Changyu Dong — accepted AI-conference papers

2026

SlaClip: Gradient Norm Slacks can be Indicator for Adaptive Clipping in DP-SGD

Shuyan Zou, Shaowei Wang, Zhanxing Zhu, Jin Li, Changyu Dong, Vladimiro Sassone +1

ICML 2026spotlight

Differentially private stochastic gradient descent (DP-SGD) achieves privacy by clipping per-sample gradients and injecting Gaussian noise, but its utility is highly sensitive to the choice of the clipping threshold $C$. A fixed $C$ often degrades performance and necessitates repeated empirical cali…

Cited by 0 Source ↗Scholar ↗

2025

Distraction is All You Need for Multimodal Large Language Model Jailbreaking

Zuopeng Yang, Jiluan Fan, Anli Yan, Erdun Gao, Xin Lin, Tao Li +2

CVPR 2025highlight

Multimodal Large Language Models (MLLMs) bridge the gap between visual and textual data, enabling a range of advanced applications. However, complex internal interactions among visual elements and their alignment with text can introduce vulnerabilities, which may be exploited to bypass safety mechan…

Cited by 1 Source ↗PDF ↗Scholar ↗