OS-Oracle：跨平台 GUI 批评模型的综合框架
OS-Oracle: A Comprehensive Framework for Cross-Platform GUI Critic Models
为视觉语言模型在 GUI 环境中的动作评估提供了可扩展的数据合成流程和两阶段训练范式（SFT + CP-GRPO），可直接用于提升多模态智能体的长程任务可靠性。
原文依据 a two-stage training paradigm combining supervised fine-tuning (SFT) and consistency-preserving group relative policy optimization (CP-GRPO)
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论在图形用户界面（GUI）环境中部署自主智能体面临显著挑战，尤其是长程任务中的误差累积以及不可逆操作的严重后果。提供实时动作评估的批评模型是一种有前景的解决方案，但其有效性受到缺乏多样、高质量的 GUI 反馈数据以及面向计算机使用的公开批评基准的制约。为弥补这些空白，我们提出 OS-Oracle，做出三项核心贡献：（1）一个可扩展的数据流程，用于合成跨平台 GUI 批评数据；（2）一种两阶段训练范式，结合监督微调（SFT）与保持一致性组相对策略优化（CP-GRPO）；（3）OS-Critic Bench，一个用于评估批评模型在移动端、网页端和桌面端平台性能的全面基准。利用该框架，我们整理了一个包含 31 万条批评样本的高质量数据集。所得批评模型 OS-Oracle-7B 取得了令人印象深刻的性能，并进一步降低了错误率，从而提升了 GUI 智能体在动态环境中的能力。所有代码、数据和检查点将公开。
英文原文摘要 1191 字符
The deployment of autonomous agents in Graphical User Interface (GUI) environments confronts significant challenges, notably error accumulation in long-horizon tasks and the severe consequences of irreversible operations. While critic models that provide real-time action assessment offer a promising solution, their effectiveness is hindered by the lack of diverse, high-quality GUI feedback data and public critic benchmarks for computer use.To bridge these gaps,we introduce OS-Oracle that makes three core contributions:(1) a scalable data pipeline for synthesizing cross-platform GUI critic data;(2) a two-stage training paradigm combining supervised fine-tuning (SFT) and consistency-preserving group relative policy optimization (CP-GRPO); (3) OS-Critic Bench, a holistic benchmark for evaluating critic model performance across Mobile, Web, and Desktop platforms.Leveraging this framework, we curate a high-quality dataset containing 310k critic samples. The resulting critic model, OS-Oracle-7B, achieves impressive performance,and further reduces error rates, which improves the capability of GUI agents in dynamic environments. All codes, data and checkpoints will be made public.