Related papers: Play Favorites: A Statistical Method to Measure Self-Bias in LLM-as-a-Judge

Play Favorites: A Statistical Method to Measure Self-Bias in LLM-as-a-Judge

URL: http://arxiv.org/abs/2508.06709v1
Date: Fri, 08 Aug 2025 21:22:12 GMT
Title: Play Favorites: A Statistical Method to Measure Self-Bias in LLM-as-a-Judge
Authors: Evangelia Spiliopoulou, Riccardo Fogliato, Hanna Burnsky, Tamer Soliman, Jie Ma, Graham Horwood, Miguel Ballesteros,
Abstract summary: Large language models (LLMs) can serve as judges that offer rapid and reliable assessments of other outputs.<n>LLMs may systematically assign overly favorable ratings to their own outputs, a phenomenon known as self-bias.<n>We present a statistical framework that explicitly formalizes assumptions under which self-bias can be identified and estimated.
Score: 17.40713507922006
License: http://creativecommons.org/licenses/by/4.0/
Abstract: Large language models (LLMs) can serve as judges that offer rapid and reliable assessments of other LLM outputs. However, models may systematically assign overly favorable ratings to their own outputs, a phenomenon known as self-bias, which can distort evaluations of true model performance. Previous studies often conflate genuine differences in model quality with bias or incorrectly assume that evaluations from LLMs and humans follow the same rating distributions. In this work, we present a statistical framework that explicitly formalizes assumptions under which self-bias can be identified and estimated. Our method models the difference in the scoring distribution that LLM-as-a-judge assigns to its own completions compared to other models, while accounting for the underlying quality of the completions provided by an independent, third-party judge (e.g., humans). Our method reliably isolates and quantifies self-bias, even when models vary in ability, ensuring that genuine performance differences are not mistaken for self-bias. We conduct an empirical analysis of self-bias on a large dataset (>5000 prompt-completion pairs) consisting of expert human annotations and judgments from nine different LLM judges. We find that some models, such as GPT-4o and Claude 3.5 Sonnet, systematically assign higher scores to their own outputs. These models also display family-bias; systematically assigning higher ratings to outputs produced by other models of the same family. Our findings highlight potential pitfalls of using LLM judges and offer practical guidance to mitigate biases when interpreting automated evaluations.

Related papers

Who can we trust? LLM-as-a-jury for Comparative Assessment [42.32900791516691]
Large language models (LLMs) are increasingly applied as automatic evaluators for natural language generation assessment.<n>LLMs judges vary substantially in performance across tasks and aspects, and their judgment probabilities may be biased and inconsistent.<n>We propose BT-sigma, a judge-aware extension of the Bradley-Terry model that introduces a discriminator parameter for each judge to jointly infer item rankings and judge reliability from pairwise comparisons alone.
arXiv Detail & Related papers (2026-02-18T17:04:02Z)
Reference-Specific Unlearning Metrics Can Hide the Truth: A Reality Check [60.77691669644931]
We propose Functional Alignment for Distributional Equivalence (FADE), a novel metric that measures distributional similarity between unlearned and reference models.<n>We show that FADE captures functional alignment across the entire output distribution, providing a principled assessment of genuine unlearning.<n>These findings expose fundamental gaps in current evaluation practices and demonstrate that FADE provides a more robust foundation for developing and assessing truly effective unlearning methods.
arXiv Detail & Related papers (2025-10-14T20:50:30Z)
Skewed Score: A statistical framework to assess autograders [2.9645858732618238]
"LLM-as-a-judge", or autograders, offer a scalable alternative to human evaluation.<n>They have shown mixed reliability and may exhibit systematic biases.<n>We propose a statistical framework that enables researchers to simultaneously assess their autograders.
arXiv Detail & Related papers (2025-07-04T18:45:10Z)
Quantitative LLM Judges [60.773734899532336]
We propose quantitative LLM judges, which align evaluation scores of existing LLM judges to humans in a given domain.<n>The models are trained to improve the score of the original judge using its rationale and score.<n>Our experiments show that quantitative judges can improve the predictive power of existing judges through post-hoc modeling.
arXiv Detail & Related papers (2025-06-03T14:44:23Z)
Self-Preference Bias in LLM-as-a-Judge [13.880151307013321]
We introduce a novel metric to measure the self-preference bias in large language models (LLMs)<n>Our results show GPT-4 exhibits a significant degree of self-preference bias.<n>This suggests that the essence of the bias lies in perplexity and that the self-preference bias exists because LLMs prefer texts more familiar to them.
arXiv Detail & Related papers (2024-10-29T07:42:18Z)
Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge [84.34545223897578]
Despite their excellence in many domains, potential issues are under-explored, undermining their reliability and the scope of their utility. We identify 12 key potential biases and propose a new automated bias quantification framework-CALM- which quantifies and analyzes each type of bias in LLM-as-a-Judge. Our work highlights the need for stakeholders to address these issues and remind users to exercise caution in LLM-as-a-Judge applications.
arXiv Detail & Related papers (2024-10-03T17:53:30Z)
Cycles of Thought: Measuring LLM Confidence through Stable Explanations [53.15438489398938]
Large language models (LLMs) can reach and even surpass human-level accuracy on a variety of benchmarks, but their overconfidence in incorrect responses is still a well-documented failure mode. We propose a framework for measuring an LLM's uncertainty with respect to the distribution of generated explanations for an answer.
arXiv Detail & Related papers (2024-06-05T16:35:30Z)
Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models [56.02275285521847]
We propose to evaluate models using a Panel of LLm evaluators (PoLL) We find that using a PoLL composed of a larger number of smaller models outperforms a single large judge, exhibits less intra-model bias due to its composition of disjoint model families, and does so while being over seven times less expensive.
arXiv Detail & Related papers (2024-04-29T15:33:23Z)
LLM Evaluators Recognize and Favor Their Own Generations [33.672365386365236]
We investigate if self-recognition capability contributes to self-preference. We find a linear correlation between self-recognition capability and the strength of self-preference bias. We discuss how self-recognition can interfere with unbiased evaluations and AI safety more generally.
arXiv Detail & Related papers (2024-04-15T16:49:59Z)
Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement [75.7148545929689]
Large language models (LLMs) improve their performance through self-feedback on certain tasks while degrade on others. We formally define LLM's self-bias - the tendency to favor its own generation. We analyze six LLMs on translation, constrained text generation, and mathematical reasoning tasks.
arXiv Detail & Related papers (2024-02-18T03:10:39Z)
Benchmarking Cognitive Biases in Large Language Models as Evaluators [16.845939677403287]
Large Language Models (LLMs) have been shown to be effective as automatic evaluators with simple prompting and in-context learning. We evaluate the quality of ranking outputs introducing the Cognitive Bias Benchmark for LLMs as Evaluators. We find that LLMs are biased text quality evaluators, exhibiting strong indications on our bias benchmark.
arXiv Detail & Related papers (2023-09-29T06:53:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.