MaPPO: Maximum a Posteriori Preference Optimization with Prior Knowledge
This work addresses the challenge of improving alignment in LLMs for users by offering a more principled approach to preference optimization, though it is incremental as it builds upon existing methods like DPO.
The authors tackled the problem of aligning large language models with human preferences by proposing MaPPO, a framework that incorporates prior reward knowledge into preference optimization, resulting in consistent performance improvements on standard benchmarks like MT-Bench, AlpacaEval 2.0, and Arena-Hard without added computational cost.
As the era of large language models (LLMs) on behalf of users unfolds, Preference Optimization (PO) methods have become a central approach to aligning LLMs with human preferences and improving performance. We propose Maximum a Posteriori Preference Optimization (MaPPO), a framework for learning from preferences that explicitly incorporates prior reward knowledge into the optimization objective. While existing methods such as Direct Preference Optimization (DPO) and its variants treat preference learning as a Maximum Likelihood Estimation (MLE) problem, MaPPO extends this paradigm by integrating prior reward estimates into a principled Maximum a Posteriori (MaP) objective. This not only generalizes DPO and its variants, but also enhances alignment by mitigating the oversimplified binary classification of responses. More importantly, MaPPO introduces no additional hyperparameter, and supports preference optimization in both offline and online settings. In addition, MaPPO can be used as a plugin with consistent improvement on DPO variants, including widely used SimPO, IPO, and CPO. Extensive empirical evaluations of different model sizes and model series on three standard benchmarks, including MT-Bench, AlpacaEval 2.0, and Arena-Hard, demonstrate consistent improvements in alignment performance without sacrificing computational efficiency.