Reinforcement Learning to Harness Approximation Errors for Long-Time Quantum Simulation
RL-Trotter uses reinforcement learning to treat unavoidable errors in quantum simulation as resources for correction. Instead of just suppressing these errors, the agent learns to adapt a single parameter—the next Trotter step size—to create self-correcting sequences over long times. This allows the simulation to maintain high accuracy without needing the full wave function, enabling practical protocols that generalize well to larger systems.