What is
Reranker?
A second-stage model that reorders retrieved results by relevance to the query.
Definition
A reranker takes the top N results from a retriever and rescores them using a more accurate (and slower) cross-encoder model. The reranker considers the query and each candidate together, producing more accurate relevance scores than the initial retrieval. This dramatically improves the precision of the top few results — which is what the LLM actually sees.
Example
A RAG system retrieves 50 candidates with hybrid search, then a reranker like Cohere Rerank or BGE-Reranker reorders them and only the top 5 are sent to the LLM.
How Vedwix uses Reranker in client work
A reranker is the single highest-leverage RAG upgrade. We add one before tuning embeddings.
We ship this.
If you're building with Reranker in production, we can help — from architecture review to full implementation.
Brief us