arXiv:2610.02137
One-Shot any Code
Andrew C. Yuan
Abstract
We first provide a simple proof showing that any single-shot code is self-correcting. We then show that any $[[n,k]]$ CSS quantum low-density parity-check (QLDPC) code can be transformed into an $[[nm,k]]$ CSS QLDPC code with single-shot (SS) quantum error correction and an efficient local decoder which runs in $O(\log m)$ parallel time. Below a constant threshold $p_{\rm RG}>0$ for joint local stochastic physical and measurement noise, the logical failure probability over $T$ correction rounds is bounded by \begin{equation*} O(T n)\exp[-Ω(m^α)] \end{equation*} for a constant $α>0$; we say that the \textit{threshold} is $p_{\rm RG}$ and \textit{error suppression} is $Ω(m^α)$. Moreover, if the input code is equipped with a single-shot decoder with threshold $p_c>0$ and error suppression $Ω(n^β)$, then the output code \textit{enhances} the error suppression to $Ω(m^α n^β)$, while maintaining the decoder-independent threshold $p_{\rm RG}$, and thus \textit{enhanced} if $p_c < p_{\rm RG}$.
AI Summary
Can arbitrary CSS QLDPC codes acquire single-shot protection without losing bounded check weight and qubit degree? This paper gives a recursive construction mapping an $[[n,k]]$ input to an $[[nm,k]]$ output, preserving $k$ and LDPC bounds up to constants depending on the input sparsity. Under joint local-stochastic data and measurement noise below a constant threshold $p_{\rm RG}>0$, the claimed failure bound over $T$ correction rounds is $O(Tn)(p/p_{\rm RG})^{\Omega(m^\alpha)}$, for a constant $\alpha>0$. The protocol uses one noisy syndrome-extraction round per correction and a final perfect measurement for assessing success. Its RG decoder takes $O(\log m)+\tau_{\rm seed}$ parallel time. If the seed has a single-shot decoder with threshold $p_c$ and suppression $\Omega(n^\beta)$, sufficient overhead gives suppression $\Omega(m^\alpha n^\beta)$, with a seed-decoder-independent threshold. The proof combines recursive code blow-ups with syndrome repair and cleaning. Cleaning contracts syndrome weight by $4/9$ per level; repair preserves a net contraction despite measurement faults. Bounded-degree hierarchical witness graphs control fault entropy and transfer the noise to a strongly suppressed effective seed process. Constant-duration local repair sweeps provide the parallel implementation. Separately, a Gibbs-syndrome local-stochastic bound establishes single-shot correction implies self-correction through decoder stability. The construction does not guarantee three-dimensional geometric locality or constant encoding rate as overhead grows. Practical constants and integration with fast surgery remain unresolved. The formal setup assumes local triviality of the seed; the explicit construction also uses simplifying incidence/connectivity hypotheses, whose removal is deferred to referenced work.
Central claim may not hold: The formal RG decoder has parallel runtime O(log m) + tau_seed, and the paper does not establish that the valid seed-decoding task can always be implemented within the abstract's unconditional O(log m) bound. The threshold proof requires a locally trivial seed and uses that assumption essentially, whereas the abstract covers arbitrary CSS QLDPC inputs without supplying a reduction that establishes this property.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- The simple proof (single-shot implies self-correction) is only 2 pages, and independent of everything else