retrieval-augmented generation (RAG)
Looking up relevant text from an external source and injecting it into the prompt before generating an answer, instead of relying only on the model's own trained-in knowledge.
Retrieval-augmented generation (RAG) answers a query by first retrieving relevant passages from an external document collection (via dense or hybrid search over chunked, embedded documents), injecting the retrieved text into the prompt as context, and then generating an answer grounded in that context -- addressing both hallucination and the knowledge cutoff by giving the model current, specific, sourced information it wasn't necessarily trained on.
Examples
Instead of guessing from trained-in memory, a RAG system answers "what's our refund window?" by first retrieving the company's actual return-policy document and grounding its answer in that text.