4 papers
Random Is Hard to Beat: Active Selection in online DPO with Modern LLMs
Giyeong Oh, Junghyun Lee, Jaehyun Park +3
Modern LLMs inherit strong priors from web-scale pretraining, which can limit the headroom of post-training data-selection strategies. While Active Preference Learning (APL) seeks…
Vicinity-Guided Discriminative Latent Diffusion for Privacy-Preserving Domain Adaptation
Jing Wang, Wonho Bae, Jiahong Chen +2
Recent work on latent diffusion models (LDMs) has focused almost exclusively on generative tasks, leaving their potential for discriminative transfer largely unexplored. We introdu…
SteeringTTA: Guiding Diffusion Trajectories for Robust Test-Time-Adaptation
Jihyun Yu, Yoojin Oh, Wonho Bae +2
Test-time adaptation (TTA) aims to correct performance degradation of deep models under distribution shifts by updating models or inputs using unlabeled test data. Input-only diffu…
What Has Been Overlooked in Contrastive Source-Free Domain Adaptation: Leveraging Source-Informed Latent Augmentation within Neighborhood Context
Jing Wang, Wonho Bae, Jiahong Chen +3
Source-free domain adaptation (SFDA) involves adapting a model originally trained using a labeled dataset ({\em source domain}) to perform effectively on an unlabeled dataset ({\em…