Few-shot

Also: Few-Shot Prompting · In-Context Learning

Giving the model a few examples in the prompt so that it adopts the pattern and format.

Few-shot prompting uses language models’ ability to learn from examples in context (“in-context learning”) – without any training. Two to five well-chosen examples are often enough to define style, structure or evaluation criteria.

The examples should be varied and typical; otherwise the model also picks up unintended quirks.

General information, not legal or professional advice.