What does a reranker do in a RAG pipeline?
Why C is the answer
It reorders retrieved results by relevance to the query
Reranking improves which passages reach the model.
In detail
- A reranker reorders results by relevance to the query.
- Rerank a wider set, then keep the best few.
Use case
An illustrative scenario, not a real incident.
Retrieval returns ten passages, with the best at number eight. A reranker moves it to the top before the model reads it.
See it in different ways
mindmap
root((RAG))
Ingest
Sync data source
Detect personal data
Chunk
By structure
Hierarchical
Search
Vector
Keyword
Rank
Rerankerflowchart LR S0["Retrieve candidates"] --> S1["Rerank"] --> S2["Keep the best"] --> S3["Generate"]
Key point: Rerankers improve what reaches the model.
Why the other options are wrong
- A. It generates the embeddings for each chunk when data is ingested
- It works after retrieval, not at ingestion.
- B. It chooses the AWS Region where the knowledge base is deployed
- It has nothing to do with Regions.
- D. It removes stale documents from the source data on a schedule
- It does not change source data.