arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.14296cs.AIcs.SE
收藏

AutoWebWorld: 通过有限状态机合成无限可验证的网页环境

AutoWebWorld: Synthesizing Infinite Verifiable Web Environments via Finite State Machines

The Hong Kong University of Science and Technology (Guangzhou)(香港科学与技术大学(广州));Peking University(北京大学)

Yifan Wu, Yiran Peng, Yiyu Chen, Jianhao Ruan, Zijie Zhuang, Cheng Yang, Jiayi Zhang, Man Chen, Yenchi Tseng, Zhaoyang Yu, Liang Chen, Yuyao Zhai, Bang Liu, Chenglin Wu, Yuyu Luo

AI总结:

AutoWebWorld通过有限状态机合成可验证的网页环境,实现自动化搜索与验证流程,提升WebGUI代理性能。

摘要:

自主网页GUI代理的性能严重依赖于其训练数据的质量和数量。然而,一个根本性的瓶颈仍然存在:从现实网站中收集交互轨迹是昂贵且难以验证的。底层的状态转换是隐藏的,导致依赖不一致且昂贵的外部验证器来评估每一步的正确性。为了解决这个问题,我们提出了AutoWebWorld,一种新的框架,通过将它们建模为有限状态机(FSMs)来合成可控且可验证的网页环境,并使用编码代理将FSMs转换为交互式网页。与现实网站不同,其中状态转换是隐含的,AutoWebWorld明确定义了所有状态、动作和转换规则。这使程序验证成为可能:动作的正确性通过预定义的规则进行检查,任务的成功通过在FSM图中达到目标状态来确认。AutoWebWorld实现了一个完全自动的搜索和验证流程,仅以每条轨迹0.04美元的成本,从29种不同的网页环境中生成超过11,663条验证过的轨迹。在合成数据上训练显著提升了现实中的性能。我们的7B Web GUI代理在WebVoyager上15步内超越了所有基线。此外,我们观察到一个清晰的扩展定律:随着合成数据量的增加,WebVoyager和Online-Mind2Web上的性能持续改善。

英文摘要:

The performance of autonomous Web GUI agents heavily relies on the quality and quantity of their training data. However, a fundamental bottleneck persists: collecting interaction trajectories from real-world websites is expensive and difficult to verify. The underlying state transitions are hidden, leading to reliance on inconsistent and costly external verifiers to evaluate step-level correctness. To address this, we propose AutoWebWorld, a novel framework for synthesizing controllable and verifiable web environments by modeling them as Finite State Machines (FSMs) and use coding agents to translate FSMs into interactive websites. Unlike real websites, where state transitions are implicit, AutoWebWorld explicitly defines all states, actions, and transition rules. This enables programmatic verification: action correctness is checked against predefined rules, and task success is confirmed by reaching a goal state in the FSM graph. AutoWebWorld enables a fully automated search-and-verify pipeline, generating over 11,663 verified trajectories from 29 diverse web environments at only $0.04 per trajectory. Training on this synthetic data significantly boosts real-world performance. Our 7B Web GUI agent outperforms all baselines within 15 steps on WebVoyager. Furthermore, we observe a clear scaling law: as the synthetic data volume increases, performance on WebVoyager and Online-Mind2Web consistently improves.

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。

↑