Pareto-Guided Optimal Transport for Multi-Reward Alignment
For practitioners of text-to-image generation, this work addresses the critical problem of multi-reward alignment and reward hacking, offering a principled framework that significantly improves both automated metrics and human preference.
Text-to-image generation models suffer from reward hacking when optimizing multiple reward models via weighted summation. The proposed PG-OT framework constructs prompt-specific Pareto frontiers and uses optimal transport to guide samples toward them, achieving an 11% gain in Joint Domination Rate and near 80% win rate in human evaluations.
Text-to-image generation models have achieved remarkable progress in preference optimization, yet achieving robust alignment across diverse reward models remains a significant challenge. Existing multi-reward fusion approaches rely on weighted summation, which is costly to tune and insufficient for balancing conflicting objectives. More critically, optimization with reward models is highly susceptible to reward hacking, where reward scores increase while the perceived quality of generated images deteriorates. We demonstrate that optimizing against a unified global target under heterogeneous reward upper bounds can induce reward hacking, a risk further exacerbated by the inherent instability of weak reward models. To mitigate this, we propose a Pareto Frontier-Guided Optimal Transport (PG-OT) framework. Our method constructs a prompt-specific Pareto frontier and maps dominated samples toward it via distribution-aware optimal transport. Furthermore, we develop both online and offline optimization strategies tailored to diverse reward signal characteristics. To provide a more rigorous assessment, we introduce the Joint Domination Rate (JDR) and Joint Collapse Rate (JCR) as principled metrics to quantify multi-reward synergy and reward hacking. Experimental results show that our approach outperforms strong baselines with an 11% gain in JDR and achieves a near 80% win rate in human evaluations.