Glossary

Logits — Glossary

Raw unnormalized scores a model outputs before converting to probabilities.

Reviewed:

Full definition

Decoding algorithms operate on logits with temperature and truncation. Logit biases can discourage or encourage specific tokens when APIs expose them.

Related encyclopedia topic

Tokens — Tokens are the discrete units language models read and generate—commonly subword pieces—used for context limits, billing, and model input encoding.

← Back to glossary · Learning hub