Zero-Shot / Few-Shot Learning in plain English.

Zero-shot and few-shot learning describe how many examples you give an AI model before asking it to perform a task. Zero-shot means you just ask directly — "classify this email as spam or not spam" — with no examples at all, relying entirely on what the model learned during training. Few-shot means you include a small number of examples in the prompt itself ("here are three emails and their correct labels; now classify this fourth one"), which reliably improves accuracy and consistency of format without any retraining. Modern frontier models are strong enough at zero-shot performance on many everyday tasks that few-shot examples are often unnecessary, but few-shot prompting remains one of the cheapest, fastest ways to improve accuracy on a specific, narrow task — no fine-tuning, no extra training cost, just better examples in the prompt.