arXiv:2610.06535
Symmetry and AI-assisted discovery of magic-state factories
Shubham P. Jain, Adam Wills, Shraddha Singh
Abstract
Magic-state distillation is a major resource cost in fault-tolerant quantum computing. The cost of a magic-state factory depends strongly on its failure rate, which grows with the number of input magic states. Although symmetry-restricted methods have recently made distance two searches tractable, distance three and above have remained elusive at moderate input counts. We develop symmetry- and AI-assisted methods to search this regime. We present a unified binary-matrix formulation encompassing both triorthogonal-code and direct circuit searches. We show that distance at least three is equivalent to nonzero, pairwise distinct syndromes, separating the choice of syndromes from the search for compatible output gates. We restrict the syndrome search using group symmetry and language-model agents, followed by deterministic solving and independent verification. Our searches yield 699 factory classes, including 564 new ones. These include factories for pure-T states and factories with entangled outputs comprising combinations of T, CS, and CCZ magic states. The pure-T factories [[63, 11, 3]] and [[850, 128, 6]] achieve the lowest overhead exponents we know among protocols with at most 100 and 1000 inputs, respectively, with $γ= 1.589$ and $γ= 1.057$. Our [[1715, 287, 6]] factory, with $γ= 0.998$, is the smallest known pure-T factory with $γ< 1$. We also provide a context directory of search briefs and campaign notes with which readers can train their own agents and tailor the search to their requirements. With these results, we begin constructing an active, open-source repository of magic-state distillation protocols for the quantum community, supplemented by our methods and data, for the practical fault-tolerant quantum computing regime.
AI Summary
The paper searches for magic-state distillation protocols with moderate input counts and distance at least three, targeting pure $T$ states and diagonal third-level magic states, including mixed $T$, $CS$, and $CCZ$ outputs. It restricts attention to postselected diagonal CSS protocols with independent input $Z$ faults. A binary-matrix formulation unifies generalized triorthogonal-code and direct circuit searches. For reduced circuits with distinct nonzero columns, distance at least three is equivalent to nonzero, pairwise distinct check syndromes. This separates syndrome-set selection from compatible-output search. Group-orbit restrictions compress the search; LLM agents propose symmetries, syndrome sets, and punctures, while deterministic solvers construct candidates and an independent verifier checks gates, rank, and distance claims. Large-block results primarily use puncture searches within existing code frameworks. The catalog contains 699 parameter/output classes, 564 identified as new, spanning distances three through nine. Highlighted pure-$T$ factories include $[[63,11,3]]$ with overhead exponent $\gamma=1.589$, $[[850,128,6]]$ with $\gamma=1.057$, and $[[1715,287,6]]$ with $\gamma=0.998$. These are reported as the best known exponent below 100 inputs, best below 1000 inputs, and smallest known pure-$T$ factory with $\gamma<1$, respectively. Explicit circuits, search tools, verification tools, and campaign context are released. Searches are heuristic rather than exhaustive beyond classified ranges; symmetry restrictions narrow their scope. Comparisons exclude catalytic and Hadamard-assisted protocols. Input counts and asymptotic exponents do not establish architectural savings: acceptance probabilities, leading error coefficients, Clifford implementation costs, and application-specific output consumption remain to be assessed.
Within that framework, large language models were the proposing half of the loop of Sec.~\ref{sec:ai-methods}: they proposed parents, symmetry groups and search directions and, in the later runs, pruned candidates as reviewers.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.