Retrieval-augmented generation (RAG)

Retrieval-augmented generation (RAG)

A technique that combines a large language model with an external, searchable source of information, so the model can retrieve relevant facts or documents before generating its answer, rather than relying solely on what it learned during training. RAG is commonly used to reduce hallucination and to let a model answer questions about material published after its training cut-off. (See also: Hallucination, Embeddings)