Corrective RAG

Corrective RAG is a retrieval-augmented generation approach that evaluates the quality of retrieved documents and, when they seem irrelevant or insufficient, takes corrective steps such as refining the query or searching elsewhere.

Corrective RAG sits in the Retrieval & memory part of the Agentik {OS} glossary, which defines the words used to build and run AI agent systems.

Also called CRAG.