Zero-shot prompting
Asking a model to do a task with no worked examples in the prompt — just the instruction.
Zero-shot prompting gives the model an instruction and no examples, relying on what it learned in training to handle the task. For clear, common tasks it often works fine and keeps the prompt short.
It struggles when the output format is specific or the task has ambiguous edge cases, because the model has nothing to copy and must guess your intent. When zero-shot output drifts, adding a single example — moving to few-shot — usually fixes it faster than rewording the instruction.