Google

An example of a follow-up prompt used to generate a final conclusion based on pre-analyzed document segments is: 'Your task is to determine whether a text discusses the risks of AI. This text has been divided into segments, and you have obtained the relevancy of each segment to the topic of AI risks. Based on this, please provide your final result.' This type of prompt directs the model to synthesize findings from segment relevancy scores rather than analyzing the original text from scratch.

Example Prompt for Synthesizing Segment Analysis

A system is designed to determine if a lengthy research paper discusses 'ethical considerations'. The system first divides the paper into paragraphs and assigns a numerical 'relevance score' to each paragraph. Which of the following follow-up prompts is correctly structured to synthesize these paragraph-level scores into a single, final conclusion for the entire paper?

A developer is building a system to determine if a long technical manual contains information about 'safety protocols'. The system first breaks the manual into 100 segments and uses a separate process to generate a 'relevancy score' for each segment. To get the final answer, the developer uses the following follow-up prompt:

'I have provided the relevancy scores for 100 segments of a technical manual regarding 'safety protocols'. Now, please re-read the original manual and, based on your reading, provide a final yes/no answer as to whether it discusses safety protocols.'

Analyze the developer's follow-up prompt. What is the primary design flaw in this prompt, considering the goal is to efficiently synthesize a conclusion from the pre-analyzed segments?

Learn Before

Related