🏢 Huazhong University of Science and Technology

New unlearnable framework secures 3D point cloud data by using class-wise transformations, enabling authorized training while preventing unauthorized access.

ADV-TRA uses adversarial trajectories to robustly fingerprint deep neural networks, outperforming state-of-the-art methods against various removal attacks.

Twin-Merging dynamically merges modular model expertise, significantly improving multitask performance without retraining, and adapting to diverse data.

Self-Distilled Depth Refinement (SDDR) tackles noisy depth maps via a novel noisy Poisson fusion approach, achieving significant improvements in depth accuracy and edge quality.

Effortlessly boost large language model performance by dynamically fusing knowledge from smaller, task-specific models – achieving near full fine-tuning results with minimal computational cost!

MoE Jetpack efficiently transforms readily available dense checkpoints into high-performing MoE models, drastically accelerating convergence and improving accuracy.

LMSPS: a novel framework efficiently leverages long-range dependencies in large heterogeneous graphs by dynamically identifying effective meta-paths, mitigating computational costs and over-smoothing.

LION: Linear Group RNNs conquer 3D object detection in sparse point clouds by enabling efficient long-range feature interaction, significantly outperforming transformer-based methods.

Lightweight Frequency Masker significantly improves cross-domain few-shot semantic segmentation by cleverly filtering frequency components of images, thereby reducing inter-channel correlation and enh…

GCFormer, a novel graph Transformer, enhances node representation learning by employing a hybrid token generator and contrastive learning, outperforming existing methods on various datasets.

New criterion maximizes remaining discrepancy after rationale removal, treating spurious features as noise, improving rationale extraction.

IR-CM: One-step image restoration using a novel consistency model for fast and universal performance.

Deep network pruning’s fundamental limits are characterized, revealing how weight magnitude and network sharpness determine the maximum achievable sparsity.

Researchers developed FSAFW, a novel framework generating universal adversarial perturbations effective against various Few-Shot Learning paradigms, surpassing baseline methods by over 16%.

FasterDiT accelerates Diffusion Transformers training 7x without architecture modification by analyzing SNR probability density functions and implementing a new supervision method.

Adaptive likelihood estimation and MAP inference during reverse diffusion tackles real-world image noise.

DarkSAM, a novel prompt-free attack, renders the Segment Anything Model incapable of segmenting objects across diverse images, highlighting its vulnerability to universal adversarial perturbations.

Coupled Mamba: Enhanced multi-modal fusion via coupled state space model boosts accuracy and efficiency.

Boosting Vision Transformer’s transferability in cross-domain few-shot learning is achieved by a simple yet effective method: strategically adjusting attention temperature to remedy ineffective target…

Node Injection-based Fairness Attack (NIFA) reveals GNNs’ vulnerability to realistic fairness attacks by injecting a small percentage of nodes, significantly undermining fairness even in fairness-awar…