Zero-Shot Learning

A machine-learning approach in which a model performs a task on classes or scenarios it has never seen explicit training examples for, relying entirely on its pre-trained knowledge and the structure of the prompt or input. In LLM-based accessibility testing, zero-shot prompting means asking the model to classify or describe an accessibility issue without providing labelled examples, which is useful when annotated accessibility datasets are scarce. Research suggests that well-engineered zero-shot prompts can match or outperform few-shot alternatives while using fewer tokens.