Videos are Sample-Efficient Supervisions: Behavior Cloning from Videos via Latent Representations | Read Paper on Bytez