Reproduce “Zephyrus-R1: Agentic Reinforcement Learning for Weather Reporting” 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
Fengfei Yu, Marshall Fisher, Isadora White, Sumanth Varambally, Yasaman Jafari, Shivam Yadav, Samuel M. Bartlett, Shawn Roj and 8 more
Zephyrus-R1 trains an AI agent to gather weather evidence with tools, first on checkable weather questions and then on open-ended forecast reports scored for quality and factual coverage. On 197 temporally disjoint continental U.S. events, Qwen3.8-27B’s curriculum-trained model reached a six-score judge average of 0.571, versus 0.506 for its base model, a 13% relative increase; this is not expert whole-report evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.