What is Few-shot prompting?
Also called: few-shot learning · in-context examples
Few-shot prompting shows the model two to five examples of the input → output you want, so it copies the pattern instead of guessing it from a description.
Describing a format is weaker than demonstrating it. A few-shot prompt includes examples — "Input: … Output: …" — before the real task, and the model matches their style, length, tone and structure with remarkable fidelity. It's the most reliable way to get consistent output across many runs.
Examples teach everything, including your mistakes. Keep them consistent, representative (simple → hard → edge case), and few: three good examples beat ten sloppy ones.
Rewrite support replies in our voice. Input: "Your order is late." → Output: "You're right, and I'm sorry — your parcel left our warehouse Tuesday; here's the tracking link, and I've added a 10% code for the wait." Input: "The app crashed." → Output: …
- Use identical formatting for every example.
- Include one example of what to do in a tricky case.
- Only show positive examples — "don't do this" examples get copied too.
Use it right now
Ask our brain anything on the homepage — it remembers the whole conversation — or write a brief in the Studio and see the prompt it compiles to.