# ChatDoctor: A Medical Chat Model Fine-Tuned on a Large Language Model Meta-AI (LLaMA) Using Medical Domain Knowledge

**Authors:** Yunxiang Li, Zihan Li, Kai Zhang, Ruilong Dan, Steve Jiang, You Zhang  
**arXiv:** [2303.14070](https://arxiv.org/abs/2303.14070)  
**Date:** 2023-03-24  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.14070v5/

## Key points

- Research aimed to overcome limitations of existing large language models (LLMs) in providing accurate medical advice
- Developed ChatDoctor, a specialized language model refined and adapted from the large language model meta-AI (LLaMA)
- Used dataset of 100,000 patient-doctor dialogues obtained from an online medical consultation platform for refinement process
- Conversations were cleaned and anonymized to ensure privacy
- Incorporated self-directed information retrieval mechanism into ChatDoctor, allowing access to real-time information from reliable online sources and curated offline medical databases
- Significant improvements observed in accuracy of responses after fine-tuning model with real-world patient-doctor interactions and equipping it with self-directed information retrieval capabilities
- ChatDoctor represents notable advancement in medical LLMs, demonstrating enhanced understanding of patient inquiries and providing accurate advice
- Improvements in providing reliable and precise information are invaluable in high-stakes field of medicine where errors can have serious consequences
- Article published under open-access license, allowing unrestricted use, distribution, and reproduction with proper attribution to original authors and source.

## Summary

The research aimed to overcome the limitations of existing large language models (LLMs) in providing accurate medical advice. To achieve this, the researchers developed ChatDoctor, a specialized language model that was refined and adapted from the large language model meta-AI (LLaMA). The refinement process involved using a dataset of 100,000 patient-doctor dialogues obtained from an online medical consultation platform. These conversations were carefully cleaned and anonymized to ensure privacy. In addition to model refinement, the researchers incorporated a self-directed information retrieval mechanism into ChatDoctor. This mechanism allowed the model to access real-time information from reliable online sources like Wikipedia and curated offline medical databases. By fine-tuning the model with real-world patient-doctor interactions and equipping it with self-directed information retrieval capabilities, significant improvements were observed in the accuracy of its responses. ChatDoctor represents a notable advancement in medical LLMs as it demonstrates enhanced understanding of patient inquiries and provides accurate advice. In the high-stakes field of medicine where errors can have serious consequences, such improvements in providing reliable and precise information are invaluable. The article is published under an open-access license, allowing unrestricted use, distribution, and reproduction as long as proper attribution is given to the original authors and source.