---
id: 20260506-T0-07
title: "GUI智能体精准点击新突破：自蒸馏策略让视觉定位更准"
title_en: "On-Policy Self-Distillation Boosts GUI Grounding Accuracy for AI Agents"
url: https://ai.daily.yangsir.net/daily/20260506-T0-07
issue_date: 2026-05-06
publish_date: 2026-05-05T04:00:00.000Z
category: research
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2605.00642
---

# GUI智能体精准点击新突破：自蒸馏策略让视觉定位更准

该研究提出了一种基于自蒸馏的强化学习策略，用于提升GUI智能体的视觉定位能力。GUI Grounding任务要求智能体根据自然语言指令，在屏幕上准确定位目标元素的坐标。虽然近期的强化学习方法（如GRPO）取得了不错的成绩，但该研究通过自身策略蒸馏进一步优化了这一过程。这使得自动化测试、RPA流程自动化等场景中的智能体能够更精准地理解和操作图形界面。

## English Version

**On-Policy Self-Distillation Boosts GUI Grounding Accuracy for AI Agents**

This research proposes an on-policy self-distillation reinforcement learning strategy to improve GUI grounding for autonomous agents. The task requires mapping natural language instructions to precise visual coordinates of target elements on screen. Building on recent RL methods like GRPO, this self-distillation approach further optimizes the grounding process. This enables AI agents in automated testing and RPA scenarios to understand and interact with graphical interfaces more accurately.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2605.00642)

**详情页**：https://ai.daily.yangsir.net/daily/20260506-T0-07

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*