate-personalize-business-processes-on-the-web","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":1013375,"content_type":"application/pdf","final_url":"https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf","first_page_render":{"method":"pdftoppm -f 1 -l 1 -scale-to 1200 -singlefile -png","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"530323fcee33226eaf8a0ad26fba21e44fa6a142128170bca733e614af694b15"},"id":"url:cccd967aa993404d74d0","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["PDF printed pp. 943–946: abstract, introduction, office-products use case, interpreter, personalization, wiki, conclusion"],"reading_status":"selected_fulltext_sections_read","receipt_key":"url-cccd967aa993404d74d0.full.pdf","retrieved_at":"2026-10-17T09:55:21Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"944b6f74a6cdb88fc253009f364dcc71cc42186b4907e04720da128c5292e357","source_id":"koala-pdf","status":200,"supports":["historical_contribution","scope_boundary"],"text_extraction":{"method":"pdftotext -layout; visual first-page checks for authored proposals","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"c3eb8a59cbee80b09d92ac2ac030b9ee14aae2f08c5eb12cc6b496fd6c2babd6"},"url":"https://plunk.org/eben/PublishedPapers/KoalaCHI.pdf","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":563747,"content_type":"text/html; charset=utf-8","final_url":"https://dspace.mit.edu/entities/publication/604a3f79-451d-4186-a94e-d12690ffe40a","id":"url:a4cdb2ee0d07d3980651","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["Complete MIT institutional abstract and issued-date/authors/published citation","Author project page: Sikuli Script and Sikuli Search"],"reading_status":"primary_abstract_or_author_project_sections_read","receipt_key":"url-a4cdb2ee0d07d3980651.metadata.html","retrieved_at":"2026-10-17T09:53:21Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"dbaaf21982522b34857be014b39d6c43fedcda20f808efe40d04958c77243a5a","source_id":"sikuli-mit","status":200,"supports":["identity","dates","narrow_historical_description"],"url":"https://dspace.mit.edu/entities/publication/604a3f79-451d-4186-a94e-d12690ffe40a","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":15270,"content_type":"text/html; charset=utf-8","final_url":"https://proceedings.mlr.press/v70/shi17a.html","id":"url:52a3261eb6a19b402e88","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"url-52a3261eb6a19b402e88.metadata.html","retrieved_at":"2026-10-17T09:53:21Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"749d2a150e4c2bcde1b219a0ecb86c95546134a057892544e8495b5e45323c44","source_id":"worldbits-page","status":200,"supports":["identity","dates"],"url":"https://proceedings.mlr.press/v70/shi17a.html","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":3383879,"content_type":"application/pdf","final_url":"https://proceedings.mlr.press/v70/shi17a/shi17a.pdf","first_page_render":{"method":"pdftoppm -f 1 -l 1 -scale-to 1200 -singlefile -png","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"5eb5344e56b53a4c282676931fa03fddcf31e0907b2aaa881ea4ab157b348a5b"},"id":"url:52a3261eb6a19b402e88","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["Publisher metadata","PDF abstract and 1 Introduction","2.1 Web as an Environment","2.2 MiniWoB","2.3 FormWoB"],"reading_status":"selected_fulltext_sections_read","receipt_key":"url-52a3261eb6a19b402e88.full.pdf","retrieved_at":"2026-10-17T09:54:39Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"087008efb9d056df5fd9fbc724413d6a3f20cb14c7f1597d00bacac53d8342d6","source_id":"worldbits-pdf","status":200,"supports":["historical_contribution","scope_boundary"],"text_extraction":{"method":"pdftotext -layout; visual first-page checks for authored proposals","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"31f04ab96ae933b3e6608a4b73d37eefda430ceeff5f33c2089db2ff42fc5382"},"url":"https://proceedings.mlr.press/v70/shi17a/shi17a.pdf","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":1949991,"content_type":"application/pdf","final_url":"https://danafergan.com/publications/deka2017rico.pdf","first_page_render":{"method":"pdftoppm -f 1 -l 1 -scale-to 1200 -singlefile -png","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"66f20562b55719b47d87f06e1b253c3163b040f458c1c087264175a1cc7a096d"},"id":"url:b560b5d6abaab98483f6","method":"https_get","ok":true,"purpose":"foundational_metadata","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"],"reading_status":"selected_fulltext_sections_read_same_bytes_as_full_copy","receipt_key":"url-b560b5d6abaab98483f6.metadata.pdf","retrieved_at":"2026-10-17T10:01:32Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"a1ec58c526b152c052c5c72f9a62725f1026b5f0e4f45c8fe85fce284295da84","source_id":"rico-author-published-pdf","status":200,"supports":["identity","dates","historical_contribution"],"text_extraction":{"method":"pdftotext -layout; visual first-page checks for authored proposals","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"3d4753479ac23e842b8bf7a68434587eefe77bb6e071f890dae9b764504b2b24"},"url":"https://danafergan.com/publications/deka2017rico.pdf","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":1949991,"content_type":"application/pdf","final_url":"https://danafergan.com/publications/deka2017rico.pdf","first_page_render":{"method":"pdftoppm -f 1 -l 1 -scale-to 1200 -singlefile -png","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"66f20562b55719b47d87f06e1b253c3163b040f458c1c087264175a1cc7a096d"},"id":"url:b560b5d6abaab98483f6","method":"https_get","ok":true,"purpose":"foundational_fulltext","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"],"reading_status":"selected_fulltext_sections_read","receipt_key":"url-b560b5d6abaab98483f6.full.pdf","retrieved_at":"2026-10-17T10:01:32Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"a1ec58c526b152c052c5c72f9a62725f1026b5f0e4f45c8fe85fce284295da84","source_id":"rico-author-published-pdf","status":200,"supports":["historical_contribution","scope_boundary"],"text_extraction":{"method":"pdftotext -layout; visual first-page checks for authored proposals","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"3d4753479ac23e842b8bf7a68434587eefe77bb6e071f890dae9b764504b2b24"},"url":"https://danafergan.com/publications/deka2017rico.pdf","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":13368,"content_type":"text/html","final_url":"https://www.cs.cmu.edu/~NatProg/sugilite.html","id":"url:392a602c660ffa91be46","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"url-392a602c660ffa91be46.metadata.html","retrieved_at":"2026-10-17T09:53:22Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"32ee05ff1a91ed4a22d4bc156a462872aa4e28f5c79abfbce5dbf3655f412033","source_id":"sugilite-cmu","status":200,"supports":["identity","dates"],"url":"https://www.cs.cmu.edu/~NatProg/sugilite.html","usable_for_claims":true},{"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":2003273,"content_type":"application/pdf","final_url":"https://toby.li/files/TobyLi-CHI2017-Sugilite.pdf","first_page_render":{"method":"pdftoppm -f 1 -l 1 -scale-to 1200 -singlefile -png","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"1db96ff72ca311632a17c42f7d46960480182328e1bf4318343994ff6807bdb5"},"id":"url:392a602c660ffa91be46","method":"https_get","ok":true,"purpose":"foundational_fulltext","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"],"reading_status":"selected_fulltext_sections_read","receipt_key":"url-392a602c660ffa91be46.full.pdf","retrieved_at":"2026-10-17T09:54:39Z","reuse_verified_at":"2026-10-18T10:10:02Z","review_status":"primary_sections_read","sha256":"cc23c19284f0a558c4b99b8b051abca8ed61e274e9e1f1fed1ae91487f4368ad","source_id":"sugilite-pdf","status":200,"supports":["historical_contribution","scope_boundary"],"text_extraction":{"method":"pdftotext -layout; visual first-page checks for authored proposals","path":"Retained in private review cache; see source URL and fingerprint.","sha256":"8dc7f8b65010b787a07f312643a327bfd8ca391dd3f4b5e84b308f596f4f13b6"},"url":"https://toby.li/files/TobyLi-CHI2017-Sugilite.pdf","usable_for_claims":true},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":43231,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/1802.08802v1","id":"arxiv:1802.08802","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-1802.08802.metadata.html","retrieved_at":"2026-10-18T10:10:05Z","sha256":"df56095550e7d7560632209a4d86ce11e2e4e219df619c20239b739bd28484c7","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/1802.08802v1"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":207313,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/1802.08802v1","id":"arxiv:1802.08802","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["Abstract","1 Introduction","2 Setup","3 Inducing workflows","6.1 Task setups","6.3.1 MiniWoB++"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-1802.08802.full.html","retrieved_at":"2026-10-18T10:10:08Z","sha256":"0bd8bfa1d6c028648e975a20df037c74e10995e02b93aece0ef18f9e05d48e8d","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/1802.08802v1"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":44083,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2207.01206v4","id":"arxiv:2207.01206","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-2207.01206.metadata.html","retrieved_at":"2026-10-18T10:10:13Z","sha256":"4df392cd073a841564fa30bc46bc1dfe9e0078b40bc2aebcc79982fcf682eda1","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/2207.01206v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":331273,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2207.01206v4","id":"arxiv:2207.01206","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["Abstract","1 Introduction","3.1 Task Formulation","3.2 Environment Implementation","Appendix E Potential Societal Impacts and Limitations"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2207.01206.full.html","retrieved_at":"2026-10-18T10:10:17Z","sha256":"2c25234743e842bcc9686d61ef007f6ca26302bd5a7fc0321a97624425c75090","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/2207.01206v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":43850,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2210.03347v2","id":"arxiv:2210.03347","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-2210.03347.metadata.html","retrieved_at":"2026-10-18T10:10:20Z","sha256":"ab98a4f14402b4d2fbccfbfa052eb9a98f505e858f23f30eee97f943532fe688","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/2210.03347v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":347171,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2210.03347v2","id":"arxiv:2210.03347","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["Abstract","1 Introduction","2.2 Architecture","2.3 Pretraining","2.5 Finetuning","3.1 Benchmarks","4.2 UIs","6 Discussion"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2210.03347.full.html","retrieved_at":"2026-10-18T10:10:25Z","sha256":"2b36f4959ea6714746a4db8b8fc13a19d15b1365fc20481f870cd53eeec14bf0","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/2210.03347v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":43113,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2306.06070v3","id":"arxiv:2306.06070","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-2306.06070.metadata.html","retrieved_at":"2026-10-18T10:10:30Z","sha256":"fe6248ffd11553e721f204cdf760e6b56362febcfe1e1cb26fcf1d5142ea2e8a","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/2306.06070v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":261502,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2306.06070v3","id":"arxiv:2306.06070","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["Abstract","1 Introduction","2.1 Task Definition","2.2 Data Collection","3 Method overview","6 Limitations and Potential Societal Impact"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2306.06070.full.html","retrieved_at":"2026-10-18T10:10:34Z","sha256":"92f1e770ec28a36dd71a3381b67fc183efffb623d0803c5d7c11e298a6f2aeec","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/2306.06070v3"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":44835,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2307.13854v4","id":"arxiv:2307.13854","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-2307.13854.metadata.html","retrieved_at":"2026-10-18T10:10:38Z","sha256":"2bdd9cdee9419df180a60202c335104a779d1b7229918f038f0e59780f8b4506","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/2307.13854v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":247063,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2307.13854v4","id":"arxiv:2307.13854","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["Abstract","1 Introduction","2 Environment including 2.1–2.4","3 benchmark overview","3.2 Evaluation Annotation"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2307.13854.full.html","retrieved_at":"2026-10-18T10:10:42Z","sha256":"15c4dcd5968f24bf21f1bdc6696be4180bdeed58f2703345ed7e11da89cedd2d","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/2307.13854v4"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":43436,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2307.10088v2","id":"arxiv:2307.10088","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-2307.10088.metadata.html","retrieved_at":"2026-10-18T10:10:47Z","sha256":"bcc253aebf71fb693401e78ccb26e8ffa1eca42d51840a1708d53c1e2a62d3ad","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/2307.10088v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":426040,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2307.10088v2","id":"arxiv:2307.10088","method":"https_get","ok":true,"purpose":"foundational_fulltext","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"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2307.10088.full.html","retrieved_at":"2026-10-18T10:10:51Z","sha256":"d326d463c42951c379dd541534e6329f4a032e8b815486503b691caf84a963e2","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/2307.10088v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":46129,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2404.07972v2","id":"arxiv:2404.07972","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-2404.07972.metadata.html","retrieved_at":"2026-10-18T10:10:57Z","sha256":"7d2e5e48d63786bc29736a610e60a83b92ef07cd36424079496a60d7aedd685b","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/2404.07972v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":820428,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2404.07972v2","id":"arxiv:2404.07972","method":"https_get","ok":true,"purpose":"foundational_fulltext","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"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2404.07972.full.html","retrieved_at":"2026-10-18T10:11:00Z","sha256":"fbd620dccd72032406ceb36d4d9e108292b1bc7fcbf4c3aee0c0b544601a05c2","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/2404.07972v2"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":44801,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/abs/2405.14573v5","id":"arxiv:2405.14573","method":"https_get","ok":true,"purpose":"foundational_metadata","read_sections":["title","ordered authors where supplied","first-submission/publication date and revision where supplied"],"reading_status":"primary_bibliographic_fields_checked","receipt_key":"arxiv-2405.14573.metadata.html","retrieved_at":"2026-10-18T10:11:03Z","sha256":"84824419227fb39f3540bc7074b7b788aec0c241c6721b60e515ca2e41c60209","status":200,"supports":["identity","dates"],"url":"https://arxiv.org/abs/2405.14573v5"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":600979,"content_type":"text/html; charset=utf-8","final_url":"https://arxiv.org/html/2405.14573v5","id":"arxiv:2405.14573","method":"https_get","ok":true,"purpose":"foundational_fulltext","read_sections":["Abstract","1 Introduction","3.1–3.5 AndroidWorld environment","Appendix A Limitations"],"reading_status":"selected_fulltext_sections_read","receipt_key":"arxiv-2405.14573.full.html","retrieved_at":"2026-10-18T10:11:07Z","sha256":"d3bdc04b692f57e3a933d67ab955fae790c1fe5deb7b5bcb71d80d24b8d2cdf6","status":200,"supports":["historical_contribution","scope_boundary"],"url":"https://arxiv.org/html/2405.14573v5"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":5999,"content_type":"text/html","final_url":"https://people.csail.mit.edu/vgod/projects.shtml","id":"url:a4cdb2ee0d07d3980651","method":"https_get","ok":true,"purpose":"author_paper_link","read_sections":["Complete MIT institutional abstract and issued-date/authors/published citation","Author project page: Sikuli Script and Sikuli Search"],"reading_status":"primary_abstract_or_author_project_sections_read","receipt_key":"sikuli-author-project.html","retrieved_at":"2026-10-18T10:11:51Z","sha256":"06c5dd8c68b6e1b4da97de0fabd575b94969dbfb7072626890e8a9718f489615","status":200,"supports":["identity","dates","narrow_historical_description"],"url":"https://people.csail.mit.edu/vgod/projects.shtml"},{"attempts":1,"body_sha256_reverified_at":"2026-10-18T10:22:08Z","bytes":531,"error":"HTTP Error 500: Internal Server Error","failure_note":"Author-linked Sikuli PDF returned HTTP 500; no full-text claim or fallback bypass.","final_url":"https://groups.csail.mit.edu/uid/","id":"url:a4cdb2ee0d07d3980651","method":"https_get","ok":false,"purpose":"foundational_fulltext_attempt","read_sections":[],"reading_status":"failed_fetch_not_evidence","receipt_key":"sikuli-author-full.pdf","retrieved_at":"2026-10-18T10:11:58Z","sha256":"2965b173e57327584895121d6015b13cdf70d071425bf46902a226d603b53c7d","status":500,"supports":[],"url":"https://uid.csail.mit.edu/projects/sikuli/sikuli-uist2009.pdf"}],"retrieval_constraints":{"cache_reuse":"Existing primary bytes reused only after SHA256 verification; original retrieval timestamp preserved and a separate reuse/read timestamp added.","failed_source":"Sikuli author-linked PDF: actual HTTP 500/531 bytes retained; not evidence of full-text reading. Web-tool link attempt also failed. No repeated fetch or access-control workaround.","new_paper_requests":"Sequential HTTPS GET, >=3.2 seconds between new requests, one attempt per URL in the cached reader. No arXiv API harvest by this agent."},"review_completed_at":"2026-10-18T10:22:08Z","root_search_boundary":"Root agent owns broad arXiv title-query search, complete retrieval totals, additional candidate screening and the main search-method subsection. This file only reports this agent’s targeted queries and reads; its counts must be deduplicated with other sections.","schema_version":1,"scope_definition":"Web/desktop/mobile software-interface action and concrete interface understanding, grounding or evaluation support. Generic QA, robotics or tool calling is not core solely on possible transfer. Hybrid GUI/API work requires a documented software-interface relationship.","section":"context","selection_basis":"Purposeful narrative sample of historical interface automation/data/environment works and complementary prior-survey perspectives. Existing reviewed inventory supplied seeds; targeted fresh primary-source search and exact-version reads checked them. No PRISMA, systematic exhaustiveness or general performance ordering claim.","targeted_search_log":[{"queries":[{"domains":["arxiv.org","aclanthology.org","openreview.net"],"q":"\"GUI agents\" \"survey\" foundation models"},{"domains":["arxiv.org"],"q":"\"computer use\" \"survey\" agents"},{"domains":["arxiv.org"],"q":"\"Efficient GUI Agents\" \"Systems Survey\""}],"query_result_mapping_note":"The tool returned one combined result stream for this batch; per-query hit allocation is not reconstructed.","receipt_sha256":"970e8ce33877db31cde2d4762325494c7c80c6823a06a0fa017e65064524b5cb","results":[{"canonical_id":"arxiv:2412.13501","returned_title_or_snippet_heading":"GUI Agents: A Survey - ACL Anthology","status":"publication_edition_of_existing_work_not_counted_as_new_paper; exact_arxiv_metadata_read","url":"https://aclanthology.org/2025.findings-acl.1158/"},{"canonical_id":"arxiv:2411.04890","returned_title_or_snippet_heading":"GUI Agents with Foundation Models: A Comprehensive Survey","status":"existing_inventory_work_exact_version_primary_page_read","url":"https://arxiv.org/abs/2411.04890"},{"canonical_id":"arxiv:2609.02309","returned_title_or_snippet_heading":"Efficient GUI Agents: A Systems Survey of Observation, Memory, Action, and Runtime Optimization","status":"existing_inventory_work_exact_version_primary_page_read","url":"https://arxiv.org/abs/2609.02309"},{"canonical_id":"arxiv:2412.13501","returned_title_or_snippet_heading":"GUI Agents: A Survey","status":"existing_inventory_work_exact_version_primary_page_read","url":"https://arxiv.org/abs/2412.13501"},{"canonical_id":null,"returned_title_or_snippet_heading":"Ryan Aponte - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/ryan-aponte/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Dang Nguyen - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/dang-nguyen/"},{"canonical_id":"arxiv:2411.04890","returned_title_or_snippet_heading":"GUI Agents with Foundation Models:","status":"same_title_candidate_edition_not_used; exact_version_arxiv_fulltext_read","url":"https://openreview.net/pdf/76a99204e90365b019e29b3e395074212a6bcbf8.pdf"},{"canonical_id":"arxiv:2503.23434","returned_title_or_snippet_heading":"Towards Trustworthy GUI Agents: A Survey","status":"existing_inventory_work_exact_version_primary_page_read","url":"https://arxiv.org/abs/2503.23434"},{"canonical_id":"arxiv:2412.13501","returned_title_or_snippet_heading":"GUI Agents: A Survey","status":"publication_edition_of_existing_work_not_counted_as_new_paper; exact_arxiv_metadata_read","url":"https://aclanthology.org/2025.findings-acl.1158.pdf"},{"canonical_id":null,"returned_title_or_snippet_heading":"Jihyung Kil - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/jihyung-kil/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Seunghyun Yoon - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/seunghyun-yoon/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Gang Wu - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/gang-wu/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Franck Dernoncourt - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/franck-dernoncourt/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Thien Nguyen - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/thien-nguyen/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Jing Shi - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/jing-shi/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Hanjia Lyu - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/hanjia-lyu/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Yu Xia - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/yu-xia/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Nesreen K. Ahmed - ACL Anthology","status":"author_profile_not_a_paper; duplicate_discovery_hit","url":"https://aclanthology.org/people/nesreen-k-ahmed/"},{"canonical_id":null,"returned_title_or_snippet_heading":"[10] Shuai Wang, Weiven 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, and Jianye Hao. Gui agents with foundation models: A comprehensive survey, 2025.","status":"reference_snippet_not_independent_paper_identity; not_counted","url":"https://openreview.net/pdf/4fe14c361104913e172e118fa05d358d3b030840.pdf"},{"canonical_id":null,"returned_title_or_snippet_heading":"MAI-UI Technical Report: Real-World Centric Foundation GUI Agents","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://openreview.net/pdf/eea51b953ed3948e699f2a4c63e3f9dba1445710.pdf"},{"canonical_id":"arxiv:2503.23434","returned_title_or_snippet_heading":"Towards Trustworthy GUI Agents: A Survey","status":"same_title_candidate_edition_not_used; exact_version_arxiv_fulltext_read","url":"https://openreview.net/pdf/9594a7359cf051689484477b8b099bff8fab66f9.pdf"},{"canonical_id":"arxiv:2503.23434","returned_title_or_snippet_heading":"Towards Trustworthy GUI Agents: A Survey","status":"same_title_candidate_edition_not_used; exact_version_arxiv_fulltext_read","url":"https://openreview.net/attachment?id=lN7UqyfSXU&name=pdf"}],"returned_result_count_in_combined_response":22,"searched_at":"2026-10-18 10:00:02 UTC"},{"queries":[{"domains":["people.csail.mit.edu","sikuli.org","mit.edu"],"q":"Sikuli Using GUI Screenshots for Search and Automation 2009 pdf"}],"query_result_mapping_note":"The tool returned one combined result stream for this batch; per-query hit allocation is not reconstructed.","receipt_sha256":"bc53fe249e0748428151ce06c6bf1a3a909f17f9b9628df5fd8f4292edd8a7d8","results":[{"canonical_id":"url:a4cdb2ee0d07d3980651","returned_title_or_snippet_heading":"Sikuli: Using GUI screenshots for search and automation","status":"existing_work_primary_abstract_and_metadata_read","url":"https://dspace.mit.edu/entities/publication/604a3f79-451d-4186-a94e-d12690ffe40a"},{"canonical_id":"url:a4cdb2ee0d07d3980651","returned_title_or_snippet_heading":"Tsung-Hsiang Chang's projects","status":"author_project_sections_read; exact_paper_link_followed; PDF_failed","url":"https://people.csail.mit.edu/vgod/projects.shtml"},{"canonical_id":null,"returned_title_or_snippet_heading":"User Interface Design Group - MIT CSAIL","status":"supporting_profile_project_or_video_result; not_new_paper; not_used_for_claims","url":"https://up.csail.mit.edu/uid.html"},{"canonical_id":null,"returned_title_or_snippet_heading":"Sikuli: Using GUI Screenshots for Search and Automation | MIT CSAIL","status":"supporting_profile_project_or_video_result; not_new_paper; not_used_for_claims","url":"https://www.csail.mit.edu/node/6181"},{"canonical_id":null,"returned_title_or_snippet_heading":"Resume - Tsung-Hsiang Chang","status":"supporting_profile_project_or_video_result; not_new_paper; not_used_for_claims","url":"https://people.csail.mit.edu/vgod/resume.html"},{"canonical_id":null,"returned_title_or_snippet_heading":"Sikuli¹ is our first attempt to","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://up.csail.mit.edu/other-pubs/vgod-thesis.pdf"},{"canonical_id":null,"returned_title_or_snippet_heading":"CV - Tsung-Hsiang Chang","status":"supporting_profile_project_or_video_result; not_new_paper; not_used_for_claims","url":"https://people.csail.mit.edu/vgod/cv.html"},{"canonical_id":null,"returned_title_or_snippet_heading":"UAP Report","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://up.csail.mit.edu/other-pubs/adaml-aup.pdf"},{"canonical_id":null,"returned_title_or_snippet_heading":"","status":"supporting_profile_project_or_video_result; not_new_paper; not_used_for_claims","url":"https://people.csail.mit.edu/vgod/resume-old.html"},{"canonical_id":null,"returned_title_or_snippet_heading":"Printed","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://up.csail.mit.edu/other-pubs/uist2011-sikuli-guide.pdf"},{"canonical_id":null,"returned_title_or_snippet_heading":"MIT Open Access Articles","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://dspace.mit.edu/bitstream/handle/1721.1/72684/Miller_GUI%20testing.pdf"},{"canonical_id":null,"returned_title_or_snippet_heading":"Using graphical representation of user interfaces as visual references","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://dspace.mit.edu/entities/publication/3c3cddc2-7b89-4e00-858f-7c347f0e6174"},{"canonical_id":null,"returned_title_or_snippet_heading":"Sikuli Slides","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://slides.sikuli.org/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Sikuli Slides","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://slides.sikuli.org/docs/"},{"canonical_id":null,"returned_title_or_snippet_heading":"Picture-driven computing | MIT News | Massachusetts Institute of Technology","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://news.mit.edu/2010/screen-shots-0120"},{"canonical_id":null,"returned_title_or_snippet_heading":"A mobile phone GuiDE : a GUI Decoder and Enricher","status":"title_or_snippet_seen_not_selected_for_primary_read","url":"https://dspace.mit.edu/entities/publication/78a6b688-0fb2-4caa-81fa-477e14b95291"}],"returned_result_count_in_combined_response":16,"searched_at":"2026-10-18 10:10:35 UTC"}],"verbatim_abstract_policy":"Primary citation_abstract payloads are retained only in private cache. Public evidence contains abstract hashes, bibliographic facts and original reviewer findings.","writing_review":{"pass_1_argument_and_meaning":[{"checked":true,"paragraph":1,"purpose":"Define computer-use scope through software-state interaction without fixing architecture or unrestricted autonomy."},{"checked":true,"paragraph":2,"purpose":"Distinguish interface modalities and document the concrete relationship required for hybrid tool use."},{"checked":true,"paragraph":3,"purpose":"Separate UI understanding, action prediction and complete task execution while allowing human intervention."},{"checked":true,"paragraph":4,"purpose":"Establish pre-foundation-model roots in editable demonstrations, collaboration and screenshot scripting."},{"checked":true,"paragraph":5,"purpose":"Distinguish mobile task automation from a reusable interface corpus."},{"checked":true,"paragraph":6,"purpose":"Trace executable web learning environments and correctly separate MiniWoB from its later extension."},{"checked":true,"paragraph":7,"purpose":"Distinguish representation learning and demonstration datasets from interactive recovery evidence."},{"checked":true,"paragraph":8,"purpose":"Compare simulated shopping and recorded real-website tasks without inflating live-action claims."},{"checked":true,"paragraph":9,"purpose":"Explain the move toward outcome evaluation while retaining environment/platform constraints."},{"checked":true,"paragraph":10,"purpose":"Synthesize overlapping historical threads without claiming linear replacement or causal superiority."},{"checked":true,"paragraph":11,"purpose":"Frame eight prior-survey comparisons by documented emphasis and exact inspected revisions."},{"checked":true,"paragraph":12,"purpose":"Recognize existing work on assets, trust and efficiency and avoid an unsupported novelty/absence claim."},{"checked":true,"paragraph":13,"purpose":"State this manuscript’s complementary paper-to-artifact emphasis as its own editorial aim."},{"checked":true,"paragraph":14,"purpose":"Disclose bounded selection and actual reading levels separately from inventory coverage."}],"pass_2_reader_and_language":{"checks":["Definitions precede historical synthesis.","No invented experiment, score or cross-benchmark ranking.","Survey comparison uses stated focus, not absence checkboxes.","No first/only survey novelty claim.","Static UI tasks, offline action matching, simulated web and controlled interactive execution remain distinct.","Readable English; no internal task IDs or filesystem paths in main prose.","Citation keys resolve to frozen canonical records.","Reading and inventory counts explicitly separated."],"completed":true,"limitations":["Root must inspect integrated rendered HTML; this agent did not perform browser/layout verification.","The 20 full-text inspections cover listed relevant sections, not every page, appendix or cited experiment."]},"skill":"paper-writing-review","visible_word_count":1960}}