Medical Reasoning with Large Language Models: A Survey and MR-Bench

Develop an open-source evaluation suite for medical reasoning that moves beyond multiple-choice to verify chains of thought. This provides the community a standard for clinical-grade AI.

Suggested repo: med-reason-bench

"Stop grading medical AI on trivia; grade them on clinical reasoning."

Estimated effort: 45h