The Comparative Trap: Pairwise Comparisons Amplifies Biased Preferences of LLM Evaluators

TL;DR

This paper investigates biases in LLM evaluators, showing pairwise comparisons amplify superficial preferences, and proposes PRePair, a new method combining pointwise reasoning to reduce bias and improve evaluation robustness.

Contribution

The paper identifies bias amplification in pairwise LLM evaluations and introduces PRePair, a novel method that mitigates this bias by integrating pointwise reasoning.

Findings

Pairwise evaluation amplifies superficial biases.

PRePair reduces biased preferences effectively.

PRePair outperforms traditional methods on benchmarks.

Abstract

As large language models (LLMs) are increasingly used as evaluators for natural language generation tasks, ensuring unbiased assessments is essential. However, LLM evaluators often display biased preferences, such as favoring verbosity and authoritative tones. Our empirical analysis reveals that these biases are exacerbated in pairwise evaluation, where LLMs directly compare two outputs and easily prioritize superficial attributes. In contrast, pointwise evaluation, which assesses outputs independently, is less susceptible to such bias because each output is judged in isolation. To address the limitations of the pairwise evaluation, we introduce a novel evaluation method, PRePair, which integrates pointwise reasoning within a pairwise framework. PRePair effectively alleviates biased preference, improving performance on the adversarial benchmark (LLMBar) while outperforming pointwise evaluation on the standard benchmark (MT-Bench).