RAG
RAG combines that retrieval step with generating a response.
Retrieval-augmented generation combines a generative model with retrieved external information. The retrieved material helps inform the response instead of relying only on information encoded in the model’s parameters.
[Lewis et al.]In practice
A museum assistant retrieves a catalog entry, then uses that entry to draft a visitor-facing answer about a painting.
[Lewis et al.]A little deeper
The original RAG research combines a pretrained generator with a retriever over an external index. Its reported improvements are experimental results, not a guarantee that every retrieved or generated claim is correct. [Lewis et al.]
A common mix-up
Using RAG guarantees a correct answer.
The retrieved evidence and the generated answer still need checking. [Lewis et al.]
Helpful to know: Language Model · Retrieval