N-shot learning

N-shot learning is a machine learning paradigm where models learn to perform new tasks using only n examples per class, where n represents a small, finite number. This approach encompasses zero-shot learning (no examples), one-shot learning (a single example), and few-shot learning (a handful of examples per class, with no fixed upper bound agreed across the literature). Many-shot in-context learning, which places hundreds or thousands of examples in a model's context window, is a related but separate idea: it does not rely on a small n. N-shot learning leverages meta-learning techniques, where models learn how to learn efficiently from limited data by training on diverse task distributions. Core methods include model-agnostic meta-learning (MAML), prototypical networks, and in-context learning with large language models. For AI agents, n-shot learning enables rapid adaptation to new domains, personalization without extensive retraining, and deployment in data-scarce environments. This capability is crucial for autonomous systems that must quickly acquire new skills, handle novel scenarios, and operate effectively when collecting large training datasets is impractical or expensive.

Related terms

• 0 shot (zero-shot)
• 0 shot learning
• Few-Shot Learning
• Few-Shot Prompting
• How does Zero shot learning work

Related services: Agentic AI consulting.

← Back to the AI Glossary