Fig. 3: Experimental shot reduction of HT readout circuits over TPBs.

The experiment is conducted with eight superconducting qubits on ibm_washington and compared to an ideal Qiskit simulation⁵³. a Error ϵ (see main text) as a function of the total number of optimally-allocated shots. Error bars show the error on the mean of ϵ averaged over ⌊5 × 10⁷/N^shots⌋ independent repetitions of the experiment. b Ratio \({N}_{{\rm{TPB}}}^{{\rm{shots}}}/{N}_{{\rm{HT}}}^{{\rm{shots}}}\) of shots needed to estimate 〈O〉 up to a total error of ϵ. In the absence of implementation errors (Theo.), this ratio is theoretically independent of ϵ and given by R_HT/R_TPB ≈ 1.87. See SM Sec. XIII for further details.