MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research

Hugging Face Daily Papers Papers

Summary

MobileGym is a browser-based simulation platform for mobile GUI agent research, featuring deterministic state evaluation and scalable parallel execution. It includes a benchmark of 416 tasks and demonstrates gains using GRPO on Qwen3-VL-4B.

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.
Original Article
View Cached Full Text

Cached at: 05/27/26, 02:47 AM

Paper page - MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research

Source: https://huggingface.co/papers/2605.26114 Authors:

,

,

,

,

,

,

,

,

,

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 throughdeterministic state-based judgingoverstructured JSON state, and scalable online RL through low-costparallel 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. Alayered state modeland adeclarative task-definition frameworkkeep state programmability and task creation practical at scale, and a singleprogrammatic judging mechanismdelivers both deterministic evaluation verdicts and dense RL rewards. The accompanying MobileGym-Bench provides 416parameterized task templates, including 256 test and 160 train templates, over 28 apps, with deterministic judges and a structuredAnswerSheet protocolthat avoids free-text matching failures. In aSim-to-Real case study,GRPOonQwen3-VL-4B-Instructgains +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.

View arXiv pageView PDFProject pageGitHubAdd to collection

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 paper0

No model linking this paper

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

Datasets citing this paper0

No dataset linking this paper

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

Spaces citing this paper0

No Space linking this paper

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

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

SimuWoB: Simulating Real-World Mobile Apps for Fast and Faithful GUI Agent Benchmarking

Hugging Face Daily Papers

SimuWoB is a synthetic benchmark with 120 challenging tasks for mobile GUI agents, using high-fidelity virtual environments and automatic reward generation. Experiments reveal that current agents achieve only 27.92% average success rate, dropping to 17.82% on long-horizon tasks, indicating substantial weaknesses in complex scenarios.