← All models · Multilingual MiniLM
Basics · feature-extraction · WASM
The fundamental capability, felt directly: the model turns each sentence into a 384-d unit vector, and cosine similarity — the dot product of those vectors — is one number from −1 to 1 that says how close in meaning they are, even across languages. A sentence and its translation should score high; an unrelated sentence in the same language should score low.
cosine = –