Concept重排序模型 / Reranker
Rerank Model
A model used in RAG systems to perform fine-grained ranking of initial recall results. It computes the relevance between query and document through more precise semantic matching, improving final retrieval quality.
Timeline (last 90 days)
Oct 28
Rerank models are usually cross-encoders, with higher accuracy than bi-encoders but high computational cost, suitable for finely ranking a small number of candidates such as the top-50.
Unverified50%