GUI-G1：理解GUI智能体中视觉定位的R1-Zero类训练
GUI-G1: Understanding R1-Zero-Like Training for Visual Grounding in GUI Agents
AI总结：
本文提出GUI-G1，通过分析R1-Zero类训练在GUI定位中的三大挑战，采用快速思考模板、框大小约束和难度感知缩放因子，在ScreenSpot上达到90.3%准确率，超越同规模及更大模型。
摘要：
近期图形用户界面（GUI）智能体复现了R1-Zero范式，将在线强化学习（RL）与显式思维链推理相结合，在对象定位之前进行推理，从而获得了显著的性能提升。本文首先对该训练流程的三个关键组件进行了广泛的分析实验：输入设计、输出评估和策略更新——每个组件都揭示了因盲目应用通用RL而未适配GUI定位任务而产生的不同挑战。输入设计：当前模板鼓励模型生成思维链推理，但更长的推理链意外地导致更差的定位性能。输出评估：基于命中信号或框面积的奖励函数允许模型利用框大小，导致奖励黑客和较差的定位质量。策略更新：在线RL因长度和样本难度偏差而倾向于过拟合简单示例，导致对较难案例优化不足。为解决这些问题，我们提出了三个针对性解决方案。首先，我们采用快速思考模板，鼓励直接生成答案，减少训练中的过度推理。其次，我们在奖励函数中加入框大小约束以缓解奖励黑客。第三，我们通过调整长度归一化并添加难度感知缩放因子来修正RL目标，从而在困难样本上实现更好的优化。我们的GUI-G1-3B，在17K公开样本上使用Qwen2.5-VL-3B-Instruct训练，在ScreenSpot上达到90.3%的准确率，在ScreenSpot-Pro上达到37.1%。这超越了所有先前类似规模的模型，甚至优于更大的UI-TARS-7B，在GUI智能体定位中确立了新的最先进水平。项目仓库可在https://github.com/Yuqi-Zhou/GUI-G1获取。
英文摘要：
Recent Graphical User Interface (GUI) agents replicate the R1-Zero paradigm, coupling online Reinforcement Learning (RL) with explicit chain-of-thought reasoning prior to object grounding and thereby achieving substantial performance gains. In this paper, we first conduct extensive analysis experiments of three key components of that training pipeline: input design, output evaluation, and policy update-each revealing distinct challenges arising from blindly applying general-purpose RL without adapting to GUI grounding tasks. Input design: Current templates encourage the model to generate chain-of-thought reasoning, but longer chains unexpectedly lead to worse grounding performance. Output evaluation: Reward functions based on hit signals or box area allow models to exploit box size, leading to reward hacking and poor localization quality. Policy update: Online RL tends to overfit easy examples due to biases in length and sample difficulty, leading to under-optimization on harder cases. To address these issues, we propose three targeted solutions. First, we adopt a Fast Thinking Template that encourages direct answer generation, reducing excessive reasoning during training. Second, we incorporate a box size constraint into the reward function to mitigate reward hacking. Third, we revise the RL objective by adjusting length normalization and adding a difficulty-aware scaling factor, enabling better optimization on hard samples. Our GUI-G1-3B, trained on 17K public samples with Qwen2.5-VL-3B-Instruct, achieves 90.3% accuracy on ScreenSpot and 37.1% on ScreenSpot-Pro. This surpasses all prior models of similar size and even outperforms the larger UI-TARS-7B, establishing a new state-of-the-art in GUI agent grounding. The project repository is available at https://github.com/Yuqi-Zhou/GUI-G1.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
