analiza aiqbench
This study demonstrates that optimizing for single-shot fidelity is suboptimal for minimizing wall-clock certification time in superconducting qubit readout. By modeling the measurement record as a stochastic channel, the authors identify a longer integration window that reduces total certification time by 9–11% while accounting for T1 relaxation and cavity memory effects. The work provides a concrete efficiency metric for hardware calibration, shifting the focus from peak fidelity metrics to throughput-relevant operational objectives.
Beyond Single-Shot Fidelity: Chernoff-Based Throughput Optimization in Superconducting Qubit Readout
arXiv:2602.22174v5 Announce Type: replace Abstract: Single-shot fidelity is the standard benchmark for superconducting qubit readout, yet it does not directly minimize the wall-clock time needed to certify a quantum state. We treat the dispersive measurement record as a stochastic communication channel and compute the classical Chernoff information governing the multi-shot error exponent, using a trajectory model that incorporates T1 relaxation with full cavity memory. The integration time that maximizes single-shot fidelity and the time that minimizes total certification time do not coincide. For representative transmon parameters and hardware overheads, the throughput-optimal window is longer, cutting certification time by roughly 9-11%, with the gain saturating near 1.13x in the high-readout-power and high-overhead regime. Benchmarking the extracted classical information against the unit-efficiency Gaussian Chernoff limit defines an information-extraction efficiency: dispersive schemes capture ~45% at short integration times, dropping to eta_info(tau_rate) ~ 12% at tau_rate ~ 1.22 us as T1-induced trajectory smearing accumulates. These results connect readout calibration directly to the operational objective of minimizing certification time in high-throughput superconducting processors.
Abstract:Single-shot fidelity is the standard benchmark for superconducting qubit readout, yet it does not directly minimize the wall-clock time needed to certify a quantum state. We treat the dispersive measurement record as a stochastic communication channel and compute the classical Chernoff information governing the multi-shot error exponent, using a trajectory model that incorporates T1 relaxation with full cavity memory. The integration time that maximizes single-shot fidelity and the time that minimizes total certification time do not coincide. For representative transmon parameters and hardware overheads, the throughput-optimal window is longer, cutting certification time by roughly 9-11%, with the gain saturating near 1.13x in the high-readout-power and high-overhead regime. Benchmarking the extracted classical information against the unit-efficiency Gaussian Chernoff limit defines an information-extraction efficiency: dispersive schemes capture ~45% at short integration times, dropping to eta_info(tau_rate) ~ 12% at tau_rate ~ 1.22 us as T1-induced trajectory smearing accumulates. These results connect readout calibration directly to the operational objective of minimizing certification time in high-throughput superconducting processors.
arXiv:2602.22174 [quant-ph]
(or arXiv:2602.22174v5 [quant-ph] for this version)
From: Sinan Bugu Mr. [view email] [v1] Wed, 25 Feb 2026 18:21:15 UTC (492 KB) [v2] Thu, 26 Feb 2026 18:36:23 UTC (483 KB) [v3] Wed, 4 Mar 2026 14:55:27 UTC (482 KB) [v4] Tue, 10 Mar 2026 17:15:49 UTC (482 KB) [v5] Tue, 2 Jun 2026 15:01:26 UTC (484 KB)
Text integral sub licența deschisă a publicației · 222 cuvinte