robomimic Multi-Human (MH) demonstrations of the "PickPlaceCan" task, recorded on a simulated Franka-style arm (robot_type = robomimic). Each episode is a teleoperated demonstration with per-frame end-effector state and delta actions, annotated with a dense, scaled reward signal (next.reward, reaching 100 at task success).… See the full description on the dataset page: https://huggingface.co/datasets/VGalaxies666/robomimic-mh-can-image-dense-scaled-100.