Coarse-to-Fine Vector Retriever

Software componentKnowledge & DataKnowledge & Dataarc:CoarseToFineVectorRetriever

A vector retriever that first searches truncated low-dimensional embeddings to shortlist candidates, then rescores the shortlist with full-dimensional embeddings.

Responsibility. Shortlists candidates with low-dimensional vectors and refines their ranking with full-dimensional vectors.

Also known as: Adaptive retrieval, Two-stage MRL retrieval

Variant of Vector Retriever abstract

When to choose. Choose when embeddings are MRL-trained and corpus scale makes full-dimensional search latency or memory prohibitive.

Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

Reranker abstract Ch6.1

Quantitative guidance

As stated by the sources; verify before use.

Two-stage 256-dim/3,072-dim retrieval typically cuts end-to-end latency by 40-60% vs. full-dimensional retrieval (Ch6.1).

Classification

Patterns: Matryoshka Representation LearningTwo-stage retrieval (low-dim first pass, high-dim rerank)
Quality attributes: Performance efficiency (ISO/IEC 25010)

Sources

Ch6.1: T. Nguyen, "Embeddings and RAG Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 6.1. ISBN: 9798244538229.