Few-shot prompting
Putting a handful of worked examples in the prompt so the model copies the pattern.
Also: few-shot learning, one-shot
In few-shot prompting, the prompt shows several examples of the task, each input with its correct output, followed by a new input for the model to complete. Benchmarks often report results as five-shot or zero-shot to show how many examples were given. Few-shot prompting is especially useful for unusual formats or labels and for smaller models. With instruction-tuned models, a clear description of the task often works nearly as well as examples.