Retrieval-augmented generation (RAG) combines a language model with a retrieval system so the model can reference specific documents or data sources when generating a response.
A standard language model generates responses based only on its training data. A RAG system first retrieves relevant documents from an external source — such as a vector database — and provides them as context. This allows the model to reference current or organization-specific information it was not trained on.
RAG reduces but does not eliminate hallucination risk. The quality of the source documents still matters.