# ChatDoctor: A Medical Chat Model Fine-tuned on LLaMA Model using Medical Domain Knowledge

**Authors:** Li Yunxiang, Li Zihan, Zhang Kai, Dan Ruilong, Zhang You  
**arXiv:** [2303.14070](https://arxiv.org/abs/2303.14070)  
**Date:** 2023-03-24  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.14070v1/

## Key points

- Recent advancements in large language models (LLMs) like ChatGPT have shown success in understanding instructions and generating human-like responses.
- Models not specifically trained for the medical domain lead to inaccuracies in diagnostic accuracy and treatment recommendations.
- A dataset of over 700 diseases with symptoms, medications, and tests was collected to address this issue.
- Integration of advanced language models into healthcare can revolutionize communication between professionals and patients.
- Open-source code, datasets, and model weights encourage dialogue model development in the medical domain.
- A novel process framework for fine-tuning LLMs in the medical field was introduced.
- ChatDoctor demonstrates higher diagnostic accuracy compared to general-domain LLMs like ChatGPT.
- ChatDoctor has the potential to improve operational efficiency in healthcare settings with limited resources.

## Summary

Recent advancements in large language models (LLMs) like ChatGPT have shown great success in understanding instructions and generating human-like responses. These models have not been specifically trained for the medical domain, leading to inaccuracies in diagnostic accuracy and recommendations for medical treatments. To address this issue, a dataset of over 700 diseases with symptoms, recommended medications, and necessary medical tests was collected. Additionally, 5,000 doctor-patient conversations were generated by fine-tuning models on this data. The integration of these advanced language models into healthcare has the potential to revolutionize communication between healthcare professionals and patients. By understanding patients' needs and providing informed advice, these models can enhance various aspects of medical-related fields. The availability of open-source code, datasets, and model weights further encourages the development of dialogue models in the medical domain. The study introduced a novel process framework for fine-tuning large language models in the medical field and validated the effectiveness of ChatDoctor through real-world clinical applications. By leveraging existing conversation datasets and training procedures, ChatDoctor demonstrates higher diagnostic accuracy compared to general-domain LLMs like ChatGPT. Overall, the utilization of advanced language models like ChatDoctor holds promise for improving operational efficiency in healthcare settings where resources are limited. With its ability to assist in initial diagnosis and patient triage, ChatDoctor showcases the potential to enhance overall quality of care and patient outcomes in the medical field.