Imagine you have asked a large language model to write a short story about a robot who discovers music. The model provides a first draft. Your task is to write a follow-up instruction that asks the model to improve its story, but without giving any specific directions on *how* to improve it.

Google

A simple, unguided instruction for refining a model's output can be a generic command like "Please refine it!". This type of prompt relies on the LLM's general ability to follow instructions without providing specific criteria for improvement.

Example of an Unguided Refinement Prompt

A user receives an initial response from a large language model and wants to ask for an improved version. Which of the following follow-up prompts is an example of an instruction that relies on the model's general capabilities without providing specific criteria for improvement?

Based on the provided scenario, evaluate the user's follow-up instruction. What is its primary limitation in achieving the desired refinement of the summary?

Learn Before

Related