Which statement best describes zero-shot prompting in language models?

Which statement best describes zero-shot prompting in language models?

• A. Zero-shot uses no examples to guide the model.
• B. Zero-shot uses a single example to demonstrate the desired format.
• C. Zero-shot relies on chain-of-thought reasoning.
• D. Zero-shot requires fine-tuning on task data.

Answer: (A)

Zero-shot prompting tests the model's ability to perform a task using only the task description or instruction, with no demonstration examples given in the prompt. It relies on the model's pre-trained knowledge to generalize from that instruction alone. Providing no examples to guide the model is the defining feature, which is exactly what zero-shot means. Using a single example would be demonstrating the task format, which is not zero-shot—it’s a one-shot (or few-shot) setup. Chain-of-thought reasoning describes including a step-by-step line of thought, which can be used in various prompting styles and isn’t what defines zero-shot. Fine-tuning on task data changes the model's parameters through training; zero-shot inference uses the fixed model without task-specific training, so it isn’t about training either.

Zero-shot prompting tests the model's ability to perform a task using only the task description or instruction, with no demonstration examples given in the prompt. It relies on the model's pre-trained knowledge to generalize from that instruction alone. Providing no examples to guide the model is the defining feature, which is exactly what zero-shot means.

Using a single example would be demonstrating the task format, which is not zero-shot—it’s a one-shot (or few-shot) setup. Chain-of-thought reasoning describes including a step-by-step line of thought, which can be used in various prompting styles and isn’t what defines zero-shot. Fine-tuning on task data changes the model's parameters through training; zero-shot inference uses the fixed model without task-specific training, so it isn’t about training either.