Home › Glossary › Presence penalty

Presence penalty

See it in the glossary A–Z

Setting between -2.0 and 2.0 applied at inference that penalises tokens which have already appeared at all. Raising it above zero encourages the model to move on to new topics.

Also called presence_penalty.

Read more: Microsoft Learn

In the Ultra Transcenders books

AI-901AI-103

Each book explains Presence penalty in context, with comparison tables and the common traps.

Terms in this definition

See Presence penalty in the full glossary