Mask2Keep: Mask-guided information transfer for backdoors resilient to compression-oriented pruning [0.03%]
Mask2Keep：用于抵抗面向压缩的剪枝的掩膜引导信息传输后门攻击
Jing Shang,Jian Wang,Kailun Wang et al.
Jing Shang et al.
Deep neural network (DNN) backdoor attacks implant hidden malicious behaviors during model training so that inputs containing a trigger are misclassified to an attacker-specified target. While prior studies have explored pruning as a defens...
JointConn-v2: Learning a joint vector field with diffusion transformers for cross-modal connectivity and dual-timestep modeling [0.03%]
联合连接v2：利用扩散变压器学习联合矢量场以进行跨模态连接和双时间步建模
Honggang Zhao,Yi-Jun Yang,Wei Zeng
Honggang Zhao
This work revisits diffusion Transformers for relative-depth-conditioned and joint image-depth synthesis, focusing on two bottlenecks: (1) cross-modal attention degrades around edges and structural regions, causing geometric distortions; (2...
Extending the scale generalization of the Vision Transformer without fine-tuning [0.03%]
无需微调即可扩展视觉变压器的尺度泛化能力
Kai Jiang,Peng Peng,Youzao Lian et al.
Kai Jiang et al.
The "train low, deploy high" paradigm offers significant practical advantages by minimizing training overhead while enabling high-fidelity inference through increased spatial resolutions. However, Vision Transformers (ViTs) often suffer fro...
Proactive structural stabilization for robust learning on heterogeneous graphs [0.03%]
基于异构图的鲁棒学习的主动结构稳定化方法研究
Jian Cao,Zeming Gan,Linlin Su et al.
Jian Cao et al.
Heterogeneous graph neural networks capture rich heterogeneous semantics through meta-path modeling, but such semantic propagation may also amplify the prediction instability under structural perturbations. Existing robustness methods are o...
TD-CAG: Enhancing adversarial transferability via curvature awareness and spatial dislocation [0.03%]
基于曲率感知和空间错位的对抗样本攻击能力增强方法
Hailing Kuang,Chen Wan,Xiaohai Lu et al.
Hailing Kuang et al.
Transfer-based adversarial attacks are widely used to evaluate the robustness of deep neural networks (DNNs) under black-box settings, yet improving their cross-model transferability remains a key challenge. This limitation arises from reli...
TD-CAG: Enhancing adversarial transferability via curvature awareness and spatial dislocation [0.03%]
基于曲率感知和空间错位的迁移攻击能力增强方法
Hailing Kuang,Chen Wan,Xiaohai Lu et al.
Hailing Kuang et al.
Transfer-based adversarial attacks are widely used to evaluate the robustness of deep neural networks (DNNs) under black-box settings, yet improving their cross-model transferability remains a key challenge. This limitation arises from reli...
Yaqi Chen,Shixun Huang,Lei Wang et al.
Yaqi Chen et al.
GNN prompting aims to adapt models across tasks and graphs without requiring extensive retraining. However, most existing graph prompt methods still require task-specific parameter updates and face the issue of generalizing across graphs, l...
DAVLMF-Seg: Vision-language model guided latent frequency-aware diffusion for semi-supervised medical image segmentation [0.03%]
基于视觉语言模型的潜频感知扩散半监督医学图像分割方法(DAVLMF-Seg)
Jiu Jiang,Qi Zhou,Nuo Chen et al.
Jiu Jiang et al.
Medical image segmentation is a fundamental task in computer-aided diagnosis and treatment planning. Fully supervised methods achieve strong performance but rely on large-scale annotated datasets. Semi-supervised learning (SSL) alleviates t...
CBS-PINN: Coordinated training to address nonlocality and spectral stiffness in the Calogero-Bogoyavlenskii-Schiff-type equation [0.03%]
CBS-PINN：协调训练以解决Calogero-Bogoyavlenskii-Schiff型方程的非局部性和谱刚度问题
Ting-Ting Jia,Ya-Juan Li,Jun Long et al.
Ting-Ting Jia et al.
The Calogero-Bogoyavlenskii-Schiff-type (CBS-type) equations can be employed to describe the properties of nonlinear wave propagation in hemodynamics, yet their simulation using Physics-Informed Neural Networks (PINNs) has been limited by s...
Fisher-Rao guided channel pruning with progressive re-estimation [0.03%]
基于 Fisher-Rao 向导的通道剪枝及逐步重新估计法
Xinjian Xiang,Mingjun Lin,Yongping Zheng et al.
Xinjian Xiang et al.
Structured channel pruning enables dense, deployment-friendly model compression, but its reliability depends on the channel-importance criterion and pruning schedule. We propose a Fisher-Rao guided framework that scores channel gates with a...