Zero-Shot Learning
Zero-shot learning is the use of a model on a task described only by natural-language instructions, with no in-context examples of the input-output mapping. Performance relies entirely on knowledge and patterns acquired during pretraining. It is faster to set up than few-shot prompting and uses fewer tokens, but is more sensitive to instruction wording and task ambiguity.
Definition
Zero-shot learning is the use of a model on a task described only by natural-language instructions, with no in-context examples of the input-output mapping. Performance relies entirely on knowledge and patterns acquired during pretraining. It is faster to set up than few-shot prompting and uses fewer tokens, but is more sensitive to instruction wording and task ambiguity.
Examples
- Asking 'Classify this email as spam or not spam' with no prior samples.
- Requesting a French translation directly without any example pairs.
- Instructing 'Extract company names as a JSON array' on raw text.
Related terms
No credit card required
Frequently asked
- What is Zero-Shot Learning?
- Zero-shot learning is the use of a model on a task described only by natural-language instructions, with no in-context examples of the input-output mapping. Performance relies entirely on knowledge and patterns acquired during pretraining. It is faster to set up than few-shot prompting and uses fewer tokens, but is more sensitive to instruction wording and task ambiguity.
- How is Zero-Shot Learning used in practice?
- Zero-Shot Learning is commonly used in scenarios such as: Asking 'Classify this email as spam or not spam' with no prior samples.; Requesting a French translation directly without any example pairs.; Instructing 'Extract company names as a JSON array' on raw text..
- How does yno.ai support Zero-Shot Learning?
- yno.ai supports Zero-Shot Learning as part of its agentic AI platform — combining multiple frontier models with extensible tool use.
- What other concepts are related to Zero-Shot Learning?
- Zero-Shot Learning relates to Few-Shot Learning, Prompt Engineering, Inference, Large Language Model. Each links to its own glossary page below.