CourionAI
EN
Newsletter
← Glossary Term

Zero-shot

When an AI model handles a task it was never specifically trained or shown examples for.

Zero-shot means asking a model to do something straight away, with no extra training and no worked examples. If you ask a chatbot to translate a sentence or summarise an email without showing it a single sample first, you are using it zero-shot.

The opposite is giving a few examples in your prompt (few-shot) or retraining the model on your own data (fine-tuning). Zero-shot performance is a useful test of how general a model really is, because it shows how well it copes with something new.