// ⌕ RAG & Vector Retrieval
Cross-Encoder Reranking
Feed query and document together into BERT to score relevance and re-rank candidates.
交叉编码器重排（BERT Reranker） aka Reranker / monoBERT / 重排序 2019 Foundational core
IN PLAIN WORDS
Vector search encodes query and document separately – fast but coarse. A cross-encoder concatenates them so every attention layer sees both, judging relevance far more accurately; it is too slow for the whole corpus, so it re-ranks the top tens to hundreds. Retrieve with vectors/BM25, then re-rank: the standard two-stage RAG setup.
Used in: The re-ranking stage of RAG and search (e.g. bge-reranker and vendor Rerank APIs)
01Original paper
02Where it sits
Learn the left column first; the right column builds on this one.
Related:ColBERT (Late Interaction)Dense Passage Retrieval (DPR)Retrieval-Augmented Generation (RAG)
Left: prerequisites · right: builds on this · dashed: related. Full map →
03Learning resources
Official docs, explainers, courses and reference code. All links checked.
Official docs
04Where the course teaches it
For AI / data
- Verification: verified, checked 2026-10-22.
- JSON: /api/learn/algorithms.json (id:
cross-encoder-rerank)