Reproduce “Decisions by Discrimination: Differential Evidence for Medical Reasoning” 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

Yunpeng Xiao, Yuzhang Xie, Junhao Huang, Carl Yang, Xiao Hu, Kai Shu

The authors introduce a framework that gives multiple-choice models option-specific evidence to distinguish competing answers. An external language model extracts anchor cues, retrieves textbook passages linked to each cue and option, and distills them for a model. Across five open-source models, MedQA five-option accuracy averages 75.53% with chain-of-thought plus differential evidence, versus 71.64% with chain-of-thought alone; this is benchmark performance, not a clinical outcome.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.