CMU, SFU, and Nvidia researchers introduce Hybrid Imitation Learning for humanoid control without relying on explicit phase variables · Digg