Hyunsoo Kim

Self-Improving Diffusion Classifiers with Minority Preference Optimization

Hyunsoo Kim, Jungmyung Wi, Soobin Um, Donghyun Kim, Suhyun Kim

Prior studies have demonstrated that diffusion classifiers achieve robust zero-shot classification performance. However, their effectiveness is strongly tied to the pretraining data distribution: they perform well in majority, high-density regions of the data manifold, but are si…Read more