Abstract
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 gamma-PO, a dynamic target margin preference optimization algorithm that adjust reward margins at the pairwise level. By introducing instance-specific margin calibration, gamma-PO strategically prioritizes high-confidence pairs (those demonstrating higher reward margins) while suppressing potential noise from ambiguous pairs. Moreover, gamma-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, gamma-PO achieves an average 4.4% improvement over other baselines, setting new benchmarks for state-of-the-art performance.Additionally, gamma-PO requires minimal code changes and has a negligible impact on training efficiency, making it a robust solution for enhancing LLMs alignment.
Citation:
@inproceedings{sun2025gammapo,
title = {Robust Preference Optimization via Dynamic Target Margins},
author = {Sun, Jie and Wu, Junkang and Wu, Jiancan and Zhu, Zhibo and Lu, Xingyu and Zhou, Jun and Ma, Lintao and Wang, Xiang},
booktitle = {Proceedings of The 63rd Annual Meeting of the Association for Computational Linguistics},
pages = {},
year = {2025}
}
Jie Sun
孙杰
Junkang Wu
吴俊康
Jiancan Wu
吴剑灿 副研究员
