Decoder. plain-English AI glossary

Reranking

▲ Rising

A second-pass sort that takes the retriever’s rough shortlist and reorders it by who actually answers the question best.

Think of it like

A hiring manager reading the résumés that HR’s keyword filter already passed through.

Example

Retrieve 50 candidate chunks cheaply, then rerank to keep the 5 that truly fit before answering.

How it actually works

Fast retrieval favors recall — cast a wide net — but the top hits aren’t always the most relevant. A reranker (often a heavier cross-encoder that reads query and chunk together) scores each candidate directly, catching the good ones the first pass buried. More compute, better precision.

For product teams

A cheap accuracy boost for RAG: keep retrieval wide, then let reranking pick the winners.

For engineers

A cross-encoder scoring (query, passage) pairs to reorder first-stage retrieval by relevance.

Retrieval — Cleans up what retrieval hands over.
RAG — A quality lever for RAG.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome