A Better LLM Evaluator for Text Generation: The Impact of Prompt Output Sequencing and Optimization

TL;DR

This paper explores how prompt design, including output sequencing and optimization, affects the accuracy and consistency of large language models in evaluating generated texts, offering insights to improve LLM-based assessment methods.

Contribution

It systematically analyzes the impact of prompt output sequencing and optimization on LLM evaluation performance, providing guidelines for better prompt design.

Findings

Order of reasons and scores affects LLM scoring accuracy

Prompt optimization can improve scoring alignment with human judgments

Different prompt structures influence model understanding of evaluation rules

Abstract

This research investigates prompt designs of evaluating generated texts using large language models (LLMs). While LLMs are increasingly used for scoring various inputs, creating effective prompts for open-ended text evaluation remains challenging due to model sensitivity and subjectivity in evaluation of text generation. Our study experimented with different prompt structures, altering the sequence of output instructions and including explanatory reasons. We found that the order of presenting reasons and scores significantly influences LLMs' scoring, with a different level of rule understanding in the prompt. An additional optimization may enhance scoring alignment if sufficient data is available. This insight is crucial for improving the accuracy and consistency of LLM-based evaluations.