What is retrieval-augmented generation (RAG)?

Retrieval-augmented generation (RAG) is a way of answering questions with a language model in two steps: first retrieve the passages from real documents that are closest to the question, then have the model answer from those passages. Because the answer is built on retrieved text, it can cite its sources, and it invents less than a model answering from memory.

A RAG system splits documents into passages, turns each into an embedding, and stores them in a vector index. A question is embedded the same way, and the nearest passages are sent to the model with instructions to answer only from them.

RAG doesn’t remove mistakes: retrieval can miss the right passage, and a model can still misread one. That is why answers should cite the passages they used.