MedRAGChecker: Claim-Level Verification for Biomedical Retrieval-Augmented Generation
MedRAGChecker is a system for medical RAG that verifies claims using Natural Language Inference and structured knowledge from DRKG. It employs distilled student models to efficiently provide detailed diagnostics like hallucination rates and reveals distinct risk profiles across different AI generators. The hosts conclude this enables reliable, verifiable information for safe clinical decision support systems.