Abstract: Reinforcement learning (RL) has enabled robots to develop complex skills, but its success in image-based tasks often depends on effective representation learning. Prior works have primarily ...