Multi-Vector (Late Interaction) Embedding Models with Sentence Transformers

Automated excerpt

Scoring uses MaxSim: for each query token, take its highest similarity against any document token, then sum those maxima across the query. It isn't one-to-one either, since several query tokens routinely settle on the same document token. Visual document retrieval models are the exception.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.