4 papers
VDE: Training-Free Accelerating Rectified Flow Model via Velocity Decomposition and Estimation
Junwen Tan, Jinglin Liang, Hongyuan Chen +1
Though rectified flow models have achieved remarkable performance in image, video, and 3D generation, their practical deployments are challenged by slow inference speeds. Prior acc…
OrbitNVS: Harnessing Video Diffusion Priors for Novel View Synthesis
Jinglin Liang, Zijian Zhou, Rui Huang +2
Novel View Synthesis (NVS) aims to generate unseen views of a 3D object given a limited number of known views. Existing methods often struggle to synthesize plausible views for uno…
Order-Level Attention Similarity Across Language Models: A Latent Commonality
Jinglin Liang, Jin Zhong, Shuangping Huang +5
In this paper, we explore an important yet previously neglected question: Do context aggregation patterns across Language Models (LMs) share commonalities? While some works have in…
ReplayCAD: Generative Diffusion Replay for Continual Anomaly Detection
Lei Hu, Zhiyong Gan, Ling Deng +4
Continual Anomaly Detection (CAD) enables anomaly detection models in learning new classes while preserving knowledge of historical classes. CAD faces two key challenges: catastrop…