MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer Environments
Quick Answer
MAI-Gym is an open platform for GUI agent evaluation and RL training, achieving state-of-the-art performance with a 9B agent across mobile and computer environments.
Quick Take
It supports 20,000 parallel instances and demonstrates a 48.1% task success rate in real-world applications after training in a mocked environment. The framework includes 2,000 verified tasks and is aimed at advancing scalable GUI agent training.
Key Points
- MAI-Gym unifies GUI , data collection, and online RL training.
- Achieved 78.4% performance on MobileWorld and 76.2% on OSWorld-Verified benchmarks.
- Supports automated synthesis of diverse GUI applications and challenging tasks.
- Trained MAI-Gym-9B agent shows significant real-world transfer improvements.
- Framework released for reproducible research and further advancements.
Paper Resources
📖 Reader Mode
~2 min readAuthors:Panrong Tong, Quyu Kong, Pengxiang Li, Tianyu Xia, Hanzhang Zhou, Lei Wang, Chenglin Cai, Long Li, Long Chen, Jianan Zhang, Zhengwen Zeng, Xu Zhang, Gongjie Zhang, Yue Wang, Steven Hoi
Abstract:Online reinforcement learning (RL) with verifiable rewards has driven rapid progress in mathematics, coding, and tool use, but scaling this success to GUI agents remains challenging due to the high cost of interactive environments and the scarcity of realistic training tasks. We introduce MAI-Gym, an open all-in-one platform that unifies GUI agent evaluation, data collection, and online RL training across mobile, computer, and web environments. Through a shared rollout pipeline, MAI-Gym ensures consistent agent behavior across evaluation and training, supports ten established benchmarks, and scales to 20,000 parallel environment instances. Beyond scalable infrastructure, MAI-Gym enables automated synthesis of diverse GUI applications and challenging tasks, including resettable mocked apps and coherent cross-app worlds with execution-validated task verifiers. Using MAI-Gym, we train MAI-Gym-9B, a unified 9B agent for mobile and computer use, achieving state-of-the-art performance among open GUI agents under 10B parameters: 78.4% on MobileWorld and 76.2% on OSWorld-Verified. Importantly, RL in a mocked train-ticketing app transfers to its real-world counterpart, improving task success on a phys- ical phone from 28.8% to 48.1%. We release the framework, synthesis skills, mocked apps, 2,000 verified tasks, and the model checkpoint to support repro- ducible research and further advances in scalable GUI agent training.
| Comments: | 31 pages, 15 pages |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.20939 [cs.AI] |
| (or arXiv:2610.20939v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20939 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tianyu Xia [view email]
[v1]
Wed, 14 Oct 2026 21:06:01 UTC (9,779 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →VEPA: Verified Evolving Formal Policy for Lifelong Agent Guardrails
VEPA introduces a framework for verified evolving formal policies for LLM agents, enhancing safety by adapting guardrails based on deployment feedback. It outperforms existing methods across benchmarks like ASSEBench-Safety and ASSEBench-Security, ensuring deterministic policy enforcement even under challenging conditions.