Cycle Benchmarking for Scalable Verification of Quantum Circuit Performance

Predicting the reliability and capacity of quantum computer hardware, under some imperfectly known noise model, to perform a quantum computation, either with or without error correction, is a major challenge towards the realization of practical quantum computation. In this talk I will describe a practical framework, comprising randomized compiling and cycle benchmarking, which provides a complete and comprehensive solution to this problem. Randomized Compiling is a scalable method to dramatically reduce the most problematic errors affecting present-day quantum processors, including coherent (calibration) errors and non-Markovian errors, which tailors the error model to stochastic Pauli errors to very good approximation. Cycle benchmarking is a more efficient and practical method of randomized benchmarking which leverages randomized compiling to enable a precise and scalable method to bound the error probability of the quantum computer output for any algorithm or application based on a characterization of the actual error model affecting the hardware. These hardware-agnostic tools, which are available as a software system, imply a rigorous means of demonstrating quantum supremacy or quantum advantage in the regime beyond classically simulability.