We provide a robust method of directly optimizing the pass at k with reinforcement learning, with theory and real world experiments.

A near-optimal quantum-accelerated Monte Carlo estimator for nested expectations