A prompt contains a task instruction and one sample of text, with no worked examples of the expected output. Which prompting pattern is this?
Show answer and explanation
Answer: B
Zero-shot means the model must rely on what it learned during pre-training, because the prompt supplies no demonstration of the desired input-to-output mapping. Modern instruction-tuned models handle many tasks this way, which makes zero-shot the cheapest and fastest starting point. If quality is uneven, adding a few examples turns the same request into a few-shot prompt without any retraining.
Why the other options are wrong
- A. Few-shot prompting requires one or more complete worked examples, and this prompt contains none.
- C. Fine-tuning updates model weights with a training dataset; nothing here changes the model itself.
- D. Chain-of-thought prompting asks for intermediate reasoning steps, which the prompt does not request.
Exam tip. Count the worked examples in the prompt: none is zero-shot, a handful is few-shot.