MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization
Summary
MobileForge presents an annotation-free adaptation system for mobile GUI agents that uses real app interaction and hierarchical feedback-guided policy optimization to improve performance, achieving near state-of-the-art results on AndroidWorld with open data.
View Cached Full Text
Cached at: 06/24/26, 05:47 AM
Paper page - MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization
Source: https://huggingface.co/papers/2606.19930
Abstract
MobileForge enables efficient adaptation of mobile GUI agents through annotation-free learning by combining real app interaction grounding with hierarchical feedback-guided policy optimization.
MLLM-based mobile GUI agentshave made substantial progress in UI understanding and action execution, but adapting them to real target apps remains costly because mobile apps are numerous, frequently updated, and hard to cover with human-written tasks, demonstrations, or reward labels. Existingannotation-free GUI learningreduces manual supervision, yet lacks a unified substrate connecting target-app exploration, curriculum mining, rollout execution, and feedback, while policy optimization often relies on isolated rollouts and coarse rewards that are hard to convert into reliable improvement signals. We present MobileForge, an annotation-free adaptation system for mobile GUI agents. MobileForge consists ofMobileGym, which grounds task generation and rollout evaluation in real mobile app interaction, andHierarchical Feedback-Guided Policy Optimization(HiFPO), which turns trajectory outcomes, step-level process feedback, and corrective hints into hint-contextualized step-levelGRPO updates. Using only automatically generated annotation-free adaptation data, MobileForge adapts Qwen3-VL-8B to 67.2%Pass@3on AndroidWorld, close to the closed-data GUI-specialized GUI-Owl-1.5-8B base model at 69.0%. The MobileForge-adapted ForgeOwl-8B further reaches 77.6%Pass@3on AndroidWorld and 41.0% success on the out-of-domain MobileWorld GUI-only split, establishing the strongest open-data mobile GUI agent in our evaluation. Code, data, and trained models will be released at https://mobile-forge.github.io/.
View arXiv pageView PDFProject pageGitHub1Add to collection
Get this paper in your agent:
hf papers read 2606\.19930
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/2606.19930 in a model README.md to link it from this page.
Datasets citing this paper4
#### lgy0404/mobileforge-exploration-trajectories #### lgy0404/mobileforge-benchmark-results Updatedabout 23 hours ago • 2 #### lgy0404/mobileforge-generated-tasks #### lgy0404/mobileforge-training-data Updatedabout 19 hours ago
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.19930 in a Space README.md to link it from this page.
Collections including this paper2
Similar Articles
CoAdapt-GUI: Joint Workflow Context and Policy Adaptation for Unseen GUI Applications
CoAdapt-GUI is a test-time adaptation framework for mobile GUI agents that jointly adapts workflow context and policy, improving performance on unseen-app benchmarks like AndroidWorld-Generalization and AndroidWorld Plus.
MAGA: Multi-Platform Self-Fusion of GUI Agents via Structured Action Distillation
This paper introduces Maga, a method for consolidating domain-specific GUI agents into a single cross-platform policy via structured action distillation, reallocating training signals to focus on erroneous actions. It achieves strong success rates across mobile, web, and desktop benchmarks.
HarnessForge: Joint Harness and Policy Evolution for Adaptive Agent Systems
HarnessForge proposes a meta-adaptive framework for evolving LLM agent systems by jointly optimizing the execution harness and reasoning policy, achieving consistent improvements on Qwen3 backbones across five benchmarks.
MobileExplorer: Accelerating On-Device Inference for Mobile GUI Agents via Online Exploration
MobileExplorer is a new framework that accelerates on-device inference for mobile GUI agents by performing lightweight parallel exploration of UI elements during model inference, reducing reasoning steps and latency by 23% while maintaining or improving task success rates.
MIRAGE: Mobile Agents with Implicit Reasoning and Generative World Models
MIRAGE is a framework for mobile GUI agents that replaces verbose chain-of-thought reasoning with compact continuous latent representations, incorporating a generative world model perspective to predict future screen states before acting. On AndroidWorld and AndroidControl benchmarks, it achieves competitive or superior performance while reducing generated tokens by over 75%.