The paper "Leveraging Large Language Model as Simulated Patients for Clinical Education" by Yanzeng Li, Cheng Zeng, Jialun Zhong, Ruoyu Zhang, Minhao Zhang, and Lei Zou explores the integration of large language models (LLMs) into clinical medical education to simulate patient interactions. The authors introduce CureFun, a model-agnostic framework that leverages LLMs to create realistic and professional dialogue flows between students and simulated patients (SPs). CureFun enhances the dialogue quality and coherence through a graph-driven context-adaptive mechanism, which dynamically adjusts the dialogue flow based on a structured case graph. The framework also includes an automatic assessment module that evaluates students' performance using multi-granularity assessment items, providing reliable and accurate scores. The study demonstrates that CureFun significantly improves the performance of LLMs in role-playing as SPs, outperforming other LLM-based chatbots in terms of authenticity and professionalism. Additionally, the framework is evaluated for its ability to assess LLMs as virtual doctors (VDs), highlighting their diagnostic capabilities and limitations. The results show that while LLMs excel in generating detailed and ethical responses, they still fall short of human experts in real-world medical scenarios. The paper concludes by discussing the potential of LLMs as pre-diagnostic and triage tools, emphasizing the need for further development and integration of VSP and VD tasks to enhance clinical education.The paper "Leveraging Large Language Model as Simulated Patients for Clinical Education" by Yanzeng Li, Cheng Zeng, Jialun Zhong, Ruoyu Zhang, Minhao Zhang, and Lei Zou explores the integration of large language models (LLMs) into clinical medical education to simulate patient interactions. The authors introduce CureFun, a model-agnostic framework that leverages LLMs to create realistic and professional dialogue flows between students and simulated patients (SPs). CureFun enhances the dialogue quality and coherence through a graph-driven context-adaptive mechanism, which dynamically adjusts the dialogue flow based on a structured case graph. The framework also includes an automatic assessment module that evaluates students' performance using multi-granularity assessment items, providing reliable and accurate scores. The study demonstrates that CureFun significantly improves the performance of LLMs in role-playing as SPs, outperforming other LLM-based chatbots in terms of authenticity and professionalism. Additionally, the framework is evaluated for its ability to assess LLMs as virtual doctors (VDs), highlighting their diagnostic capabilities and limitations. The results show that while LLMs excel in generating detailed and ethical responses, they still fall short of human experts in real-world medical scenarios. The paper concludes by discussing the potential of LLMs as pre-diagnostic and triage tools, emphasizing the need for further development and integration of VSP and VD tasks to enhance clinical education.