Exploration is Harder than Prediction: Cryptographically Separating Reinforcement Learning from Supervised Learning.

Noah Golowich, Ankur Moitra, Dhruv Rohatgi

Browse the full FOCS paper archive.