Geunhyeok Yu

Geunhyeok Yu, Hyoseok Hwang

Deep Neural Networks (DNNs) have become pivotal in various fields, especially in computer vision, outperforming previous methodologies. A critical challenge in their deployment is the bias inherent in data across different domains, such as image style and environmental conditions, leading to domain gaps. This necessitates techniques for learning general representations from biased training data, known as domain generalization. This paper presents Attend to eXpert Prompts (A2XP), a novel approach for domain generalization that preserves the privacy and integrity of the network architecture. A2XP consists of two phases: Expert Adaptation and Domain Generalization. In the first phase, prompts for each source domain are optimized to guide the model towards the optimal direction. In the second phase, two embedder networks are trained to effectively amalgamate these expert prompts, aiming for an optimal output. Our extensive experiments demonstrate that A2XP achieves state-of-the-art results over existing non-private domain generalization methods. The experimental results validate that the proposed approach not only tackles the domain generalization challenge in DNNs but also offers a privacy-preserving, efficient solution to the broader field of computer vision.

Geunhyeok Yu, Minwoo Jeon, Hyoseok Hwang

The susceptibility of deep neural networks (DNNs) to adversarial examples has prompted an increase in the deployment of adversarial attacks. Image-agnostic universal adversarial perturbations (UAPs) are much more threatening, but many limitations exist to implementing UAPs in real-world scenarios where only binary decisions are returned. In this research, we propose Decision-BADGE, a novel method to craft universal adversarial perturbations for executing decision-based black-box attacks. To optimize perturbation with decisions, we addressed two challenges, namely the magnitude and the direction of the gradient. First, we use batch loss, differences from distributions of ground truth, and accumulating decisions in batches to determine the magnitude of the gradient. This magnitude is applied in the direction of the revised simultaneous perturbation stochastic approximation (SPSA) to update the perturbation. This simple yet efficient method can be easily extended to score-based attacks as well as targeted attacks. Experimental validation across multiple victim models demonstrates that the Decision-BADGE outperforms existing attack methods, even image-specific and score-based attacks. In particular, our proposed method shows a superior success rate with less training time. The research also shows that Decision-BADGE can successfully deceive unseen victim models and accurately target specific classes.

Geunhyeok Yu, Hyoseok Hwang

Recent deep learning models increasingly rely on depth without structural guarantees on the validity of intermediate representations, rendering early stopping and adaptive computation ill-posed. We address this limitation by formulating a structural requirement for state-space model's scale-consistent latent dynamics across iterative refinement, and derive Fractal of Stationary Transformations (FROST), which enforces a self-similar representation manifold through a fractal inductive bias. Under this geometry, intermediate states correspond to different resolutions of a shared representation, and we provide a geometric analysis establishing contraction and stable convergence across iterations. As a consequence of this scale-consistent structure, halting naturally admits a ranking-based formulation driven by intrinsic feature quality rather than extrinsic objectives. Controlled experiments on ImageNet-100 empirically verify the predicted scale-consistent behavior, showing that adaptive efficiency emerges from the aligned latent geometry.

Geunhyeok Yu, Sunjae Jeong, Yoonyoung Choi et al.

Vision Transformers are widely adopted as the backbone of vision foundation models, but they are known to produce high-norm artifacts that degrade representation quality. When knowledge distillation transfers these features to students, high-norm artifacts dominate the objective, so students overfit to artifacts and underweight informative signals, diminishing the gains from larger models. Prior work attempted to remove artifacts but encountered an inherent trade-off between artifact suppression and preserving informative signals from teachers. To address this, we introduce Singular Nullspace-Guided Energy Reallocation (SiNGER), a novel distillation framework that suppresses artifacts while preserving informative signals. The key idea is principled teacher feature refinement: during refinement, we leverage the nullspace-guided perturbation to preserve information while suppressing artifacts. Then, the refined teacher's features are distilled to a student. We implement this perturbation efficiently with a LoRA-based adapter that requires minimal structural modification. Extensive experiments show that \oursname consistently improves student models, achieving state-of-the-art performance in multiple downstream tasks and producing clearer and more interpretable representations.