Abstract:Objective：To construct a new virtual patient（VP）system based on artificial intelligence（AI）large language models（LLMs）technology，which can be used for medical education，and to explore the application effect of the system in standardized training for surgical residents. Method：Professional master’s students who rotated in the Department of Surgery at the First Affiliated Hospital of Nanjing Medical University from September 2024 to October 2024 were selected as the research subjects（N=20），randomly divided into an experimental group and a control group，and underwent a one week clinical consultation training in surgery simultaneously. The experimental group used LLM virtual patient system teaching format，while the control group used traditional infusion teaching. After the course，both groups underwent a 2-month clinical rotation. After the rotation，evaluate the quality of medical records and patient satisfaction of 20 students’ bed medical records within 2 months. Result：There was no statistically significant difference between the two groups in terms of age，gender，education，and test scores（P > 0.05）. The average score of hospitalization medical record quality for the experimental group was 86.2 points，which was higher than the control group’s 80.4 points（P < 0.05）. In terms of patient satisfaction，the experimental group also scored an average of 84.5 points higher than the control group’s 78 points（P < 0.05）. Conclusion：The VP system based on LLM has improved students’ mastery of consultation and admission record writing，and also promoted the cultivation of their clinical thinking ability. This teaching method can provide new teaching tools and ideas for other medical higher education groups.