The cost of measuring quantum expectation values of an operator can be reduced by grouping
the Pauli string (SU(2) tensor product) decomposition of the operator into maximally commuting
sets. We detail an algorithm, presented in [1], to partition the full set of m-qubit Pauli strings
into the minimal number of commuting families, and benchmark the performance with dense
Hamiltonians on IBM hardware. Here we also compare how our method scales compared to
graph-theoretic techniques for the generally commuting case.