Cross-encoder
Cross-encoder: A model that jointly processes query and document to estimate their relevance accurately. Cross-attention captures detailed relations but prevents independent document precomputation. It is therefore mainly used on candidates already retrieved. Examples include Query-document reranking, Evaluating text pairs, Second-stage retrieval.
In italiano: Cross-EncoderLegge I · Embedding, feature e spazi vettorialiCross-attention captures detailed relations but prevents independent document precomputation. It is therefore mainly used on candidates already retrieved.
Examples
- Query-document reranking
- Evaluating text pairs
- Second-stage retrieval