Self-Supervised ‘Plan2Explore’ RL Agent Achieves SOTA Zero-Shot and Adaptation Performance

Enter Plan2Explore — a self-supervised RL agent designed to quickly generalize to unseen tasks in a zero or few-shot manner.