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

• DOI: 10.48550/arxiv.2303.14070
• Publication Date: 2023-01-01
• Authors: Yunxiang Li; Li Zihan; Kai Zhang; Dan Ruilong; You Zhang
• Keywords: Medical advice; Advice (programming); Medical record

The primary aim of this research was to address the limitations observed in medical knowledge prevalent large language models (LLMs) such as ChatGPT, by creating a specialized model with enhanced accuracy advice. We achieved adapting and refining meta-AI (LLaMA) using dataset 100,000 patient-doctor dialogues sourced from widely used online consultation platform. These conversations were cleaned anonymized respect privacy concerns. In addition refinement, we incorporated self-directed information retrieval mechanism, allowing access utilize real-time sources like Wikipedia data curated offline databases. fine-tuning real-world interactions significantly improved model's ability understand patient needs provide informed By equipping reliable sources, substantial improvements its responses. Our proposed ChatDoctor, represents significant advancement LLMs, demonstrating improvement understanding inquiries providing accurate Given high stakes low error tolerance field, enhancements are not only beneficial but essential.