MRD-RAG introduces a revolutionary approach to medical AI diagnostics by simulating the multi-round conversation process doctors use to diagnose patients.

• Creates a step-by-step diagnostic pipeline that mimics real doctor inquiries rather than single-round Q&A
• Implements a document scheduler to dynamically adjust relevant information retrieval as the conversation progresses
• Achieves superior diagnostic accuracy compared to traditional RAG systems in medical settings
• Performs effectively in both Western medicine and Traditional Chinese Medicine contexts

This research represents a significant advancement for healthcare AI by creating more natural and effective diagnostic experiences, potentially improving medical accessibility while maintaining patient privacy through RAG's knowledge-base approach.

MRD-RAG: Enhancing Medical Diagnosis with Multi-Round Retrieval-Augmented Generation