The control and handling of errors arising from cross-talk and unwanted
interactions in multi-qubit systems is an important issue in quantum
information processing architectures. Weintroduce a benchmarking protocol that
provides information about the amount of addressability present in the system
and implement it on coupled superconducting qubits. The protocol consists of
randomized benchmarking each qubit individually and then simultaneously, and
the amount of addressability is related to the difference of the average gate
fidelities of those experiments. We present the results on two similar samples
with different amounts of cross-talk and unwanted interactions, which agree
with predictions based on simple models for the amount of residual coupling.