Skip to content
BENIBIX Learn

Reranking

Also known as Re-ranking, Reranker, Cross-Encoder Reranking

A second, slow look at the top hits. It puts them in a better order.

Think of it like

Think of it like a coach who picks 50 kids fast. Then the coach watches each one to choose the best 5.

Example

Jack's search pulls 50 help articles in 20 milliseconds. A reranker reads each with the question and moves the one true answer from 14th place to 1st.

Why it matters

It often gives the biggest jump in search quality for the least work.

Definition

Reranking is a second pass that puts search results in a better order. A fast search first pulls, say, the top 50 candidates. Then a reranker model reads the question together with each candidate and gives it a new score. That is slower per text but much more accurate, so it is only run on the short list. The best few after reranking are the ones passed on to the user or the model.
Fifty quick results go in. The reranker reads each with the question, and the best five come out on top.

Common mistake

Reranking too few candidates. If the right text is not in the first pull, no reranker can bring it back.

Key takeaways

  1. 1
    Two passes. Fast, then careful.
  2. 2
    Reads pairs. Question with each result.
  3. 3
    Short list only. Too slow for all.

Compare