RAG connects generation to retrieved documents
A research paper combines document retrieval with language generation.
A historical account of this event. Usage notes are YLEM editorial analysis; check the provider for current features and prices.
What changed
RAG combines a language model with a retrievable document index. The paper studies knowledge-intensive tasks using a Wikipedia index, giving generated answers access to information outside model parameters.
Sources: [1] Lewis et al. · RAG paper
What this means in practice
For a document assistant, ask for the supporting passage as well as the answer. A useful trial is to compare answers to the same question with and without the relevant document.
How to read the result
Retrieval and generation are separate failure points: the system may miss a passage or misread it. Check whether each citation actually supports the adjacent claim.