Optimizing quantum error correction experiments with flux-tunable transmonsPart 2: higher accuracy error decoding*

In this second part, we present results from running small surface code error correction experiments on a 17-qubit flux-tunable transmon, with fixed-frequency couplers, quantum computer. We compare the performance of different decoding algorithms, including a neural-network-based one. We further demonstrate improvement in the logical fidelity by making use of additional analogue information obtained during readout, widely called soft information. Our results lead to a greater understanding of the performance of the superconducting device in surface code error correction contexts.