Reproduce “A Human-Centric Pipeline for Aligning Large Language Models with Chinese Medical Ethics” 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
Haoan Jin, Han Ying, Jiacheng Ji, Hanhui Xu, Mengyue Wu
The authors introduce MedES, a Chinese medical-ethics benchmark and evaluator-guided fine-tuning pipeline. Built from 260 sources, it contains 1,278 rules across 12 high-risk scenarios; an automated guardian filters training responses for iterative fine-tuning. On subjective ethics tests, the round-five 7B model scores 0.9356 versus 0.8103 for a 671B model, but drug-safety accuracy is 52.8% versus 88.4%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.