Glossary

What is RAG (Retrieval-Augmented Generation)?

RAG (retrieval-augmented generation) is a method in which a language model first retrieves relevant passages from a document collection and then bases its answer on those passages. This lets the AI work with a company's own, current knowledge without retraining the model, and every statement can be traced back to a source.

In simple terms

Think of an open-book exam. Without RAG, the AI answers from memory: usually right, sometimes wrong, and with no evidence. With RAG, it first looks in the right binder, reads the relevant pages and only then writes the answer. Because it knows where it looked, it can name the page, and you can check for yourself whether the answer holds up.

In the enterprise

At the engineering firm Muster & Partner, thousands of expert reports, standards excerpts and project reports sit on the file server. A project manager asks: "Which fire safety requirements applied to the 2023 warehouse conversion?" The RAG system finds the relevant sections in the project files, the language model summarises them and names the file and page. The project manager opens the source and checks the quote before using it.

Related terms

Sources

  1. Lewis et al. (2020): Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (arXiv:2005.11401)

How we use this: UNID AI Second Brain · The second brain for your company's knowledge

← All terms

FAQ

Frequently asked questions

Does the AI model have to be trained on our documents for RAG?

No. With RAG the model stays unchanged. The documents are made searchable, and at query time only the relevant passages are passed to the model. New or changed documents are available as soon as they have been indexed.

See it working on your own material.

In a 30-minute call we show the products live and work out which use case makes sense for you first. No obligation, in German or English, in Vienna or online.

UNID AI · Leonard-Bernstein-Straße 10 · 1220 Vienna · +43 660 476 30 89