Semantic Search Guide

Home / Reranking

Reranking guides

7 published guides. Start with What is reranking.

Start hereWhat is reranking

Reranking rescores the top results with a stronger model. How it fits after retrieval and what it gains.

Learning path

All guides

Understand it1 guides

Start hereWhat is rerankingReranking rescores the top results with a stronger model. How it fits after retrieval and what it gains.

Build and run it1 guides

TutorialCross-encoder rerankersSetting up a cross-encoder reranker, how many candidates to rerank and the latency it adds.

Compare and choose2 guides

ComparisonBest reranking modelsHosted and open rerankers compared on accuracy, latency, cost and language support.Decision guideReranking latency and costMeasuring the latency and cost reranking adds and finding the candidate count that pays off.

How it works3 guides

ExplainerLLM rerankingUsing an LLM as a reranker, the three prompting styles and when it is worth the cost.ExplainerReranking in RAGWhy reranking is often the biggest single accuracy gain in RAG and how to add it.ExplainerMaximal marginal relevanceHow MMR balances relevance against redundancy so the top results are not five copies of one idea.

Related sections

Query understanding How it works Embeddings Keyword search