6 papers
Understanding In-Context Multimodal Jailbreaks via Posterior Reweighting
Xu Zhang, Dev Mistry, Xiang Xu +1
In-context learning (ICL) jailbreaks reveal a critical vulnerability in multimodal large language models (MLLMs): harmful demonstrations in the prompt can induce unsafe outputs wit…
Communication-Efficient Personalized Federated Learning via Layer-Wise Multi-Threshold Random Sketching
Xu Zhang, Xingyu Hou, Jiacheng Cheng +2
Personalized federated learning (PFL) is a promising paradigm for collaborative learning over distributed devices, where edge nodes collaboratively train personalized models withou…
Exploring Dynamic Properties of Backdoor Training Through Information Bottleneck
Xinyu Liu, Xu Zhang, Can Chen +1
Understanding how backdoor data influences neural network training dynamics remains a complex and underexplored challenge. In this paper, we present a rigorous analysis of the impa…
Personalized Federated Learning with Bidirectional Communication Compression via One-Bit Random Sketching
Jiacheng Cheng, Xu Zhang, Guanghui Qiu +3
Federated Learning (FL) enables collaborative training across decentralized data, but faces key challenges of bidirectional communication overhead and client-side data heterogeneit…
Optimizing Robustness and Accuracy in Mixture of Experts: A Dual-Model Approach
Xu Zhang, Kaidi Xu, Ziqing Hu +1
Mixture of Experts (MoE) have shown remarkable success in leveraging specialized expert networks for complex machine learning tasks. However, their susceptibility to adversarial at…
FGP: Feature-Gradient-Prune for Efficient Convolutional Layer Pruning
Qingsong Lv, Jiasheng Sun, Sheng Zhou +6
To reduce computational overhead while maintaining model performance, model pruning techniques have been proposed. Among these, structured pruning, which removes entire convolution…