Suhyun Kim

Jailbreaking on Text-to-Video Models via Scene Splitting Strategy

Static, Initialization-based Layer-wise Learning Rates

Rethinking Regularization in Federated Learning: An Initialization Perspective

ReSafe: Enhancing Safety of Text-to-Image Diffusion via Post-Hoc Image Back Translation

Revisiting the Role of Homophily in Fair Graph Representation Learning

Lipschitz-aware Linearity Grafting for Certified Robustness

Shortcut Features as Top Eigenfunctions of NTK: A Linear Neural Network Case and More

Rethinking Entropy in Test-Time Adaptation: The Missing Piece from Energy Duality

When Model Knowledge meets Diffusion Model: Diffusion-assisted Data-free Image Synthesis with Alignment of Domain and Class

ELITE: Enhanced Language-Image Toxicity Evaluation for Safety

Initializing the Layer-wise Learning Rate