Year: 2026
Type: preprint
Abstract: We simulate an effective change in a single qubit's inner product (IP) by implementing the channel $\mathcal{F}^η$ on a superconducting quantum processor. Whereas previous experiments simulated parity-time (PT)-symmetric dynamics with a fixed IP, this experiment implements the IP-change channel itself for metrics associated with PT-symmetric systems. Building on an earlier adversarial verification approach, we derive a process-fidelity threshold that rules out mixtures of qubit unitary operations with random discarding of runs. We test the channel for 40 metrics using a single qutrit and a two-qubit Stinespring dilation. The measured fidelities exceed the threshold in 24 of 40 qutrit instances after readout error mitigation, compared with 16 of 40 unmitigated two-qubit instances. Without mitigation, 14 of 40 qutrit instances pass. These results demonstrate a direct simulation of an effective Hilbert-space IP change and its verification against a specified adversarial class, while showing the importance of readout treatment in comparing the encodings.
Source: arXiv (Cornell University)
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: G585908603, G716412294, G86560745
OpenAlex ID: W7235749193
