AReA: Adaptive Receptive-Field Attention for Efficient High-Resolution Diffusion Transformers Paper • 2610.17227 • Published 4 days ago • 49
Puppeteer: Object-Grounded Posture-Aware Co-Speech Gesture Generation Paper • 2609.00369 • Published Aug 31 • 21
SynthGait-19K: A Physically Grounded Synthetic Video Dataset for Gait Parameter Estimation Paper • 2609.08108 • Published Sep 8 • 18