AlphaNIST

AlphaNIST studies how dexterity and coordination can be learned in simulation with binary task rewards, cached starting states, and large-scale training, without prescribed manipulation stages.