SE-GUI: Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learning

Bo Li (Nanjing University) · Jie Chen (Hong Kong Baptist University) · Lujian Yao (East China University of Science and Technology) · Xinbin Yuan (Nankai University) · Zhuoxuan Cai (Fudan University) · Jinwei Chen (vivo Mobile Communication Co., Ltd.) · Peng-Tao Jiang (vivo Mobile Communication (Hangzhou) Co., Ltd.) · Jian Zhang (Xiamen University) · Kaixin Li (National University of Singapore) · Enguang Wang (Nankai University) · Qibin Hou (Nankai University)

attention mapsdense policy gradientgrounding benchmarksgrounding instructionsgui agentshigh-resolution environmentsmodel refinement prediction accuracy reinforcement learningscreenspot-pro datasetseed data curationself-evolutionary reinforcement finetuningstate-of-the-art results supervised fine-tuning training samples

Graphical User Interface (GUI) agents have made substantial strides in understanding and executing user instructions across diverse platforms. Yet, grounding these instructions to precise interface elements remains challenging—especially in complex, high-resolution, professional environments. Traditional supervised fine-tuning (SFT) methods often require large volumes of diverse data and exhibit weak generalization. To overcome these limitations, we introduce a reinforcement learning (RL)-based framework that incorporates three core strategies: (1) seed data curation to ensure high-quality training samples, (2) a dense policy gradient that provides continuous feedback based on prediction accuracy, and (3) a self-evolutionary reinforcement finetuning mechanism that iteratively refines the model using attention maps. With only 3k training samples, our 7B-parameter model achieves state-of-the-art results among similarly sized models on three grounding benchmarks. Notably, it attains 47.3\% accuracy on the ScreenSpot-Pro dataset—outperforming much larger models, such as UI-TARS-72B, by a margin of 24.2\%. These findings underscore the effectiveness of RL-based approaches in enhancing GUI agent performance, particularly in high-resolution, complex environments.