Model-based characterization on 10 qubits

Predictive gate-level models provide a valuable tool for characterizing the noise in quantum processors. Insight regarding the nature of the errors afflicting a system of qubits often identifies means of compensating for the errors and may suggest design choices for future hardware. Model-based characterization, however, is known to scale poorly due to the large number of model parameters and difficulty of circuit simulation. In this talk, we discuss the necessary tradeoffs between the number of model parameters, the depth of test circuits, and the precision of the result. We demonstrate on a 10-qubit system how model fitting and testing can be used to provide insight into underlying error mechanisms.