Glossary · RAG

Reranker

Also known as: Re-ranker, Cross-encoder reranker

Two-stage retrieval: cheap retriever shortlists, expensive reranker reorders for higher precision.

A second-stage model in a retrieval pipeline that takes the top-K candidates from a cheap retriever (BM25 / dense) and re-scores them with a more expensive cross-encoder or LLM. Dramatically improves top-N precision at modest extra cost.

In practice

Standard quality lever for production RAG. Interviews probe cohere-rerank vs ColBERT vs LLM-as-judge approaches.

Comparisons that include Reranker

Reranker vs RetrieverRetrievers fetch a shortlist fast; rerankers score the shortlist accurately. Production RAG uses both. A bi-encoder retrieves the top 100, a cross-encoder reranks to the top 5.

Related terms

Retrieval-Augmented Generation (RAG)

Fetch relevant docs at inference time and stuff them into the prompt so the model can answer from real data.

Chunking

Cutting documents into retrievable pieces before embedding them.

Context Window

The max number of tokens a model can attend to at once.

Hallucination

When a model confidently makes up something that isn't true.

AI System Design

End-to-end design of production LLM systems: ingestion, retrieval, serving, eval, monitoring.

HNSW (Hierarchical Navigable Small World)

A graph-based ANN index with log-time approximate search, excellent recall/latency trade-off.