Reranker
Also called reranking model, reranking
A model that scores each passage a search returned against the question and keeps only the best few for the AI model to read.
The first search in a feature that answers from documents has to run fast across thousands of passages, so it's rough. A reranker then reads each candidate together with the question and gives it a relevance score, which is slower but sharper, and passes only the top few to the model.
Fewer, better passages mean better answers and less for the model to read. The catch is the extra step.
For example
Search brings back 150 passages about returns; a reranker scores each against "Can I return opened headphones?" and passes the top 20 to the model.
Put it to use
Where you'll meet reranker in ChatGPT, Gemini and Claude, how to use it there, and the mix-up most people make.
Daily Bytes Plus is in the Android app.