AI Glossary Term
What does “Few-Shot Learning” mean?
Definition
Learning or adapting behavior from only a small number of examples.
How Few-Shot Learning works
Example
You show a model two product reviews labeled with sentiment, then paste a third review with no label. The model infers the pattern and labels it, without any retraining.
Why it matters
Few-shot prompting lets people adapt a model quickly with a handful of examples rather than a large labeled dataset.
Common questions
How is few-shot different from fine-tuning?
Few-shot gives examples in the prompt at run time and changes no weights. Fine-tuning updates the model's weights.
How many examples count as few?
Usually between about two and a dozen.
Does the choice of examples matter?
Yes. Clear, varied, correctly labeled examples tend to help, and sloppy ones can mislead the model.