← Back to Glossary
Dense Retrieval
Phase 03 · RAGtechnique

Retrieving documents based on semantic meaning using vector embeddings rather than matching exact words.

Why it exists

Catches synonyms and conceptual matches (e.g. matching 'doctor' with 'physician') even when no exact words overlap.

How it works

Query and documents are passed through embedding models; vector distance search returns closest concept matches.

Where you'll use this