Re-ranking with Cross-Encoders for Better AI Database Retrieval
Re-ranking with cross-encoders is a practical way to improve the precision of AI database search without asking the…
Read moreKnowledge Base
Learn
Structured articles covering everything from core concepts to production operations.
Voronoi Cells and Space Partitioning in Vector Search
Voronoi cells are one of the simplest ways to understand how many vector indexes make search faster: a…
Read more
How to Measure Recall@K
Recall@K measures how many of the true top K nearest neighbors a vector search index returns for a…
Read more
Recall vs Latency: Understanding the ANN Trade-off
Approximate nearest neighbor indexes trade some search accuracy for speed because they avoid comparing a query vector with…
Read more
How k-Nearest Neighbours Works
k-nearest neighbours, often shortened to k-NN, is the retrieval operation that finds the k stored items most similar…
Read more
Maximum Inner Product Search (MIPS) in AI Databases
Maximum Inner Product Search, usually shortened to MIPS, is the task of finding the stored vectors that produce…
Read more
ACORN and Adaptive Filtered Traversal in Vector Search
ACORN is a filtered vector search approach designed to keep graph-based nearest neighbor search fast when metadata filters…
Read more
Filtered ANN Search Algorithms: Why Metadata Constraints Change Vector Search
Filtered approximate nearest neighbor search is the problem of finding the most similar vectors while also honoring metadata…
Read more
Graph-Based vs Cluster-Based Indexes: HNSW vs IVF for Vector Search
HNSW and IVF are two common ways to make vector search fast without scanning every embedding. HNSW is…
Read more
The Flat Brute-Force Index Explained
A flat, or brute-force, index is the simplest way to perform vector search: store every vector as-is, compare…
Read more