Giving a language model a handful of worked examples inside the prompt so it can follow the same pattern for a new input.
Few-shot prompting shows a model what good output looks like instead of only describing it. The prompt includes a small number of input and output pairs, often two to five, followed by the new input the model should handle. It sits between zero-shot prompting, which gives no examples, and fine-tuning, which retrains the model on many examples.
A small online store might classify customer emails this way. The prompt lists three sample emails labelled refund, delivery question, and product question, then adds a new email and asks for its label. The model picks up the format and categories from the examples. Few-shot prompting is quick and cheap to try, but the examples use up part of the context window, and poorly chosen examples can bias results.
Further training a pre-trained AI model on a smaller, specific dataset so it performs better on a particular task or style.
Prompt EngineeringThe practice of designing and refining instructions given to AI models to get more accurate, useful, and reliable outputs.
Zero-Shot LearningThe ability of an AI model to perform a task or recognize categories it was not explicitly trained on, without any examples.
Prompt patterns and editing techniques that keep your own voice in AI-assisted emails, posts and reports, plus a checklist of robotic tells to cut before you hit send.
What prompt engineering is, the 6 parts of a strong prompt and before-and-after examples for common business tasks you can copy and adapt.