Hybrid search

LLM foundationsRetrieval and dataPublished By Simon Budziak

Hybrid search runs keyword search and vector search for the same query, then combines their ranked results into one list. It gives a RAG system both exact matching for names, codes, and rare terms and semantic matching for passages that express the same meaning with different words.

Azure AI Search runs full-text and vector queries in parallel and merges them with reciprocal rank fusion. Other search engines expose similar combinations with different tuning controls.

How does hybrid search improve RAG?

Keyword search protects exact matches while semantic search broadens recall. The combined list is more robust across mixed business data, where a question may contain both natural language and an invoice number. A vector database must support the text and vector paths or delegate one of them to another index.

Does hybrid search remove the need for reranking?

No. Fusion combines initial rankings, while reranking applies a stronger second-pass model to the leading candidates. Measure the complete RAG pipeline, because better retrieval can still produce poor answers when chunking or generation fails. Metadata filtering should also run before results cross tenant or permission boundaries.

Frequently asked questions

Why is hybrid search often better than vector search alone?

Vector search handles meaning well but can miss exact identifiers. Keyword search catches exact terms, while the combined ranking benefits from both signals.

How are hybrid search results combined?

A common method is reciprocal rank fusion, which gives each document points based on its position in every result list and then sums those scores.

Summarize this page with

Train your team to build this