MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research
Abstract
MobileGym presents a browser-based mobile environment enabling deterministic evaluation and scalable reinforcement learning through JSON-based state management and parallel execution.
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.
Community
MobileGym is a browser-hosted, lightweight, fully controllable, and highly parallel environment for everyday mobile use.
that core idea of representing the whole mobile world as a public, structured json state and forking identical initial conditions for parallel rollouts is genuinely clever. it lets you run verifiable judges and dense rl rewards without real devices, which obviously helps scale safety and robustness work. but how do they bound or handle subtle non-determinism in os services or timing like network callbacks or animation frames across forks, especially for longer sims? btw the arxivlens breakdown helped me parse the method details: https://arxivlens.com/PaperView/Details/mobilegym-a-verifiable-and-highly-parallel-simulation-platform-for-mobile-gui-agent-research-2227-e12f8863
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agents (2026)
- KnowU-Bench: Towards Interactive, Proactive, and Personalized Mobile Agent Evaluation (2026)
- Towards Scalable Lightweight GUI Agents via Multi-role Orchestration (2026)
- OpenComputer: Verifiable Software Worlds for Computer-Use Agents (2026)
- Computer Use at the Edge of the Statistical Precipice (2026)
- WindowsWorld: A Process-Centric Benchmark of Autonomous GUI Agents in Professional Cross-Application Environments (2026)
- Weblica: Scalable and Reproducible Training Environments for Visual Web Agents (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend
Get this paper in your agent:
hf papers read 2605.26114 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper