MobileGym is a browser-hosted, lightweight, fully controllable, and highly parallel environment for everyday mobile use.</p>\n","updatedAt":"2026-05-27T02:46:19.801Z","author":{"_id":"649ecf9827145c4463240177","avatarUrl":"/avatars/27696cf31790a3d58d8be2e0c983800e.svg","fullname":"Lue Fan","name":"Abyssaledge","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":5,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8790509104728699},"editors":["Abyssaledge"],"editorAvatarUrls":["/avatars/27696cf31790a3d58d8be2e0c983800e.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.26114","authors":[{"_id":"6a1549d5b57a1823d5708c64","name":"Dingbang Wu","hidden":false},{"_id":"6a1549d5b57a1823d5708c65","name":"Rui Hao","hidden":false},{"_id":"6a1549d5b57a1823d5708c66","name":"Haiyang Wang","hidden":false},{"_id":"6a1549d5b57a1823d5708c67","name":"Shuzhe Wu","hidden":false},{"_id":"6a1549d5b57a1823d5708c68","name":"Han Xiao","hidden":false},{"_id":"6a1549d5b57a1823d5708c69","name":"Zhenghong Li","hidden":false},{"_id":"6a1549d5b57a1823d5708c6a","name":"Bojiang Zhou","hidden":false},{"_id":"6a1549d5b57a1823d5708c6b","name":"Zheng Ju","hidden":false},{"_id":"6a1549d5b57a1823d5708c6c","name":"Zichen Liu","hidden":false},{"_id":"6a1549d5b57a1823d5708c6d","name":"Lue Fan","hidden":false},{"_id":"6a1549d5b57a1823d5708c6e","name":"Zhaoxiang Zhang","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/649ecf9827145c4463240177/e4MW08ivBpq8AQMlrIINL.mp4"],"publishedAt":"2026-05-25T00:00:00.000Z","submittedOnDailyAt":"2026-05-27T00:00:00.000Z","title":"MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research","submittedOnDailyBy":{"_id":"649ecf9827145c4463240177","avatarUrl":"/avatars/27696cf31790a3d58d8be2e0c983800e.svg","isPro":false,"fullname":"Lue Fan","user":"Abyssaledge","type":"user","name":"Abyssaledge"},"summary":"We present MobileGym, a browser-hosted, lightweight, fully controllable environment for everyday mobile use, targeting interaction fidelity without replicating proprietary backends. It enables two capabilities previously out of reach for everyday apps: verifiable outcome signals through deterministic state-based judging over structured JSON state, and scalable online RL through low-cost parallel rollouts. The full environment state is captured, configured, forked, and compared as structured JSON, and a single server can host hundreds of parallel instances, with about 400 MB memory per instance and about 3 s cold start. A layered state model and a declarative task-definition framework keep state programmability and task creation practical at scale, and a single programmatic judging mechanism delivers both deterministic evaluation verdicts and dense RL rewards. The accompanying MobileGym-Bench provides 416 parameterized task templates, including 256 test and 160 train templates, over 28 apps, with deterministic judges and a structured AnswerSheet protocol that avoids free-text matching failures. In a Sim-to-Real case study, GRPO on Qwen3-VL-4B-Instruct gains +12.8 percentage points on the 256-task test set, and on a 59-task real-device signal subset, real-device execution retains 95.1% of the simulation-side training gain. Project page: https://mobilegym.github.io.","upvotes":43,"discussionId":"6a1549d6b57a1823d5708c6f","projectPage":"https://mobilegym.github.io","githubRepo":"https://github.com/Purewhiter/mobilegym","githubRepoAddedBy":"user","ai_summary":"MobileGym presents a browser-based mobile environment enabling deterministic evaluation and scalable reinforcement learning through JSON-based state management and parallel execution.","ai_keywords":["browser-hosted environment","deterministic state-based judging","structured JSON state","parallel rollouts","layered state model","declarative task-definition framework","programmatic judging mechanism","Sim-to-Real case study","GRPO","Qwen3-VL-4B-Instruct","parameterized task templates","AnswerSheet protocol"],"githubStars":24},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"649ecf9827145c4463240177","avatarUrl":"/avatars/27696cf31790a3d58d8be2e0c983800e.svg","isPro":false,"fullname":"Lue Fan","user":"Abyssaledge","type":"user"},{"_id":"688f20de20339f1b229ad00a","avatarUrl":"/avatars/b003bb03901835f6c5161edb818f37dc.svg","isPro":false,"fullname":"Rui Hao","user":"haorui1234","type":"user"},{"_id":"6a165efd01014d9a39ae1953","avatarUrl":"/avatars/c4a1aa90406b1aa5977349aa346effb1.svg","isPro":false,"fullname":"wangfeng","user":"iamresearcher","type":"user"},{"_id":"6531293cd780e9ba7420eafc","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/qtsgrhqkplLQss7X79FQI.jpeg","isPro":false,"fullname":"Weikai Xie","user":"Shrelic","type":"user"},{"_id":"66a889cbe0420835a19e6f12","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/66a889cbe0420835a19e6f12/ImhW2qwlUAm7j8FdunA2B.jpeg","isPro":false,"fullname":"Weizhi Zhao","user":"WeizhiZhao","type":"user"},{"_id":"680e712a2b588ca79da6efa3","avatarUrl":"/avatars/3a6ea6eff251792425a6bbaa0c902397.svg","isPro":false,"fullname":"Ruitong GAN","user":"ganruitong1933","type":"user"},{"_id":"68b16999cf872b3fd0451557","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/aYtIBZ0LDWsRtCRE5Msbw.png","isPro":false,"fullname":"BingZhan","user":"bigbigzi","type":"user"},{"_id":"695b55355b1e2d9ce64d6928","avatarUrl":"/avatars/c5024551bb6f506d554b96995dae04f8.svg","isPro":false,"fullname":"Junran Peng","user":"JrPeng13124","type":"user"},{"_id":"67fc60bbcba7e1d6e45f70a5","avatarUrl":"/avatars/07c3870146f60ead360201c362e903f6.svg","isPro":false,"fullname":"lmj","user":"llmmjj13","type":"user"},{"_id":"67d247f804e727ca41b0b5d5","avatarUrl":"/avatars/8e7a9666a5cbb172a482b8fdbabd0f73.svg","isPro":false,"fullname":"Xiaogong zhang","user":"RookieZxg","type":"user"},{"_id":"68ec95e063255e14ac27450c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/RzqQKNRj0I8UivJbdYOGv.png","isPro":false,"fullname":"Yingyan Li","user":"liyingyan","type":"user"},{"_id":"695b5b47a934de056c75facf","avatarUrl":"/avatars/702261c2e040abc992796e7d4ce97219.svg","isPro":false,"fullname":"Wenbo Chen","user":"wenbo-chen","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":2,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.26114.md"}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2605.26114 in a model README.md to link it from this page.

Cite arxiv.org/abs/2605.26114 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2605.26114 in a Space README.md to link it from this page.