Reproduce “WebHarbor: Docking Real Websites for Evolving GUI Agent Environments” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhaoyang Wang, Qianhui Wu, Boyu Gou, Demi Wang, Jiacheng Zhang, Peichun Hua, Qize Li, Shi Qiu and 49 more
WebHarbor builds task-scoped local copies of real websites as executable environments, using coding agents and human review. Its suite includes 153 environments and 3,518 tasks, including authenticated, state-changing workflows. In a controlled 4B comparison, WebHarbor-4B scores a 75.8 average on two live-web benchmarks, versus 70.6 for OpenWebRL-4B. Because both tasks and environments differ, the comparison does not isolate causes or establish live-site safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.