HumanVLA: Towards Vision-Language Directed Object Rearrangement by Physical Humanoid | Read Paper on Bytez