VA-AR: Learning Velocity-Aware Action Representations with Mixture of Window Attention

Re-assign community