# Reranking

> Reranking is a second retrieval stage in which a more accurate but slower model, typically a cross-encoder that reads the query and each candidate together, rescores the top results from a fast first-stage search and reorders them.

- Identifier: PTL-0129
- Category: Retrieval & Tool Use
- Canonical URL: https://protologue.com/t/reranking/
- Also known as: re-ranking, cross-encoder reranking, reranker
- Introduced: 2019

## Description

First-stage retrievers embed queries and documents separately to stay fast, which limits accuracy. A reranker sees both texts jointly, so it judges relevance more precisely, and applying it only to a few dozen candidates keeps cost manageable. Rerankers can also be LLMs prompted to judge or order passages.

## Related terms

- [Dense Retrieval](https://protologue.com/t/dense-retrieval/)
- [Hybrid Search](https://protologue.com/t/hybrid-search/)
- [ColBERT](https://protologue.com/t/colbert/)
- [Retrieval-Augmented Generation](https://protologue.com/t/retrieval-augmented-generation/)

## Sources

- Nogueira and Cho (2019). Passage Re-ranking with BERT. https://arxiv.org/abs/1901.04085

## Cite this entry

Protologue. (2026). Reranking. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0129). https://protologue.com/t/reranking/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)