arXiv:2610.06606
Staying Sober: A Shot-Free Predictor of Logical Error Rates for Concatenated Quantum Error Correcting Codes
Sayam Sethi, Aditi Awasthi, Maxwell Poster, Joshua Viszlai, Jonathan Mark Baker
Abstract
Estimating the logical error rate (LER) of a quantum error-correcting (QEC) code by Monte Carlo sampling takes $O(p^{-\lceil d/2 \rceil}/\varepsilon^2)$ samples at physical error rate $p$, code distance $d$ and relative error $\varepsilon$, which becomes intractable at large distances and low error rates, and especially for concatenated codes, whose distance grows as $d^{\ell}$ with the number of levels $\ell$. The fail fast method reduces this cost by sampling the failures of each weight separately, but it only works for unconcatenated codes, and does not model the soft information that the levels of a concatenated code pass to each other. We propose a predictor that takes the stabilizers, the minimum-weight logical operators and, for odd $d$, the logical operators of weight $d + 1$ of a code, and computes, without sampling, the leading monomial $c\,p^{o}$ of the LER when expressed as a polynomial in $p$, at every level of concatenation, in time polynomial in the code parameters and linear in $\ell$ for fixed budgets. We prove that our prediction is exact for any unconcatenated stabilizer code, for every logical observable and arbitrary per-qubit error rates, whenever no budget is exhausted. Furthermore, for concatenated codes, we construct a recursive approach that computes the leading monomial exactly in exponential time, and our predictor approximates this approach. Our predictions match exact enumeration at level 1 on every code we evaluate, and up to three levels for the iceberg code. Using our predictor, we evaluate and compare $37$ code configurations under four noise biases in $48$ core-hours, where Monte Carlo sampling would need more than $10^{12}$ shots at $p = 10^{-3}$ for $113$ of these $148$ cases.
AI Summary
Estimating very small logical error rates (LERs) by direct Monte Carlo becomes impractical as code distance and concatenation depth increase. This paper targets the low-noise leading term $c p^o$ for maximum-likelihood decoding of stabilizer codes and soft tree decoding of concatenated codes, rather than the full finite-$p$ LER. For an unconcatenated distance-$d$ code, it identifies leading-weight failures using weight-$d$ logical operators, weight-$(d+1)$ logical operators when $d$ is odd, and low-weight stabilizers. Möbius inversion over logical-operator intersections avoids double counting. The resulting coefficient is exact, subject to the stated leading-order tie convention, for observables flipped by a minimum-weight logical operator, with $o=\lceil d/2\rceil$ and arbitrary positive per-qubit Pauli biases. An exact but exponential recursion propagates distributions of likelihood-order gaps and coefficient ratios through concatenation. A budgeted dynamic program approximates this recursion; for fixed budgets its runtime is polynomial in distance and supplied logical-operator count, and linear in concatenation depth. Small-code enumeration agrees within $0.1\%$ at level 1 and $p=3.2\times10^{-4}$, and within $0.2\%$ for iceberg concatenation through level 3. Comparing 37 configurations under four biases takes 48 core-hours. The main setting assumes independent Pauli errors and perfect syndrome measurements. Higher-level approximations discard correlations and truncate configurations without an accuracy guarantee. Tree decoding is ML only for single-logical-qubit blocks. Complete logical-operator enumeration can be costly or unavailable; exhausted budgets can trigger sampling. The asymptotic predictions have no general finite-$p$ validity bound.
Central claim may not hold: The abstract promises exactness for every logical observable, whereas the unconcatenated theorem only covers observables flipped by a minimum-weight logical operator. The predictor uses sampling when certain budgets are exceeded, and some evaluated inputs are sampled, so its sampling-free description needs explicit qualification. The leading-term tie rule can disagree with actual maximum-likelihood decoding and produce an incorrect leading logical-error coefficient for an individual observable, even under the unconcatenated theorem's assumptions.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 25 pages, 10 figures