How does hybrid search merge results in Azure AI Search?
Why C is the answer
Reciprocal Rank Fusion
Hybrid search runs full-text and vector search in parallel, then merges results with RRF.
In detail
- Azure AI Search merges keyword and vector results with Reciprocal Rank Fusion.
- Rank positions work across different scoring systems.
Use case
An illustrative scenario, not a real incident.
A hybrid query returns two ranked lists. Reciprocal Rank Fusion merges them by rank, because their raw scores cannot be compared.
See it in different ways
flowchart TD A["Query"] --> B["Full-text search"] A --> C["Vector search"] B --> D["Reciprocal Rank Fusion"] C --> D D --> E["Semantic ranker"] E --> F["Top results"]
sequenceDiagram participant Q as Query participant K as Keyword search participant V as Vector search participant R as Ranking Q->>K: Search Q->>V: Search K->>R: Ranked results V->>R: Ranked results R->>R: Reciprocal Rank Fusion R-->>Q: Merged, then semantic ranker
flowchart LR S0["Keyword results"] --> S1["Vector results"] --> S2["Reciprocal Rank Fusion"] --> S3["Merged list"]
Key point: Hybrid search merges with Reciprocal Rank Fusion.
Why the other options are wrong
- A. By asking the model to pick the best results
- A model is not used to merge results.
- B. By keeping only the vector results
- Keeping only vector results loses the keyword matches.
- D. By adding the raw scores from each query
- Raw scores from different methods are not comparable.