Self-Augmented Preference Optimization: Off-Policy Paradigms for Language Model Alignment

Yueqin Yin, Zhendong Wang, Yujia Xie, Weizhu Chen, Mingyuan Zhou

Abstract: Traditional language model alignment methods, such as Direct Preference Optimization (DPO), are limited by their dependence on static, pre-collected paired preference data, which restricts their adaptability and practical applicability. To address this limitation, we introduce Self-Augmented Preference Optimization (SAPO), an effective and scalable training paradigm without the need of existing paired data. Built upon the self-play concept that autonomously generate negative responses, we further involve the off-policy learning pipeline to improve the data exploration and exploitation. Specifically, we employ an Exponential Moving Average (EMA) model along with a replay buffer to enable dynamic updates of response segments, effectively integrating real-time feedback with historical data insights. Our comprehensive evaluations of the LLaMA3-8B and Mistral-7B models across benchmarks—including the Open LLM Leaderboard, IFEval, AlpacaEval 2.0, and MT-Bench—demonstrate that SAPO matches or surpasses established offline contrastive baselines, such as DPO and Odds Ratio Preference Optimization (ORPO), and outperforms offline self-play methods like SPIN.