arXiv:2610.10531
Efficient Estimation of Logical Sensitivities Through Fault-Counting
Winston Fu, J. Wilson Staples, Jeff D. Thompson
Abstract
Quantum error-correcting circuits are affected by multiple physical noise mechanisms, whose contributions to logical failure must be understood to evaluate code performance and guide improvements in hardware. The individual error budget contributions for an error type $i$ can be characterized by its logical sensitivity $ν_i = \frac{\partial p_L}{\partial p_i}$, which measures the response of the logical error rate $p_L$ to each physical noise parameter $p_i$. Normally, $ν_i$ is measured using linear fits such as finite differences, where $p_L$ is measured at two or more values of $p_i$ to calculate partial derivatives. In this paper, we develop a differentiable estimator to obtain all components of $\nabla_{\mathbf p}p_L$ simultaneously from a single Monte Carlo data set at one noise configuration, using information about the underlying fault configurations. In surface code simulations with circuit-level noise, the estimator agrees with conventional finite differences while requiring one to two orders of magnitude fewer shots to achieve the same variance. We apply this technique to quantify error budgets, resolve sensitivities at the individual qubit level, and infer effective code distance. Finally, we incorporate the sensitivities into Newton root finding to locate and trace threshold contours in multidimensional noise models. These results provide an efficient method for extracting and applying logical sensitivity information from standard quantum error correction simulations.
AI Summary
The paper addresses how to estimate logical-error sensitivities $\nu_i=\partial p_L/\partial p_i$ for many physical noise parameters without separate simulation sweeps for each parameter. It applies the score-function identity to sampled physical fault configurations, giving $\nu_i=\mathbb E[L(E)\,\partial_{p_i}\log P_{\mathbf p}(E)]$, where $L(E)$ indicates logical failure. All gradient components can therefore be estimated from one Monte Carlo dataset, with faults grouped afterward by error type, location, or circuit round. Circuit-level simulations of unrotated surface codes at distances $3,5,7$, using physical error rates from $10^{-3}$ to $10^{-2}$, compare this estimator with central finite differences while holding the decoder fixed. The estimates agree within statistical uncertainty. For five noise types and a finite-difference step of $0.1p_i$, the reported shot reduction at matched variance is one to two orders of magnitude. An approximate variance-ratio calculation supports the observed reduction. Applications include spatial sensitivity maps, error-budget decomposition, and estimating effective distance from the local logarithmic slope of $p_L$. The sensitivities also drive Newton root finding and tangent-based tracing of two-parameter logical-error crossing contours, demonstrated with PyMatching, correlated PyMatching, and Tesseract. Independent samples validate the converged crossings. These finite-distance contours only approximate asymptotic thresholds. The estimator requires access to sampled faults and assumes independently sampled stochastic mechanisms; demonstrations use Pauli noise. Extensions to other stochastic mechanisms and higher derivatives are proposed but not demonstrated. The paper also notes concurrent similar work.
The authors acknowledge the use of LLMs to assist in ideation, coding, data analysis, and proofreading the manuscript.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.