[论文] WebHarbor: Docking Real Websites for Evolving GUI Agent Environments

论文概要

研究领域: NLP
作者: Zhaoyang Wang, Qianhui Wu, Boyu Gou, Demi Wang, Jiacheng Zhang, Peichun Hua, Qize Li, Shi Qiu, Yifan Jiang, Yifan Zhang, Zexu Jin, Baolin Peng, Bolin Shen, Chetan Bansal, Dehan Kong, Deren Lei, Dongshen Peng, Fengqing Jiang, Hao Cheng, Haoqing Wang, Jiaming Shan, Jicheng Wang, Jingguang Zhou, Jixuan He, Kaide Zeng, Kangyu Zhu, Lama Moukheiber, Mufan Qiu, Peng Xia, Qihan Sun, Qingwei Lin, Qiushi Sun, Ruichen Zhang, Shuqing Luo, Si Qin, Sicheng Fan, Tao Xiong, Wenlin Yao, Xinyu Zhang, Xinyue Zeng, Xuanrui Li, Xuchao Zhang, Yian Wang, Yichao Liu, Yiming Liang, Yiyang Zhou, Yizhu Jiao, Yuetai Li, Yurun Chen, Yuxuan Zhang, Zheyuan Yang, Zhiyuan Yao, Zhongyang Li, Zihao Li, Zili Wang, Jianfeng Gao, Huaxiu Yao
发布时间: 2026-11-02
arXiv: 2611.03634

中文摘要

GUI Web 智能体需要逼真、交互式、在评估与训练中都保持稳定的环境。对真实网站的评估 suffers from 内容变化、访问限制、认证工作流覆盖不足；而现有的合成训练环境往往功能深度与视觉保真度都不够。我们提出 WebHarbor：一套 153 个可执行、Docker 化的网站环境，能在确定性控制下高保真重建真实 Web 服务的视觉界面、内容与功能工作流。WebHarbor 通过任务驱动的人机协作构建——环境随新任务扩展功能范围，编码智能体实现所需改动，人类贡献者审计视觉保真度、功能正确性与任务有效性。最终套件包含 3,518 个任务，覆盖多步与认证后工作流——这正是已有工作常覆盖不到的。除可靠评估外，WebHarbor 还为低成本强化学习（RL）提供高保真合成 Web 环境。我们证明：在 WebHarbor 中用 RL 训练的 Qwen 4B 级 GUI 智能体能泛化到未见任务，并在四个基准上与领先的开源 GUI Web 智能体竞争。代码与数据已开源：https://github.com/aiming-lab/WebHarbor

原文摘要

GUI web agents require realistic and interactive environments that remain stable during both evaluation and training. Live website evaluation suffers from changing content, access restrictions, and limited coverage of authenticated workflows, while existing synthetic environments for training often lack functional depth and visual fidelity. We propose WebHarbor, a suite of 153 executable and dockerized website environments that faithfully reconstruct the visual interfaces, content, and functional workflows of real-world web services under deterministic control. We build WebHarbor through task-driven human-agent collaboration, where environments evolve as new tasks expand their functional scope, coding agents implement the required changes, and human contributors audit visual fidelity, func...

自动采集于 2026-11-04

#论文 #arXiv #NLP #小凯