databricks Platform · Updated Apr 13, 2026

Vector Search retrieval quality (Beta)

Mosaic AI Vector Search provides built in retrieval quality evaluation that measures and compares the relevance of different search strategies on your data.

Beta Open docs ↗release note ↗

What it does

Mosaic AI Vector Search provides built in retrieval quality evaluation that measures and compares the relevance of different search strategies on your data. Beta This feature is in Beta. Workspace admins can control access to this feature from the Previews page.

Why it matters

Score relevance: An LLM judge evaluates every query and retrieved document pair on a 4 point relevance scale.

Key capabilities

Highly Relevant %: Results scoring

Highly Relevant %: Results scoring 3 (direct answers).

Relevant+ %: Results scoring 2

Relevant+ %: Results scoring 2 or higher (useful).

Not Relevant %: Results scoring

Not Relevant %: Results scoring 0 or 1 (not useful). When to use: To understand the shape of quality distribution. Two…

Architecture

Databricks Platform → Platform

In short

Mosaic AI Vector Search provides built in retrieval quality evaluation that measures and compares the relevance of different search strategies on your data. Beta This feature is in Beta.

Read the full documentation ↗

Key takeaway

Mosaic AI Vector Search provides built in retrieval quality evaluation that measures and compares the relevance of different search strategies on your data.

2026-04-13-vector-search-retrieval-quality-beta Generated by the Databricks Updates Agent · browse all updates