Zero-Shot Learning

Zero-shot learning asks a model to perform a task from an instruction alone, with no worked examples provided. It relies entirely on knowledge acquired during pre-training and instruction tuning, making it the simplest and cheapest baseline before adding examples or heavier prompting scaffolds.

Definition

Zero-Shot Learning

In a zero-shot setting the prompt describes the task in plain language and the model responds directly, without demonstrations. Instruction-tuned models are trained to follow such directives, so for common, well-understood tasks a clear instruction is often sufficient.

Because nothing anchors the expected output format, zero-shot quality depends heavily on how precisely the instruction is phrased and how familiar the task is. It is the natural first thing to try; if results are inconsistent, adding a few examples (few-shot) or a reasoning directive typically helps.

Frequently asked questions

What is Zero-Shot Learning?

Zero-shot learning asks a model to perform a task from an instruction alone, with no worked examples provided. It relies entirely on knowledge acquired during pre-training and instruction tuning, making it the simplest and cheapest baseline before adding examples or heavier prompting scaffolds.

Sources

Related terms

← Full AI & prompt engineering glossary

Put Zero-Shot Learning to work with Prompeteer, the Agentic Contextual AI Platform →