Motus

“DreamZero [52] and LingBot-VA [26] adapt bidirectionally pretrained video generators to causal video–action prediction”