Saved in:
Bibliographic Details
Main Authors: Sun, Jie, Wu, Junkang, Wu, Jiancan, Zhu, Zhibo, Lu, Xingyu, Zhou, Jun, Ma, Lintao, Wang, Xiang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.03690
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914120554512384
author Sun, Jie
Wu, Junkang
Wu, Jiancan
Zhu, Zhibo
Lu, Xingyu
Zhou, Jun
Ma, Lintao
Wang, Xiang
author_facet Sun, Jie
Wu, Junkang
Wu, Jiancan
Zhu, Zhibo
Lu, Xingyu
Zhou, Jun
Ma, Lintao
Wang, Xiang
contents The alignment of Large Language Models (LLMs) is crucial for ensuring their safety and reliability in practical applications. Direct Preference Optimization (DPO) has emerged as an efficient method that directly optimizes models using preference pairs, significantly reducing resource demands. However, the effectiveness of DPO heavily depends on the data quality, which is frequently compromised by noise. In this work, we propose $γ$-PO, a dynamic target margin preference optimization algorithm that adjust reward margins at the pairwise level. By introducing instance-specific margin calibration, $γ$-PO strategically prioritizes high-confidence pairs (those demonstrating higher reward margins) while suppressing potential noise from ambiguous pairs. Moreover, $γ$-PO is a plug-and-play method, compatible with variants of DPO that rely on reward margin between preference pairs. Across benchmarks such as AlpacaEval2 and Arena-Hard, $γ$-PO achieves an average 4.4\% improvement over other baselines, setting new benchmarks for state-of-the-art performance. Additionally, $γ$-PO requires minimal code changes and has a negligible impact on training efficiency, making it a robust solution for enhancing LLMs alignment. Our codes are available at \href{https://github.com/sunjie279/gammaPO}{https://github.com/sunjie279/gammaPO}.
format Preprint
id arxiv_https___arxiv_org_abs_2506_03690
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Robust Preference Optimization via Dynamic Target Margins
Sun, Jie
Wu, Junkang
Wu, Jiancan
Zhu, Zhibo
Lu, Xingyu
Zhou, Jun
Ma, Lintao
Wang, Xiang
Computation and Language
The alignment of Large Language Models (LLMs) is crucial for ensuring their safety and reliability in practical applications. Direct Preference Optimization (DPO) has emerged as an efficient method that directly optimizes models using preference pairs, significantly reducing resource demands. However, the effectiveness of DPO heavily depends on the data quality, which is frequently compromised by noise. In this work, we propose $γ$-PO, a dynamic target margin preference optimization algorithm that adjust reward margins at the pairwise level. By introducing instance-specific margin calibration, $γ$-PO strategically prioritizes high-confidence pairs (those demonstrating higher reward margins) while suppressing potential noise from ambiguous pairs. Moreover, $γ$-PO is a plug-and-play method, compatible with variants of DPO that rely on reward margin between preference pairs. Across benchmarks such as AlpacaEval2 and Arena-Hard, $γ$-PO achieves an average 4.4\% improvement over other baselines, setting new benchmarks for state-of-the-art performance. Additionally, $γ$-PO requires minimal code changes and has a negligible impact on training efficiency, making it a robust solution for enhancing LLMs alignment. Our codes are available at \href{https://github.com/sunjie279/gammaPO}{https://github.com/sunjie279/gammaPO}.
title Robust Preference Optimization via Dynamic Target Margins
topic Computation and Language
url https://arxiv.org/abs/2506.03690