Zero-Shot, One-Shot, and Few-Shot Prompting (for task examples)
These terms refer to how many examples you give an AI to guide its response for a specific task. 'Zero-shot' means you give no examples, just the instruction. 'One-shot' means you provide one example of the desired input-output format. 'Few-shot' means you provide a small set of input-output examples. For instance, if you want to summarize text, a zero-shot prompt might just say 'Summarize this:' A one-shot prompt would be 'Text: [original text] Summary: [example summary] Text: [new text to summarize] Summary:' This helps the AI understand the *format* and *style* you expect, even for tasks it hasn't been explicitly trained on in that specific way. The more examples (up to a point), the better the AI understands your specific task requirement.
It's like teaching someone a new game: zero-shot is just telling them the rules; one-shot is showing them one round played correctly; few-shot is playing a few rounds with them so they grasp the nuances.
By providing even a single example, you dramatically improve the AI's ability to match your desired output format and style, saving you significant editing time and reducing irrelevant responses.
Learn one new AI thing every day.
Daily Deck sends you seven plain-English cards like this every morning. Free.
Start free