Zero-shot prompting is the technique of instructing a language model to perform a task using only a description, without providing worked examples. It relies on knowledge the model acquired during pretraining.
Semantic Classification
Content
- Zero-shot prompting asks a model to carry out a task from instructions alone, such as classifying or summarising text, without including example input-output pairs in the prompt. It contrasts with few-shot prompting, which supplies several demonstrations to guide the model.
- The approach works because large language models generalise from broad pretraining, though clear task descriptions and output formats improve reliability. It is the simplest form of in-context learning and a baseline against which few-shot methods are compared.