Glossary
What is Cross-encoder reranking? Definition and meaning
A model reads the query and each candidate together and produces a more accurate relevance score for the shortlist.
Definition
Cross-encoder reranking: A model reads the query and each candidate together and produces a more accurate relevance score for the shortlist.
In short: Re-scoring top results with a joint query and document model.
When it is used
Raising precision of the final context passed to a language model.
Learn more
The full guide to Cross-encoder reranking explains how it works, when to use it and how it compares with the alternatives, with examples. The glossary lists every term on the site from A to Z.
Previous term: CrewAI · Next term: Cross-validation