Benchmarking Quantum Instruments

Quantum measurements with feed-forward are crucial components of fault-tolerant quantum computers. We show how the error rate of such a measurement can be directly estimated by fitting the probability that successive randomly compiled measurements all return the ideal outcome. Unlike conventional randomized benchmarking experiments and alternative measurement characterization protocols, all the data can be obtained using a single sufficiently large number of successive measurements. We also prove that generalized Pauli fidelities are invariant under randomized compiling and can be combined with the error rate to characterize the underlying errors up to a gauge transformation that introduces an ambiguity between errors happening before or after measurements.