RAG (retrieval-augmented generation)
RAG, retrieval-augmented generation, is a method for looking up outside information and handing it to a model before it answers.
What it tells you about a developer
RAG experience means the developer has combined search with a model, which involves databases and data pipelines alongside the AI part. It is one of the most common ways companies put a model on top of their own data.
- Type
- Framework
- Released
- 2020
- Used in
- AI & machine learning
- Applied to
- AI agents & tooling
RAG stands for retrieval-augmented generation. Researchers at Meta's AI research lab described the method in a 2020 paper: retrieval means searching a store of documents for the relevant ones, and generation is the model writing an answer, so RAG augments, or adds to, what the model generates with what was retrieved. A developer who builds RAG systems is usually solving the problem of a model needing facts, like a company's own documents, that it was never trained on.