Traditional Chinese Medicine Instruction Data for Multimodal LLM Fine-Tuning

245,000 instruction examples across text, visual, and signal modalities support the fine-tuning of ShizhenGPT, a specialized model for Traditional Chinese Medicine. FreedomIntelligence created and released this collection, with its latest update in August 2025.

Use Cases

• Fine-tuning a multimodal LLM to generate TCM text instructions based on visual inputs like tongue images.
• Training a model to correlate TCM signal instructions, such as pulse readings, with corresponding diagnostic text.
• Developing a retrieval-augmented generation system for TCM knowledge using the dataset's structured instruction-response pairs.
• Benchmarking the performance of multimodal models on domain-specific tasks involving TCM text, images, and signals.

Strengths

• Contains 245,000 total instruction examples for model training.
• Integrates three distinct data modalities: text, visual, and signal instructions.

Limitations

• The exact distribution of examples across the three modalities is unspecified, making task-specific volume unclear.
• No sample data or column details are provided to assess data structure or quality.
• Potential geographic or cultural bias exists as the data is focused on Traditional Chinese Medicine practices.

Provenance

Source

FreedomIntelligence.

Collection Method

Created for fine-tuning the ShizhenGPT model; specific collection methodology is not detailed.

Freshness

Last updated on 2025-08-25.