Yang Hu

Yang Hu

Auto-generated profile, based on publicly available information

Yang Hu

Beihang University, Beijing, China

Research Focus & Metrics

Representative Publications

View All Publications
11 July 2026
Verified
Conference Paper
Open access
YHYang Hu
Yang Hu
HYHaotian Ye
Haotian Ye
YHYudi Huang
Yudi Huang
QSQingyun Sun
Qingyun Sun
XBXiuli Bi
Xiuli Bi
JLJianxin Li
Jianxin Li
Multimodal diffusion models have become a central approach for high-quality content generation. Some recent studies explore cross-timestep gradient relations to guide improvements in diffusion training. However, commonly used full-gradient cosine similarity can be distorted by parameter heterogeneity and layer-wise scale differences, leading to misleading estimates of the coupling among timestep-specific objectives. This paper revisits cross-timestep gradient measurement and proposes a corrected protocol for multimodal diffusion training. The corrected measurement shows that gradient similarity decays rapidly with timestep distance, and gradients from distant timesteps become nearly orthogonal as training progresses. These results suggest that diffusion training consists of weakly coupled timestep-wise objectives. They further imply that directly applying history-accumulating mechanisms across timesteps may introduce interference among weakly related timestep objectives. Our study refines the empirical basis for timestep-wise optimization dynamics analysis and provides guidance for optimization design in multimodal content generation.

9 November 2026
Verified
Conference Paper
Open access
YHYang Hu
Yang Hu
TYTao Yang
Tao Yang
YHYuheng He
Yuheng He
QSQingyun Sun
Qingyun Sun
XBXiuli Bi
Xiuli Bi
JLJianxin Li
Jianxin Li

28 October 2024
Verified
Conference Paper
Open access
XBXiuli Bi
Xiuli Bi
YHYang Hu
Yang Hu
BLBo Liu
Bo Liu
WLWeisheng Li
Weisheng Li
PCPamela Cosman
Pamela Cosman
BXBin Xiao
Bin Xiao
As machine learning advances, machine learning as a service (MLaaS) in the cloud brings convenience to human lives but also privacy risks, as powerful neural networks used for generation, classification or other tasks can also become privacy snoopers. This motivates privacy preservation in the inference phase. Many approaches for preserving privacy in the inference phase introduce multi-objective functions, training models to remove specific private information from users' uploaded data. Although effective, these adversarial learning-based approaches suffer not only from convergence difficulties, but also from limited generalization beyond the specific privacy for which they are trained. To address these issues, we propose a method for privacy preservation in the inference phase by removing task-irrelevant information, which requires no knowledge of the privacy attacks nor introduction of adversarial learning. Specifically, we introduce a metric to distinguish task-irrelevant information from task-relevant information, and achieve more efficient metric estimation to remove task-irrelevant features. The experiments demonstrate the potential of our method in several tasks. Our code will be available at: https://github.com/iwhoyoung/PriFU.

6 November 2026
Verified
Conference Paper
Open access
TYTao Yang
Tao Yang
YHYang Hu
Yang Hu
FLFeihong Lu
Feihong Lu
QSQingyun Sun
Qingyun Sun
JLJianxin Li
Jianxin Li