Explaining Length Bias in LLM-Based Preference Evaluations

Published: 06 Mar 2025, Last Modified: 07 Mar 2025ICLR 2025 Workshop Data Problems PosterEveryoneRevisionsBibTeXCC BY 4.0
Keywords: length bias, llm evaluation
TL;DR: This study investigates the biases in large language model-based preference evaluation, particularly its preference for longer responses.
Abstract: The use of large language models (LLMs) as judges, particularly in preference comparisons, has become widespread, but this reveals a notable bias towards longer responses, undermining the reliability of such evaluations. To better understand such bias, we propose to decompose the preference evaluation metric, specifically the *win rate*, into two key components: *desirability* and *information mass*, where the former is length-independent and related to trustworthiness such as correctness, toxicity, and consistency, and the latter is length-dependent and represents the amount of information in the response. We empirically demonstrated the decomposition through controlled experiments and found that response length impacts evaluations by influencing information mass. To derive a reliable evaluation metric that assesses content quality without being confounded by response length, we propose **AdapAlpaca**, a simple yet effective adjustment to win rate measurement. Specifically, **AdapAlpaca** ensures a fair comparison of response quality by aligning the lengths of reference and test model responses under equivalent length intervals.
Submission Number: 5
Loading

OpenReview is a long-term project to advance science through improved peer review with legal nonprofit status. We gratefully acknowledge the support of the OpenReview Sponsors. © 2025 OpenReview