1 paper · 1 filter
Youngjoon Jeong, Jihwan Yu, Minsoo Jo +2
Latent action pretraining learns representations of visual change from pairs of observations, but existing methods typically encode each transition as a single unstructured represe…