Combining keyword (BM25) and vector search for better recall.
1 min readupdated 2026-07-04
/ quick answer
Hybrid search runs a lexical query and a vector query in parallel, then fuses the scores. It catches exact-match terms that dense retrieval misses and semantic matches that keyword search misses. Combining keyword (BM25) and vector search for better recall.
Combining keyword (BM25) and vector search for better recall. Hybrid search runs a lexical query and a vector query in parallel, then fuses the scores. It catches exact-match terms that dense retrieval misses and semantic matches that keyword search misses. In practice: Query 'ACME-9000 firmware v2.3' hits BM25 on the SKU and dense retrieval on 'firmware update guide'. This dictionary node is part of the Onexial knowledge graph and links to related concepts, workflows and tools below.
Definition
Hybrid search runs a lexical query and a vector query in parallel, then fuses the scores. It catches exact-match terms that dense retrieval misses and semantic matches that keyword search misses.
Example
Query 'ACME-9000 firmware v2.3' hits BM25 on the SKU and dense retrieval on 'firmware update guide'.
Hybrid search runs a lexical query and a vector query in parallel, then fuses the scores. It catches exact-match terms that dense retrieval misses and semantic matches that keyword search misses.
What is an example of Hybrid Search?
Query 'ACME-9000 firmware v2.3' hits BM25 on the SKU and dense retrieval on 'firmware update guide'.
Why does Hybrid Search matter for AI and automation?
Combining keyword (BM25) and vector search for better recall. It connects to the workflows, prompts and tool stacks linked on this page, so you can move from definition to execution without leaving Onexial.