Which prompt engineering technique provides input to a language model without specific training examples, expecting it to generate a response based on its preexisting knowledge?

Which prompt engineering technique provides input to a language model without specific training examples, expecting it to generate a response based on its preexisting knowledge?

• A. Zero-shot
• B. Few-shot
• C. Chain-of-thought (CoT)
• D. Contextual understanding

Answer: (A)

The key idea here is prompting the model to perform a task using only its existing knowledge, without any demonstrations or examples in the prompt. This approach relies on the model’s pretraining to generalize to the task from the description alone, which is exactly what zero-shot prompting does. If we were providing examples to guide the model, that would be a few-shot setup, which isn’t the case here. Chain-of-thought prompts focus on eliciting a step-by-step reasoning process, not on whether any training examples were given. Contextual understanding isn’t a standard prompt technique name for this scenario, so it doesn’t fit as well. So, the technique described is zero-shot prompting: you give the task as instructions, and the model answers using its preexisting knowledge without seeing any concrete examples.

The key idea here is prompting the model to perform a task using only its existing knowledge, without any demonstrations or examples in the prompt. This approach relies on the model’s pretraining to generalize to the task from the description alone, which is exactly what zero-shot prompting does.

If we were providing examples to guide the model, that would be a few-shot setup, which isn’t the case here. Chain-of-thought prompts focus on eliciting a step-by-step reasoning process, not on whether any training examples were given. Contextual understanding isn’t a standard prompt technique name for this scenario, so it doesn’t fit as well.

So, the technique described is zero-shot prompting: you give the task as instructions, and the model answers using its preexisting knowledge without seeing any concrete examples.