Few-Shot Learning

Few-shot learning is one of the most effective prompt engineering techniques. Instead of describing what you want in abstract terms, you show the model 2-5 concrete examples of the desired input-output pattern. The model generalizes from these examples and applies the pattern to new inputs. For example, to classify customer feedback sentiment, you include: "Great product, love it! -> positive", "Terrible experience, want a refund -> negative", "It works fine, nothing special -> neutral" in your prompt. The model then correctly classifies new feedback without any fine-tuning. Few-shot is the sweet spot between zero-shot (no examples, less reliable) and fine-tuning (expensive, requires training data). Use it when: (1) the task has a specific output format, (2) the model struggles with zero-shot accuracy, (3) you need consistent formatting across responses. Most production AI features use few-shot prompting for data extraction, classification, and structured generation tasks.