Large language models excel at answering medical questions but fail to recognize their knowledge gaps, posing risks in ...