VideoRLCS. Learning to Identify Critical States for Reinforcement Learning from Videos (Accepted to ICCV'23)