Fig. 4: Simulation of QIRB on few-qubit QPUs with a simple error model.
a, b RB curves from two two-qubit [(a)] and two [(b)] six-qubit QIRB-r simulations performed using \({p}_{{\mathsf{cnot}}}=.35\) and \({p}_{{\mathsf{MCM}}}=.01\) (blue) or \({p}_{{\mathsf{MCM}}}=.5\) (orange). The points show \({\overline{F}}_{d}\), and the violin plots show the distributions of \(\langle {s}_{\tilde{C}}\rangle\) over circuits of that depth. QIRB has succeeded in all four cases as evidenced by the exponential decay of \({\overline{F}}_{d}\). c A plot comparing rΩ from eight two-, four-, and six-qubit simulations (blue, orange, and green, respectively) with varying \({p}_{{\mathsf{MCM}}}\) and fixed \({p}_{{\mathsf{cnot}}}=.35\) (the points) against our predictions for rΩ (the lines) based on the data-generating model. The error bars are 1σ error bars generated by performing each simulation eight times. These results validate QIRB as the rΩ observed in each simulation matches that predicted by our theory. d Table containing the extracted one-qubit gate, \({\mathsf{cnot}}\), and MCM error rates (resp. \({\varepsilon }_{{{{\rm{1Q}}}}}{{,}}\,{\varepsilon }_{{{{\rm{2Q}}}}}{{,and}}\,{\varepsilon }_{{\mathsf{MCM}}}\)). We see close agreement between the estimated error rates and those used to generate the model. See Supplementary Note 5 for additional details.
