Reproduce “HydroAgent: Closing the Gap Between Frontier LLMs and Human Experts in Hydrologic Model Calibration via Simulator-Grounded RL” 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
Zhi Li, Songkun Yan, Jie Cao, Mofan Zhang, Anjiang Wei, Jinwoong Yoo, Yang Hong
The authors train HydroAgent, a 4B-parameter Qwen3 model, to calibrate streamflow in CREST by learning tool use from GPT-5-generated examples and then refining parameters through online simulator feedback. On four held-out U.S. gauges, it improves NSE over untuned Qwen3 on each gauge, raising the reported panel mean from −0.14 to +0.20; this limited test does not establish broader generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.