1 paper · 1 filter
Zuojin Tang, Feifan Luo, Haoyun Liu +10
The paper introduces DLAM, a distributional latent-action model that encodes video transitions as diagonal Gaussians with temporal constraints, improving reconstruction consistency…