Threshold-Guided Optimization for Visual Generative Models
This work addresses the scalability limitation of pairwise preference optimization for aligning visual generative models with human feedback, offering a principled alternative that works directly from scalar ratings.
The authors propose a threshold-guided alignment framework that replaces the intractable instance-specific baseline in KL-regularized alignment with a global threshold estimated from scalar feedback, enabling optimization without paired comparisons. The method consistently improves preference alignment over previous methods across diffusion and masked generative models on three test sets and five reward models.
Aligning large visual generative models with human feedback is often performed through pairwise preference optimization. While such approaches are conceptually simple, they fundamentally rely on annotated pairs, limiting scalability in settings where feedback is collected as independent scalar ratings. In this work, we revisit the KL-regularized alignment objective and show that the optimal policy implicitly compares each sample's reward to an instance-specific baseline that is generally intractable. We propose a threshold-guided alignment framework that replaces this oracle baseline with a data-driven global threshold estimated from empirical score statistics. This formulation turns alignment into a binary decision task on unpaired data, enabling effective optimization directly from scalar feedback. We also incorporate a confidence weighting term to emphasize samples whose scores deviate strongly from the threshold, improving sample efficiency. Experiments across both diffusion and masked generative paradigms, spanning three test sets and five reward models, show that our method consistently improves preference alignment over previous methods. These results position our threshold-guided framework as a simple yet principled alternative for aligning visual generative models without paired comparisons.