Few-shot

Overview

Few-shot prompting includes a handful of (input, output) pairs in the prompt, letting the LLM infer the task's format, style, and constraints from context. It is one of the simplest ways to adapt a model to a specific task without fine-tuning.

Practical tips

Two to eight examples is a practical range — more examples consume context window. Diversity and representativeness matter: a set of examples that all share the same bias will skew outputs. Use zero-shot first; add few-shot examples only when zero-shot quality is insufficient.