In the Pyrana versus naive RAG comparison, naive RAG uses text chunks as its retrieval unit, whereas the Pyrana Context Engine uses atomic CXUs so that retrieval operates on finer-grained knowledge objects.

4.4 Pyrana versus naive RAG

— Beyond Retrieval - Pyrana Context Engine Whitepaper.pdf

Retrieval unit

— Beyond Retrieval - Pyrana Context Engine Whitepaper.pdf

Naive RAG: Text chunk

— Beyond Retrieval - Pyrana Context Engine Whitepaper.pdf

Pyrana Context Engine: Atomic CXU

— Beyond Retrieval - Pyrana Context Engine Whitepaper.pdf