RAG (Retrieval-Augmented Generation)

Definition: A technique where the model first retrieves relevant documents from a specific source — your files, a database, a website — and then generates its answer grounded in what it actually found, rather than relying only on what it memorized during training.

Why it matters: It sharply reduces hallucination for anything the retrieval step actually covers, because the model is summarizing a real document in front of it rather than generating an answer purely from memory — though it's only as accurate as the documents it retrieves from.

Example: A support chatbot that quotes your actual help-center articles when answering, instead of generating a plausible-sounding answer from general training data.