Skip to content

What is Zero-Shot Learning?

Also known as: Zero-Shot

Zero-shot learning (or zero-shot prompting) is when an AI model completes a task it has never seen examples of during training, relying only on a natural language description of what to do.

The model draws on knowledge acquired during pre-training to interpret the task instructions and generate an appropriate response without any task-specific examples or fine-tuning.

It works by framing the request as a clear prompt that describes the desired input-output behavior, allowing the model to generalize from its broad understanding of language and concepts.

This approach contrasts with supervised learning, which requires labeled data, and few-shot learning, which supplies a handful of examples in the prompt.

Example

A user asks a language model to classify movie reviews as positive or negative by simply writing 'Classify the sentiment of this review as positive or negative' followed by the review text, without showing any labeled examples.

Why it matters

Zero-shot capabilities let users apply powerful pretrained models to new tasks instantly, reducing the need for data collection and retraining and making advanced AI more accessible and flexible.

Frequently asked questions

Zero-shot uses only instructions with no examples, while few-shot includes a small number of labeled examples in the prompt to guide the model.