RailwailRailwail
Technique

Few-shot Prompting

Including a small number of solved examples in the prompt to teach the model a pattern.

Definition

Few-shot prompting embeds 1–32 (input, output) examples before the actual query. The model imitates the format and reasoning style. It usually beats zero-shot on niche tasks and avoids the cost and complexity of fine-tuning.

Common use cases

  • Classification
  • Information extraction
  • Format imitation

Related terms

Few-shot
Zero-shot Prompting
In-context Learning
Back to glossary
    Few-shot Prompting — AI Glossary | Railwail