不同生成式大语言模型在中医神志病诊疗领域的应用能力比较研究OA

With the rapid advancement of artificial intelligence,generative large language models(LLMs)have attracted growing attention for their potential applications in the medical field.This study evaluates six representative generative LLMs—including four general-purpose models and two specialized models for traditional Chinese medicine(TCM)—on a constructed dataset of TCM medical records related to mental disorders.A comprehensive comparison was conducted from two aspects:the models'diagnostic and therapeutic capabilities,and the quality of their generated outputs,aiming to assess their applicability in the diagnosis and treatment of mental disorders within TCM.The results show that while all models demonstrate certain competence in this domain,their performance varies significantly.Under the specific test conditions and specific dataset of this study,DeepSeek,Ernie Bot,and ChatGLM exhibited relatively outstanding overall applicability.Specialized TCM models,however,still show room for improvement in areas such as depth of data processing,integration of domain knowledge,and clinical logical reasoning.Looking forward,with the continuous expansion of high-quality TCM datasets,ongoing optimization of model architectures,and the gradual improvement of clinical validation systems,the application prospects of generative LLMs in the field of TCM are expected to broaden considerably.