Jingfang: An LLM-Based Multi-Agent System for Precise Medical Consultation and Syndrome Differentiation in Traditional Chinese Medicine

TL;DR

This work presents JingFang, an LLM-based multi-agent system for Traditional Chinese Medicine that addresses the limitations of existing TCM LLMs by enabling patient-tailored, multi-round consultations and explicit syndrome differentiation. It introduces the Multi-Agent Collaborative Consultation Mechanism (MACCM) to coordinate specialist and general agents, a Syndromes Agent trained on a curated dataset, and a Dual-Stage Retrieval Scheme (DSRS) within the Treatment Agent to deliver syndrome-specific prescriptions. Experimental results show JingFang achieves substantial gains in syndrome differentiation precision (up to 124% over domain-specific baselines and 21.1% over SOTA models) and robustness in multi-round consultations, with strong semantic and lexical fidelity when the General Agent is included. The framework promises practical impact for AI-assisted TCM by providing interpretable reasoning, targeted questioning, and precise, syndrome-based treatments, and suggests future integration of broader multimodal diagnostics and enhanced collaboration among agents.

Abstract

The practice of Traditional Chinese Medicine (TCM) requires profound expertise and extensive clinical experience. While Large Language Models (LLMs) offer significant potential in this domain, current TCM-oriented LLMs suffer two critical limitations: (1) a rigid consultation framework that fails to conduct comprehensive and patient-tailored interactions, often resulting in diagnostic inaccuracies; and (2) treatment recommendations generated without rigorous syndrome differentiation, which deviates from the core diagnostic and therapeutic principles of TCM. To address these issues, we develop \textbf{JingFang (JF)}, an advanced LLM-based multi-agent system for TCM that facilitates the implementation of AI-assisted TCM diagnosis and treatment. JF integrates various TCM Specialist Agents in accordance with authentic diagnostic and therapeutic scenarios of TCM, enabling personalized medical consultations, accurate syndrome differentiation and treatment recommendations. A \textbf{Multi-Agent Collaborative Consultation Mechanism (MACCM)} for TCM is constructed, where multiple Agents collaborate to emulate real-world TCM diagnostic workflows, enhancing the diagnostic ability of base LLMs to provide accurate and patient-tailored medical consultation. Moreover, we introduce a dedicated \textbf{Syndrome Differentiation Agent} fine-tuned on a preprocessed dataset, along with a designed \textbf{Dual-Stage Recovery Scheme (DSRS)} within the Treatment Agent, which together substantially improve the model's accuracy of syndrome differentiation and treatment. Comprehensive evaluations and experiments demonstrate JF's superior performance in medical consultation, and also show improvements of at least 124% and 21.1% in the precision of syndrome differentiation compared to existing TCM models and State of the Art (SOTA) LLMs, respectively.

Figures (8)

• Figure 1: Overall workflow framework of JingFang, including three main modules: TCM Consultation, TCM Syndrome Differentiation, and TCM Treatment, which enables the model to simulate the real-world diagnosis and treatment scenarios of TCM.
• Figure 2: An example of the workflow of the proposed MACCM that enhances the accurate and targeted medical consultation.
• Figure 3: Comparison of multi-round consultation capabilities among different models.
• Figure 4: An ablation experiment on syndrome differentiation precision of different foundation models that are integrated with the developed multi-agent system, which demonstrates the effectiveness and applicability of the developed method and dataset in this work.
• Figure 5: Comparison result of the retrieval similarity between the DSRS and the Single-Stage approach in a total of 100 TCM clinical cases.