HumanPlus provides two transformer-based systems for robot learning: the Humanoid Shadowing Transformer (HST) for reinforcement learning training in IsaacGym simulation, and the Humanoid Imitation Transformer (HIT) for real-world imitation learning from human demonstrations. The system uses whole-body pose estimation to enable a humanoid robot to replicate human motions, combining techniques from legged robotics, ACT, and Mobile ALOHA.