LingLanMedBench: A Unified Benchmark for Evaluating Large Language Models in Traditional Chinese Medicine
LingLanMedBench created a unified, large-scale benchmark to evaluate how well Large Language Models understand and reason about Traditional Chinese Medicine (TCM). It combines 13 subtasks across five domains like knowledge and clinical decision-making. The results show models excel at basic recall but struggle with complex reasoning, highlighting a gap between factual knowledge and clinically faithful application.