🏢 Nanyang Technological University

Tri-Hom disentangles graph homophily into label, structural, and feature aspects, providing a more comprehensive and accurate metric for predicting GNN performance.

UMI-GRAT: A universal meta-initialized and gradient robust adversarial attack effectively exploits vulnerabilities in the Segment Anything Model (SAM) and its fine-tuned downstream models, even withou…

Variance Reduction Fine-tuning (VRF) simultaneously boosts in-distribution and out-of-distribution accuracy in fine-tuned zero-shot models, overcoming the ID-OOD trade-off.

BeTopNet uses braid theory to create a topological representation of multi-agent future driving behaviors, improving prediction and planning accuracy in autonomous driving systems.

Language Hierarchical Self-training (LHST) enhances weakly-supervised object detection by integrating language hierarchy, mitigating label mismatch, and improving generalization across diverse dataset…

MVGamba: A unified, feed-forward 3D content generation model achieving state-of-the-art quality and speed using an RNN-like state space model for efficient multi-view Gaussian reconstruction.

Concentric Causal Attention (CCA) significantly reduces object hallucination in LVLMs by cleverly reorganizing visual tokens to mitigate the impact of long-term decay in Rotary Position Encoding.

Proactive Infeasibility Prevention (PIP) framework significantly improves neural methods for solving complex Vehicle Routing Problems by proactively preventing infeasible solutions and enhancing const…

Animates diverse garments realistically from a single cloth’s trajectory using a disentangled learning approach and Energy Unit Network (EUNet).

Hybrid Mamba Network (HMNet) boosts few-shot segmentation accuracy by efficiently fusing support and query features using a novel hybrid Mamba architecture, significantly outperforming current state-o…

HisTPT: Historical Test-Time Prompt Tuning memorizes past learning, enabling robust online prompt adaptation for vision models, overcoming performance degradation in continuously changing data streams…

Generalizable Implicit Motion Modeling (GIMM) revolutionizes video frame interpolation by accurately predicting optical flows at any timestep, surpassing existing methods and achieving state-of-the-ar…

FASTopic: a pretrained transformer-based topic model achieving superior speed, adaptivity, stability, and transferability compared to existing methods.

DRAGON: A novel GNN framework using distributed-order fractional calculus surpasses traditional methods by capturing complex graph dynamics with enhanced flexibility and performance.

ContextGS: Revolutionizing 3D scene compression with an anchor-level autoregressive model, achieving 15x size reduction in 3D Gaussian Splatting while boosting rendering quality.

Researchers developed AdvRM, a new adversarial patch attack against monocular depth estimation models, which effectively camouflages patches as road markings to mislead depth predictions for any obsta…

ART: A novel automatic red-teaming framework reveals safety vulnerabilities in popular text-to-image models by identifying unsafe outputs even from seemingly harmless prompts.