<ul data-eligibleForWebStory="true"><li>Autonomous agents are changing GUI interactions using natural language as an intermediary.</li><li>Supervised Fine-Tuning methods in GUI agents struggle with accurate positional data perception.</li><li>Reinforcement learning methods often fall short in assessing positional accuracy effectively.</li><li>Location Preference Optimization (LPO) is introduced to optimize interaction preferences using locational data.</li><li>LPO uses information entropy to predict interaction positions and introduces a dynamic location reward function based on physical distance.</li><li>LPO, supported by Group Relative Preference Optimization (GRPO), enhances interaction precision in GUI environments.</li><li>Experiments demonstrate LPO's superior performance, achieving state-of-the-art results in offline benchmarks and real-world online evaluations.</li><li>The code for LPO will be publicly available soon on GitHub at https://github.com/AIDC-AI/LPO.</li></ul>

LPO: Towards Accurate GUI Agent Interaction via Location Preference Optimization

Discover more