RAG (Retrieval-Augmented Generation)

A technique where a model first searches external documents, then uses what it finds to craft its answer — reducing hallucinations.