Retrieval-Augmented Generation (RAG)

A technique that lets the LLM ground its responses in retrieved documents (knowledge base, FAQs) instead of relying on training data alone.