arXiv:2610.24196 · EPJ Quantum Technology
Nonlocality as a Benchmarking Tool for Quantum Processors
Andrés Camilo Granda Arango, Carla Sophie Rieger, Carlo Cuccu, Giuseppe Sergioli, Luigi Iapichino, Tobias Valentin Bauer, Roberto Giuntini, Federico Hernán Holik
Abstract
Benchmarking quantum processors is a central task in every roadmap aimed at quantum advantage. Here, we discuss a benchmarking protocol based on the violation of the Svetlichny inequality, which is designed to simulate genuine multipartite nonlocality. We compare it with Mermin's inequality, and present the results of applying this framework to two different quantum processing units -the superconducting system Garnet (from IQM) and the ion-trap system Marmot (from AQT), testing different families of quantum circuits for three to five qubits under identical experimental conditions. We also run simultaneous triplet tests across the full register, including a comparison between GHZ and linear cluster states, to probe how Svetlichny-inequality violations depend on qubit location and calibration. Our results demonstrate that tests based on genuine multipartite nonlocality provide a powerful tool to characterize the collective performance of quantum devices and to guide their calibration and optimization.
AI Summary
The paper examines whether Svetlichny-inequality tests can benchmark collective processor performance more stringently than Mermin tests. It extends an existing protocol using GHZ states, Clifford circuits, and Clifford+$T$ circuits on superconducting Garnet and trapped-ion Marmot processors, principally with 3–5 qubits. Measurement directions are optimized numerically for ideal states, then the corresponding correlations are reconstructed experimentally. Random-circuit experiments select low-depth representatives across ideal violation ranges rather than executing the full sampled ensemble. The Svetlichny hybrid-local bound is $2^{N-1}$, with quantum maximum $2^{N-1}\sqrt{2}$. For GHZ states, Garnet yields $4.39\pm0.10$, $7.72\pm0.09$, and $12.3\pm0.50$ for 3, 4, and 5 qubits, respectively: only the three-qubit result exceeds its bound. Marmot yields $5.39\pm0.05$, $9.83\pm0.26$, and $18.68\pm0.74$, exceeding all three bounds. Both devices retain Mermin violations through five qubits. These values summarize 20 repetitions; uncertainties are standard deviations. Depolarizing-noise simulations show that the fraction of circuits retaining Svetlichny violations decreases faster than for Mermin violations, alongside a comparison with Grover-search success. Additional three-qubit tests resolve location- and time-dependent performance on Garnet. Simultaneous tests on four disjoint Marmot triplets show similar performance across triplets but greater dispersion than isolated tests. Stated limitations include the absence of spacelike-separated measurements, precluding loophole-free nonlocality certification; unequal shot counts between platforms; exponential measurement resources; and the fact that failure to violate Svetlichny does not exclude genuine multipartite nonlocality detectable by other inequalities.
Error found: The abstract's assertion of identical experimental conditions overlooks the tenfold difference in shots per circuit between Garnet and Marmot. The explanation that Svetlichny violations disappear because higher-order correlations degrade preferentially is unsupported: both operators contain full N-party correlators, and the GHZ data exhibit nearly equal fractional attenuation with different violation thresholds. The claim that random circuits make the benchmark unbiased overlooks explicit selection for ideal violations, shallow circuits, and low entangling-gate counts. Several reported normalized violation ratios have signs inconsistent with their definition, and that definition does not preserve the stated sign interpretation for states whose ideal expectation is below the classical bound.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Journal reference
- EPJ Quantum Technology, 2026
- Comments
- 23 pages, 14 figures, 5 tables