MUVERA
Definition
MUVERA (Multi-Vector Retrieval via Fixed Dimensional Encodings) collapses a Late Interaction multi-vector representation (e.g. ColBERT’s per-token matrix) into a single fixed-dimensional vector that approximates the MaxSim score under an ordinary inner product. This lets a multi-vector document be retrieved with standard single-vector ANN (HNSW) in the first stage, sidestepping the cost of full MaxSim during candidate generation.
Why It Matters
The central problem of late interaction is that MaxSim compares every query vector against every document vector — too expensive for first-stage retrieval over a large corpus. MUVERA provides a cheap, ANN-indexable proxy:
- Stage 1 — encode query and documents as fixed-dimensional MUVERA vectors, retrieve top-k candidates with fast single-vector ANN.
- Stage 2 — rerank those candidates with the full multi-vector MaxSim (ColBERT).
This is the single-vector “approximation of the multivector” that plays the same role as Elastic’s average vectors — a compressed stand-in for the first stage.
In Practice
Qdrant Vector DB supports MUVERA as a first-stage prefetch representation, then reranks with a ColBERT multivector field in the same Query API call (using="muvera" → using="colbert"). See Late Interaction in Qdrant.
Related Concepts
- Late Interaction — the architecture MUVERA accelerates
- ColBERT — the multi-vector model MUVERA approximates
- HNSW — single-vector ANN index MUVERA makes usable for late interaction
- Token Pooling — alternative multi-vector compression (clusters vectors rather than collapsing to one)
- Bi-Encoder — the single-vector baseline MUVERA imitates for stage 1
Related Topics
- Late Interaction in Qdrant — where MUVERA is used as the prefetch stage
- Late Interaction in Elasticsearch — “average vectors” play the analogous role there
- Vector Search Tradeoffs — the one-vector-or-many axis, and why the multi-vector cost is moved to stage two