Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Zhuo, Li, Moxin, Deng, Xun, Wang, Qifan, Feng, Fuli
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2505.19176
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914045346447360
author Liu, Zhuo
Li, Moxin
Deng, Xun
Wang, Qifan
Feng, Fuli
author_facet Liu, Zhuo
Li, Moxin
Deng, Xun
Wang, Qifan
Feng, Fuli
contents LLM-as-a-Judge employs large language models (LLMs), such as GPT-4, to evaluate the quality of LLM-generated responses, gaining popularity for its cost-effectiveness and strong alignment with human evaluations. However, training proxy judge models using evaluation data generated by powerful teacher models introduces a critical yet previously overlooked issue: teacher preference bias, where the proxy judge model learns a biased preference for responses from the teacher model. To tackle this problem, we propose a novel setting that incorporates an additional assistant model, which is not biased toward the teacher model's responses, to complement the training data. Building on this setup, we introduce AGDe-Judge, a three-stage framework designed to debias from both the labels and feedbacks in the training data. Extensive experiments demonstrate that AGDe-Judge effectively reduces teacher preference bias while maintaining strong performance across six evaluation benchmarks. Code is available at https://github.com/Liuz233/AGDe-Judge.
format Preprint
id arxiv_https___arxiv_org_abs_2505_19176
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Assistant-Guided Mitigation of Teacher Preference Bias in LLM-as-a-Judge
Liu, Zhuo
Li, Moxin
Deng, Xun
Wang, Qifan
Feng, Fuli
Computation and Language
LLM-as-a-Judge employs large language models (LLMs), such as GPT-4, to evaluate the quality of LLM-generated responses, gaining popularity for its cost-effectiveness and strong alignment with human evaluations. However, training proxy judge models using evaluation data generated by powerful teacher models introduces a critical yet previously overlooked issue: teacher preference bias, where the proxy judge model learns a biased preference for responses from the teacher model. To tackle this problem, we propose a novel setting that incorporates an additional assistant model, which is not biased toward the teacher model's responses, to complement the training data. Building on this setup, we introduce AGDe-Judge, a three-stage framework designed to debias from both the labels and feedbacks in the training data. Extensive experiments demonstrate that AGDe-Judge effectively reduces teacher preference bias while maintaining strong performance across six evaluation benchmarks. Code is available at https://github.com/Liuz233/AGDe-Judge.
title Assistant-Guided Mitigation of Teacher Preference Bias in LLM-as-a-Judge
topic Computation and Language
url https://arxiv.org/abs/2505.19176