Saved in:
Bibliographic Details
Main Authors: Xu, Mingyuan, Tan, Xinzi, Wu, Jiawei, Zhou, Doudou
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.21817
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914291296239616
author Xu, Mingyuan
Tan, Xinzi
Wu, Jiawei
Zhou, Doudou
author_facet Xu, Mingyuan
Tan, Xinzi
Wu, Jiawei
Zhou, Doudou
contents Evaluating large language models (LLMs) on open-ended tasks without ground-truth labels is increasingly done via the LLM-as-a-judge paradigm. A critical but under-modeled issue is that judge LLMs differ substantially in reliability; treating all judges equally can yield biased leaderboards and misleading uncertainty estimates. More data can make evaluation more confidently wrong under misspecified aggregation. We propose a judge-aware ranking framework that extends the Bradley-Terry-Luce model by introducing judge-specific discrimination parameters, jointly estimating latent model quality and judge reliability from pairwise comparisons without reference labels. We establish identifiability up to natural normalizations and prove consistency and asymptotic normality of the maximum likelihood estimator, enabling confidence intervals for score differences and rank comparisons. Across multiple public benchmarks and a newly collected dataset, our method improves agreement with human preferences, achieves higher data efficiency than unweighted baselines, and produces calibrated uncertainty quantification for LLM rankings.
format Preprint
id arxiv_https___arxiv_org_abs_2601_21817
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Judge-Aware Ranking Framework for Evaluating Large Language Models without Ground Truth
Xu, Mingyuan
Tan, Xinzi
Wu, Jiawei
Zhou, Doudou
Machine Learning
Evaluating large language models (LLMs) on open-ended tasks without ground-truth labels is increasingly done via the LLM-as-a-judge paradigm. A critical but under-modeled issue is that judge LLMs differ substantially in reliability; treating all judges equally can yield biased leaderboards and misleading uncertainty estimates. More data can make evaluation more confidently wrong under misspecified aggregation. We propose a judge-aware ranking framework that extends the Bradley-Terry-Luce model by introducing judge-specific discrimination parameters, jointly estimating latent model quality and judge reliability from pairwise comparisons without reference labels. We establish identifiability up to natural normalizations and prove consistency and asymptotic normality of the maximum likelihood estimator, enabling confidence intervals for score differences and rank comparisons. Across multiple public benchmarks and a newly collected dataset, our method improves agreement with human preferences, achieves higher data efficiency than unweighted baselines, and produces calibrated uncertainty quantification for LLM rankings.
title A Judge-Aware Ranking Framework for Evaluating Large Language Models without Ground Truth
topic Machine Learning
url https://arxiv.org/abs/2601.21817