TechPrufer
Get started
Weaviate
Weaviate logoWeaviate · Feature trail

PathSeer: Adaptive Filtered Vector Search

1 post · Oct 2026

What this trail covers

PathSeer is a new predicate-agnostic filtered vector search algorithm integrated into Weaviate's HNSW index. It dynamically adapts its search strategy per node, combining 'distance first' and 'filter first' approaches to optimize performance across varying filter selectivities and correlations. This aims to improve throughput and recall for filtered vector searches, offering a more efficient alternative to existing sweeping and ACORN strategies.

All Weaviate trails
  1. PathSeer: Adaptive Filtered Vector Search | WeaviateThis post introduces PathSeer, a novel algorithm for filtered vector search in HNSW. It details the algorithm's two phases: a 'two-hop burst' phase while the result heap is filling, and a 'skip distances between non-matching nodes' phase once the heap is full. The implementation in Weaviate is compatible with existing HNSW indices and does not require a rebuild. Benchmarks demonstrate significant throughput improvements over sweeping and ACORN across various datasets and filter selectivities, particularly in correlated filter scenarios.