Quantifying and Mitigating Self-Preference Bias of LLM Judges

• URL: http://arxiv.org/abs/2604.22891v2
• Date: Tue, 28 Apr 2026 13:54:05 GMT
• Title: Quantifying and Mitigating Self-Preference Bias of LLM Judges
• Abstract summary: Self-Preference Bias (SPB) is a directional evaluative deviation in which LLMs systematically favor or disfavor their own generated outputs during evaluation.<n>Existing measurements rely on costly human annotations and conflate generative capability with evaluative stance, and thus are impractical for large-scale deployment in real-world systems.<n>We introduce a fully automated framework to quantifying and mitigating SPB, which constructs equal-quality pairs of responses with negligible quality differences.
• Score: 5.017055831476582
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: LLM-as-a-Judge has become a dominant approach in automated evaluation systems, playing critical roles in model alignment, leaderboard construction, quality control, and so on. However, the scalability and trustworthiness of this approach can be substantially distorted by Self-Preference Bias (SPB), which is a directional evaluative deviation in which LLMs systematically favor or disfavor their own generated outputs during evaluation. Existing measurements rely on costly human annotations and conflate generative capability with evaluative stance, and thus are impractical for large-scale deployment in real-world systems. To address this issue, we introduce a fully automated framework to quantifying and mitigating SPB, which constructs equal-quality pairs of responses with negligible quality differences, enabling statistical disentanglement of discriminability from bias propensity without human gold standards. Empirical analysis across 20 mainstream LLMs reveals that advanced capabilities are often uncorrelated, or even negatively correlated, with low SPB. To mitigate this bias, we propose a structured multi-dimensional evaluation strategy grounded in cognitive load decomposition, which reduces SPB by 31.5\% on average.

Related papers

• (Towards) Scalable Reliable Automated Evaluation with Large Language Models [0.0]
  This work introduces a novel evaluation framework designed to approximate expert-level assessments of Large Language Models (LLMs) outputs.<n>An Elo rating system is used to generate stable and interpretable rankings.<n>Preliminary results show that automatically derived rankings correlate well with expert judgments.
  arXiv  Detail & Related papers  (2026-07-30T14:31:07Z)
• Quantifying and Auditing LLM Evaluation via Positive--Unlabeled Learning [4.114698130306098]
  Large Language Models (LLMs) are increasingly used as judges for scalable evaluation.<n>LLMs exhibit systematic biases that are decoupled from semantic quality.<n>Human supervision is costly and typically selective, yielding reliable positive judgments but leaving most outputs unlabelled and potentially mixed in quality.
  arXiv  Detail & Related papers  (2026-06-17T13:26:04Z)
• Automated Self-Testing as a Quality Gate: Evidence-Driven Release Management for LLM Applications [51.56484100374058]
  We present an automated self-testing framework that introduces quality gates with evidence-based release decisions.<n>We evaluate the framework through a longitudinal case study of an internally deployed multi-agent conversational AI system.
  arXiv  Detail & Related papers  (2026-03-13T20:44:15Z)
• Robust LLM Performance Certification via Constrained Maximum Likelihood Estimation [14.465019747508604]
  We propose a new, practical, and efficient approach to LLM failure rate estimation based on constrained maximum-likelihood estimation (MLE)<n>We validate our approach through a comprehensive empirical study, benchmarking it against state-of-the-art baselines like Prediction-Powered Inference (PPI)
  arXiv  Detail & Related papers  (2026-03-11T21:48:40Z)
• CDRRM: Contrast-Driven Rubric Generation for Reliable and Interpretable Reward Modeling [61.75914342638658]
  CDRRM (Contrast-Driven Reward Model) is a framework for high-quality rubric generation and guided preference judgment.<n>This work offers a scalable, interpretable, and data-efficient path for reward modeling.
  arXiv  Detail & Related papers  (2026-03-09T07:15:23Z)
• Mitigating Reward Hacking in RLHF via Bayesian Non-negative Reward Modeling [49.41422138354821]
  We propose a principled reward modeling framework that integrates non-negative factor analysis into the Bradley-Terry preference model.<n>BNRM represents rewards through a sparse, non-negative latent factor generative process.<n>We show that BNRM substantially mitigates reward over-optimization, improves robustness under distribution shifts, and yields more interpretable reward decompositions than strong baselines.
  arXiv  Detail & Related papers  (2026-02-11T08:14:11Z)
• BiasScope: Towards Automated Detection of Bias in LLM-as-a-Judge Evaluation [31.32711926164484]
  BiasScope is a framework for automatically and at scale discovering potential biases during model evaluation.<n>JudgeBench-Pro is a more challenging benchmark for evaluating the robustness of LLM-as-a-judge.
  arXiv  Detail & Related papers  (2026-02-10T03:51:03Z)
• Doubly-Robust LLM-as-a-Judge: Externally Valid Estimation with Imperfect Personas [31.16720541398267]
  We propose a doubly-robust estimation framework designed to address evaluation sampling bias.<n>Key to our approach is the use of "persona" ratings produced by prompting an evaluator to behave as a human rater.<n>We show that our approach yields valid system quality estimates when either (i) a model trained to predict human ratings using persona ratings and source data observed under sampling bias, or (ii) a reweighting model that corrects for sampling bias is of sufficient quality.
  arXiv  Detail & Related papers  (2025-09-26T21:42:51Z)
• 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)
• T2I-Eval-R1: Reinforcement Learning-Driven Reasoning for Interpretable Text-to-Image Evaluation [60.620408007636016]
  We propose T2I-Eval-R1, a novel reinforcement learning framework that trains open-source MLLMs using only coarse-grained quality scores.<n>Our approach integrates Group Relative Policy Optimization into the instruction-tuning process, enabling models to generate both scalar scores and interpretable reasoning chains.
  arXiv  Detail & Related papers  (2025-05-23T13:44:59Z)
• Beyond Single-Point Judgment: Distribution Alignment for LLM-as-a-Judge [24.862965044243168]
  Previous methods rely on single-point evaluations, overlooking the inherent diversity and uncertainty in human evaluations.<n>We propose a novel training framework that explicitly aligns the LLM-generated judgment distribution with empirical human distributions.<n>Our framework significantly outperforms existing closed-source LLMs and conventional single-point alignment methods.
  arXiv  Detail & Related papers  (2025-05-18T08:33:09Z)
• Pairwise or Pointwise? Evaluating Feedback Protocols for Bias in LLM-Based Evaluation [57.380464382910375]
  We show that the choice of feedback protocol for evaluation can significantly affect evaluation reliability and induce systematic biases.<n>We find that generator models can flip preferences by embedding distractor features.<n>We offer recommendations for choosing feedback protocols based on dataset characteristics and evaluation objectives.
  arXiv  Detail & Related papers  (2025-04-20T19:05:59Z)
• Identifying and Mitigating Social Bias Knowledge in Language Models [52.52955281662332]
  We propose a novel debiasing approach, Fairness Stamp (FAST), which enables fine-grained calibration of individual social biases.<n>FAST surpasses state-of-the-art baselines with superior debiasing performance.<n>This highlights the potential of fine-grained debiasing strategies to achieve fairness in large language models.
  arXiv  Detail & Related papers  (2024-08-07T17:14:58Z)

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

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.