A team designing a policy assistant wants the model to follow a specific questions–answer pattern. They provide two complete example Q&A pairs in the…

This technique is known as few-shot prompting. In large language models (LLMs), "shot" refers to an example provided within the prompt to guide the model's response. Zero-shot prompting provides no examples, while few-shot prompting provides a small number (typically 2-5) of complete examples. By showing the model two Q&A pairs, the team is conditioning it to follow a specific format, tone, and structure for its subsequent answers, which is the exact purpose of few-shot prompting.

B. Zero-shot prompting involves asking the model to perform a task without providing any examples in the prompt, which is contrary to the scenario described.

C. Chain-of-thought prompting guides the model to break down its reasoning process step-by-step, which is not the primary goal here; the focus is on output format.

D. Temperature is a model parameter that controls the randomness or creativity of the output, not a method of structuring the prompt with examples.

1. Microsoft Azure Documentation. (2024). Introduction to prompt engineering. "Prompting techniques" section. It defines few-shot prompting as providing "a few examples to the model to help it understand the context and pattern."

2. Brown, T. B., et al. (2020). Language Models are Few-Shot Learners. arXiv:2005.14165. Section 2.2 describes few-shot learning as providing a few demonstrations of the task at inference time as part of the prompt. https://doi.org/10.48550/arXiv.2005.14165

3. Stanford University. (CS224N). Lecture 12: Natural Language Generation. The courseware distinguishes between zero-shot, one-shot, and few-shot prompting based on the number of examples provided in the context.