【摘要】 为评估人工智能语言模型ChatGPT在医学考试中的表现，采用横断面研究方法，将ChatGPT回答的心血管主治医师资格考试的模拟试题结果与标准答案进行比对。在排除2道图片题后，ChatGPT正确回答了333道问题中的189道，准确率为56.8%，其在四个单元的回答正确率分别为67.0%、61.0%、58.2%及11.4%。逻辑错误是最常见的回答错误原因。题目属于第四单元、案例分析题是回答错误的影响因素。研究显示ChatGPT未通过心内科主治医师资格模拟考试，尚不推荐将其用于心血管内科医师的日常教育。更多还原

【Abstract】 In order to evaluate the performance of the artificial intelligence language model ChatGPT in medical examinations, a cross-sectional study was conducted comparing ChatGPT’s answers to a set of simulated tests of Board Certification Examination for Cardiovascular Medicine with the standard answers. After excluding 2 picture questions, ChatGPT correctly answered 189 out of 333 questions, with an accuracy rate of 56.8%. The correct response rates across four sections were 67.0%,61.0%, 58.2% and 11.4% respectively. Logical errors were the most common cause of incorrect answers. Factors influencing incorrect responses included questions from the fourth section and case analysis questions. Research shows that ChatGPT failed to pass the simulation examination of Board Certification Examination for Cardiovascular Medicine, and it is not recommended to be used in the routine education of cardiovascular physicians.更多还原