Reproduce “BenCao: An Instruction-Tuned Large Language Model for Traditional Chinese Medicine” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Jiacheng Xie, Yang Yu, Yibo Chen, Hanyao Zhang, Lening Zhao, Jiaxuan He, Lei Jiang, Xiaoting Tang and 2 more

BenCao is an instruction-tuned, multimodal large language model for Traditional Chinese Medicine (TCM) that integrates textual and visual inputs, leveraging a comprehensive knowledge base and expert human feedback. It provides context-aware, interpretable reasoning for TCM theory, health guidance, and preliminary discomfort conditioning, demonstrating superior performance in domain-specific tasks compared to general-purpose LLMs.

Key insights

  • Achieving expert-level reasoning and alignment in specialized medical domains is feasible and cost-effective through natural language instruction tuning and human feedback, without needing extensive parameter retraining.
  • Multimodal integration, specifically incorporating visual cues like tongue images with textual data, is crucial for developing AI systems that truly align with the diagnostic practices of Traditional Chinese Medicine.
  • Transparent Chain-of-Thought mechanisms, combined with strict safety disclaimers and uncertainty-aware safeguards, are essential for building trustworthy and ethically compliant medical AI applications.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.