Almost device-independent calibration beyond Born's rule: Bell tests for cross-talk detection

Abstract

In quantum information, device-independent protocols offer a new approach to information processing tasks, making minimal assumptions about the devices used. Typically, since these protocols draw conclusions directly from the data collected in a meaningful Bell test, the no-signaling conditions, and often even Born's rule for local measurements, are taken as premises of the protocol. Here, we demonstrate how to test such premises in an (almost) device-independent setting, i.e., directly from the raw data and with minimal assumptions. In particular, for IBM's quantum computing cloud services, we implement the prediction-based ratio protocol to characterize how well the qubits can be accessed locally and independently. More precisely, by performing a variety of Clauser-Horne-Shimony-Holt-type experiments on these systems and carrying out rigorous hypothesis tests on the collected data, we provide compelling evidence showing that some of these qubits suffer from cross-talks, i.e., their measurement statistics are affected by the choice of measurement bases on another qubit. Unlike standard randomized benchmarking, our approach does not rely on assumptions such as gate-independent Markovian noise. Moreover, despite the relatively small number of experimental trials, the direction of "signaling" may also be identified in some cases. Our approach thus serves as a complementary tool for benchmarking the local addressability of quantum computing devices.