Reproduce “LegalEvolver: Agentic Legal Judgment Prediction via Rubric-Guided Self-Evolution” 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
Xinyu Yang, Jiejun Tan, Chenlong Deng, Chenghao Zhang, Tongyu Wen, Zhicheng Dou
LegalEvolver trains agents to improve legal judgment prediction by searching for unresolved legal issues and updating query criteria as policies change. It contrasts same-case trajectories with different outcomes to derive rubrics, then rewards intermediate searches alongside judgment outcomes. On two Chinese datasets, it scores 61.5 overall, versus SmartSearch’s 60.4; results cover benchmarked cases and do not establish real-world judicial performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.