Fig. 7: Error mitigation and distinguishability.

The top schematic illustrates the probability distribution of an observable A for two noisy states \({{{\mathcal{E}}}}(\psi )\) and \({{{\mathcal{E}}}}(\phi )\). The expectation values are shifted from the true values due to the noise effects. As in the bottom schematic, error mitigation converts them to other distributions whose expectation values are closer to the true values than the initial noisy distributions are. However, the converted distributions get broader, and the overlap between two distributions increases in general.