Omni-RAG introduces a comprehensive framework for medical LLMs that intelligently selects and retrieves the most appropriate knowledge sources based on the specific context.

• Addresses hallucination issues in medical LLMs by implementing context-appropriate knowledge retrieval
• Introduces MedOmniKB, a specialized repository of diverse medical knowledge sources
• Frames knowledge acquisition as a source planning problem that determines which references to use when
• Demonstrates improved performance across medical diagnosis, clinical decision-making, and health inquiries

This research is critical for healthcare applications where factual accuracy can directly impact patient outcomes and clinical decisions. By developing more reliable AI systems for medical contexts, Omni-RAG helps bridge the gap between AI capabilities and healthcare requirements.

Towards Omni-RAG: Comprehensive Retrieval-Augmented Generation for Large Language Models in Medical Applications