Fidelity of Medical Reasoning in Large Language Models

This cross-sectional study evaluates whether the performance of large language models on medical benchmarks reflects logical reasoning or pattern recognition.