Candidate depth 10→500 lifts score by just 0.01: Qdrant on diagnosing before tuning vector search

qdrant_engine · x · 2026-09-08

Qdrant's team argues that vector search can "work" while still being badly tuned — with knobs like hnswef, candidate depth, RRF k, quantization, and reranking, it's unclear which to change first. Testing across five public datasets, they found the right knob depends on where the problem is:

Their new article lays out a systematic approach to retrieval tuning without guessing.

Original post →

More from coding & agent

coding & agent channel →