What does “Cross-Encoder” mean?
A model that scores a query and document together in one pass for high-accuracy relevance judgments.
How Cross-Encoder Works
Cross-Encoder comparing Bi-encoder (Query and document embedded separately, Fast, precomputed) versus Cross-encoder (Query and document read together, Slower, more accurate).
Practical Example
A cross-encoder reads 'how do I reset my password' together with a help article and outputs a single relevance score. Because it sees both at once, it can pick up on subtle matches.
Why It Matters
Cross-encoders are a standard choice for reranking. They are too slow to run on an entire collection.
Common Questions About Cross-Encoder
Q: Why more accurate?
It can compare words in the query with words in the document directly.
Q: Why slower?
Each query-document pair needs its own pass, so nothing can be precomputed.
Q: Where is it used?
Reranking search results and judging relevance.