AgentCPM-GUI: Building Mobile-Use Agents with Reinforcement Fine-Tuning

TL;DR

AgentCPM-GUI is a robust, multilingual mobile GUI agent trained with grounding-aware pre-training, supervised fine-tuning, and reinforcement learning, achieving state-of-the-art performance on multiple benchmarks.

Contribution

The paper introduces AgentCPM-GUI, a novel 8B-parameter GUI agent with a comprehensive training pipeline and a compact action space for mobile environments, addressing prior limitations.

Findings

Achieves 96.9% Type-Match and 91.3% Exact-Match on benchmarks.

Outperforms existing models on five public benchmarks and a new Chinese GUI benchmark.

Demonstrates effective multilingual and cross-scenario GUI interaction.

Abstract

The recent progress of large language model agents has opened new possibilities for automating tasks through graphical user interfaces (GUIs), especially in mobile environments where intelligent interaction can greatly enhance usability. However, practical deployment of such agents remains constrained by several key challenges. Existing training data is often noisy and lack semantic diversity, which hinders the learning of precise grounding and planning. Models trained purely by imitation tend to overfit to seen interface patterns and fail to generalize in unfamiliar scenarios. Moreover, most prior work focuses on English interfaces while overlooks the growing diversity of non-English applications such as those in the Chinese mobile ecosystem. In this work, we present AgentCPM-GUI, an 8B-parameter GUI agent built for robust and efficient on-device GUI interaction. Our training pipeline includes grounding-aware pre-training to enhance perception, supervised fine-tuning on high-quality Chinese and English trajectories to imitate human-like actions, and reinforcement fine-tuning with GRPO to improve reasoning capability. We also introduce a compact action space that reduces output length and supports low-latency execution on mobile devices. AgentCPM-GUI achieves state-of-the-art performance on five public benchmarks and a new Chinese GUI benchmark called CAGUI, reaching $96.9\%$ Type-Match and $91.3\%$ Exact-Match. To facilitate reproducibility and further research, we publicly release all code, model checkpoint, and evaluation data.