BenCao: An Instruction-Tuned Large Language Model for Traditional Chinese Medicine

TL;DR

This work addresses the challenge of applying LLMs to Traditional Chinese Medicine by introducing BenCao, a ChatGPT-based multimodal assistant tailored to TCM reasoning. BenCao combines a large, curated knowledge base with a four-scenario instruction framework, a chain-of-thought simulation for interpretable reasoning, and human expert feedback, all without parameter retraining. It leverages external APIs for tongue-image classification and multimodal data retrieval, achieving superior accuracy in diagnostics and herb recognition relative to general-domain and some TCM-domain models, and it has been deployed publicly via the GPTs Store with substantial early usage. The study demonstrates a practical, scalable pathway for domain-specific AI that harmonizes traditional medical reasoning with modern multimodal AI, offering educational, consultative, and health-guidance utilities while outlining clear avenues for broader validation and modality expansion.

Abstract

Traditional Chinese Medicine (TCM), with a history spanning over two millennia, plays a role in global healthcare. However, applying large language models (LLMs) to TCM remains challenging due to its reliance on holistic reasoning, implicit logic, and multimodal diagnostic cues. Existing TCM-domain LLMs have made progress in text-based understanding but lack multimodal integration, interpretability, and clinical applicability. To address these limitations, we developed BenCao, a ChatGPT-based multimodal assistant for TCM, integrating structured knowledge bases, diagnostic data, and expert feedback refinement. BenCao was trained through natural language instruction tuning rather than parameter retraining, aligning with expert-level reasoning and ethical norms specific to TCM. The system incorporates a comprehensive knowledge base of over 1,000 classical and modern texts, a scenario-based instruction framework for diverse interactions, a chain-of-thought simulation mechanism for interpretable reasoning, and a feedback refinement process involving licensed TCM practitioners. BenCao connects to external APIs for tongue-image classification and multimodal database retrieval, enabling dynamic access to diagnostic resources. In evaluations across single-choice question benchmarks and multimodal classification tasks, BenCao achieved superior accuracy to general-domain and TCM-domain models, particularly in diagnostics, herb recognition, and constitution classification. The model was deployed as an interactive application on the OpenAI GPTs Store, accessed by nearly 1,000 users globally as of October 2025. This study demonstrates the feasibility of developing a TCM-domain LLM through natural language-based instruction tuning and multimodal integration, offering a practical framework for aligning generative AI with traditional medical reasoning and a scalable pathway for real-world deployment.

Figures (2)

• Figure 1: Model performance across single-choice question categories in TCM. Accuracy of different LLMs on single-choice questions across seven TCM categories, including diagnostics, pharmacognosy, surgery, herbal formulas, and internal medicine. The x-axis indicates the question categories, and the y-axis shows the corresponding accuracy achieved by each model.
• Figure 2: Cross-task performance on the herb recognition and TCM constitution classification tasks. Each point denotes an individual LLM’s accuracy on two TCM tasks: herb recognition (x-axis) and TCM constitution classification (y-axis). The dashed diagonal indicates equal performance on both tasks; points toward the upper-right reflect stronger cross-task performance.