Based on the scenario below, formulate a single, specific instruction (a prompt) that would be more effective at fixing the output than the user's original attempt. Then, briefly explain why your proposed instruction is better.

Google

To guide an LLM's refinement process toward a specific issue, a prompt can provide targeted feedback. For instance, the instruction "Please correct all the grammatical errors in the translation" directs the model to focus its improvements on grammatical accuracy.

Example of a Guided Refinement Prompt for Grammatical Correction

An AI model produced the following sentence: 'The undertaking of the project was commenced with significant alacrity by the team members, who demonstrated a high level of collaborative synergy.' While the sentence is grammatically correct, the goal is to make it sound more natural and less formal. Which of the following instructions would be the most effective and targeted way to guide the model toward the desired revision?

Improving an LLM's Translation Output

An AI assistant produced the following text: 'Me go to park. Dog run fast. We has fun.' Write a single, concise instruction for the AI that specifically targets the correction of the errors in this text while preserving its original simple meaning.

Learn Before

Related