Skip to content

AI Glossary · Concepts

What is Zero-Shot Learning?

Zero-Shot Learning

Zero-shot learning applies a model to target categories without labeled training examples for those categories, using information such as descriptions or attributes. In language-model discussions, zero-shot prompting means giving a task instruction without worked examples in the prompt.

Also known as: ZSL

· Updated · Chain of Thought

  • 3 sources

A recognition model might learn from horse and tiger images while also learning associations with attributes such as stripes and a mane. A description of a zebra can then help it recognize a category with no labeled zebra examples in the training set. The description connects the unseen category to learned visual features.

An LLM asked “Label this review positive or negative” with no demonstrations is instead being used with zero-shot prompting. This describes the prompt, not proof that the model never encountered sentiment labels or similar reviews during training.

The distinction matters for evaluation. For unseen-class recognition, inspect training and pretraining overlap with test categories. For prompting, record which examples and instructions the model received. Neither setup guarantees reliable performance; weak descriptions or an unsuitable model can leave the task unresolved.

Google’s glossary also calls zero-shot prompting direct prompting (Google). Adding a few worked examples turns it into few-shot prompting; both depend on in-context learning, the model’s ability to follow a task described in its prompt.

Sources

Go deeper