aiqbench analysis
This paper proposes QMemTwin, a digital twin for predicting entanglement decay in quantum repeater memories. It combines physics-constrained models, Akaike weighting, and an abstention gate for unsupported dynamics. The results are from synthetic data, and several key figures are unfilled placeholders in this version, so the claimed gains can't yet be assessed. Readers building quantum network controllers should treat this as a methods proposal, not a validated result.
Model-Uncertainty-Aware Digital Twins for Non-Markovian Quantum Repeater Memories
arXiv:2610.34506v1 Announce Type: new Abstract: Quantum-repeater controllers need reliable estimates of stored entanglement as adjacent links are generated. Fixed exponential lifetimes miss non-Markovian revivals, while complex surrogates are unnecessary when the microscopic family is known. We present QMemTwin, a physics-constrained digital twin that accounts for model uncertainty. It calibrates a bank of contractive passive state-space models, combines their amplitudes with finite-sample-corrected Akaike weights, and abstains when temporal-holdout data indicate unsupported dynamics. Passivity guarantees a completely positive, trace-preserving amplitude-damping channel for each candidate and their convex amplitude average. We evaluate [N cases] synthetic memories: [N supported] supported passive time-invariant cases and [N structural] nonstationary structural-shift cases. In the matched one-pseudomode regime, the correctly specified one-pseudomode fit performs best. Across [N unknown] supported cases with uncertain families, QMemTwin reduces population RMSE by [population reduction]% and Bell-fidelity RMSE by [Bell reduction]% relative to the two-pseudomode model; both paired effects remain significant after Holm correction. The flexible passive model has a slightly lower mean error for unknown families but is statistically indistinguishable from QMemTwin and loses accuracy in the matched regime. The temporal-holdout gate detects [detection]% of structural shifts, with [false alarm]% false alarms on supported families and AUROC [AUC]. In 2-, 4-, and 8-link repeater chains, the mixed cohort shows no significant 8-link goodput gain. A predeclared high-backflow stress cohort yields an 8-link goodput gain of [goodput gain]% over the two-pseudomode model, with paired p = [p value]. These results favor using a known family when available, accounting for model uncertainty otherwise and abstaining on unsupported dynamics.
Abstract:Quantum-repeater controllers need reliable estimates of stored entanglement as adjacent links are generated. Fixed exponential lifetimes miss non-Markovian revivals, while complex surrogates are unnecessary when the microscopic family is known. We present QMemTwin, a physics-constrained digital twin that accounts for model uncertainty. It calibrates a bank of contractive passive state-space models, combines their amplitudes with finite-sample-corrected Akaike weights, and abstains when temporal-holdout data indicate unsupported dynamics. Passivity guarantees a completely positive, trace-preserving amplitude-damping channel for each candidate and their convex amplitude average. We evaluate [N cases] synthetic memories: [N supported] supported passive time-invariant cases and [N structural] nonstationary structural-shift cases. In the matched one-pseudomode regime, the correctly specified one-pseudomode fit performs best. Across [N unknown] supported cases with uncertain families, QMemTwin reduces population RMSE by [population reduction]% and Bell-fidelity RMSE by [Bell reduction]% relative to the two-pseudomode model; both paired effects remain significant after Holm correction. The flexible passive model has a slightly lower mean error for unknown families but is statistically indistinguishable from QMemTwin and loses accuracy in the matched regime. The temporal-holdout gate detects [detection]% of structural shifts, with [false alarm]% false alarms on supported families and AUROC [AUC]. In 2-, 4-, and 8-link repeater chains, the mixed cohort shows no significant 8-link goodput gain. A predeclared high-backflow stress cohort yields an 8-link goodput gain of [goodput gain]% over the two-pseudomode model, with paired p = [p value]. These results favor using a known family when available, accounting for model uncertainty otherwise and abstaining on unsupported dynamics.
arXiv:2610.34506 [quant-ph]
(or arXiv:2610.34506v1 [quant-ph] for this version)
https://doi.org/10.48550/arXiv.2610.34506 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 340 words