Reproduce “WebCoder: Scaling General Frontend Development with Artifact-Grounded Agentic Reinforcement Learning” 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
Zijun Dou, Zhen Yang, Xijun Liu, Dongzhuoran Zhou, Aohan Zeng, Jie Tang
WebCoder trains frontend agents to develop applications while coordinating multiple requirements. It uses an independent verifier to assess the finished app and estimates requirement progress between development segments to distribute learning credit. On Webdev-Bench, it scores 72.8 RS, versus 64.8 for its base model and 65.1 for Claude-Opus-4.8, with runnable apps on all 40 tasks. These results are under an LLM-judged benchmark protocol.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.