Pattern
ColBERT and late-interaction models
Provide a middle ground between bi-encoders (single-vector embeddings, fast but limited representation) and cross-encoders (full cross-attention, expressive but slow): per-token embeddings with MaxSim aggregation, document representations precomputable and indexable, query-time computation lighter than cross-encoders.
Classification — Late-interaction retrieval and reranking with per-token embeddings.
Full entry
The complete treatment — problem, how it works, when to use it, sources, and examples — lives in the volume below.
Browse the full pattern index or the catalog.