Language & LLMs

What Is Dense Retrieval?

Dense retrieval represents both queries and documents as dense embedding vectors and finds matches by comparing these vectors. It captures semantic similarity, allowing relevant results even without shared keywords. It contrasts with sparse retrieval methods that rely on exact term matching.