AnyFlow: Any-Step Video Diffusion Model with On-Policy Flow Map Distillation Paper โข 2605.13724 โข Published 20 days ago โข 101
HumanNet: Scaling Human-centric Video Learning to One Million Hours Paper โข 2605.06747 โข Published 26 days ago โข 52
MHLA: Restoring Expressivity of Linear Attention via Token-Level Multi-Head Paper โข 2601.07832 โข Published Jan 12 โข 53