Google

A prompt designed to instruct a Large Language Model to refine a previous answer typically provides the model with three key pieces of information: the original question, the model's initial response, and the feedback generated for that response. For instance, the instructional part of the prompt might state: 'You will be provided with a question, its corresponding response, and some feedback on this response.' This structure guides the model to produce an improved version.

Example of a Refinement Prompt Structure

A user wants to instruct a language model to improve its previous answer based on a critique. Which of the following prompt structures is most complete and effective for this task, as it provides the model with all the necessary context to understand and implement the required changes?

A prompt engineer provides a language model with the following instruction: 'Your last answer was not detailed enough. Please provide a better one.' Critically analyze this instruction. Identify the two essential pieces of contextual information that are missing from this prompt for the model to effectively refine its previous response, and explain why their absence is problematic.

Critique of a Refinement Prompt

A user is trying to improve a language model's response. They have the following three pieces of information:

1.  **Original Question:** 'Explain the water cycle.'
2.  **Model's Initial Response:** 'The water cycle is how water moves around the Earth. It includes evaporation, condensation, and precipitation.'
3.  **Feedback on Response:** 'This is correct but very basic. Please add the stage of collection and briefly describe what happens during each of the four stages.'

Based on this scenario, write the complete instructional text for a new prompt that will guide the language model to improve its initial response. Your prompt should clearly state what information the model will receive and what it is expected to do with it.

Learn Before

Related