Autonomous Quantum Computer Calibration

Short summary

Google researchers demonstrated a reinforcement learning system that continuously calibrates the Willow quantum processor during computation, eliminating the need to pause for recalibration. The system uses quantum error-correction data as training feedback, adjusting over 1,000 control parameters in real time. Results showed 3.5x more stable logical error rates under hardware drift and approximately 20% reduction in logical error rates beyond conventional calibration.

• •RL agent continuously calibrates Google's Willow quantum processor during computation
• •Error-correction data doubles as training signal for the AI system
• •Logical error rates improved ~20% beyond conventional expert tuning

Generated with AI, which can make mistakes.