Abstract Covariate-constrained cluster randomization draws a treatment assignment from a constrained set of balanced allocations, and randomization-based inference relies on that set's distribution. We study what happens when an approximate sampler replaces the exact target. On one fixed, reproducible benchmark built from the 16-county Dickinson cluster-randomization example, we define an exact accepted set of 1,287 of 12,870 balanced allocations, obtained by exact rational scoring with a canonical index tie-break. We certify sampler outputs by total variation (TV) distance to this target, using full-statevector probabilities where available. On the anchor, ideal statevector results show that a fixed-step amplification control retains raw leakage of 0.00144, while two fixed-depth QAOA headline runs remain far from the target (raw TV 0.833 and 0.948). On four small synthetic instances, the energy-selected XY-QAOA outputs also remain far from the target (raw TV 0.59 to 0.72). The penalty control places 51% of its raw mass on wrong-weight strings. Its randomization-test p-value is undefined without an extension rule; we report an envelope from two fully defined extensions. For any fixed outcome realization and fully defined randomization-test event, the difference between the p-value under a sampler and under the exact target is bounded by TV. In a simulation anchored on the observed covariates with simulated outcomes, the exact reference holds sharp-null rejection near 5%, a biased classical control inflates type-I error, and the two statistics respond differently to the same distortion. The study demonstrates a certification workflow and its inferential consequences on one empirical-anchor benchmark, with a synthetic extension that is pre-specified, hash-recorded, and frozen before the relevant execution; it was not externally preregistered. Primary inference in the extension is limited to equal-size Gaussian outcomes; binary, heterogeneous, and unequal-size analyses are sensitivity-only. The study does not establish quantum advantage, hardware performance, noise robustness, or clinical benefit.
Abstract Covariate-constrained cluster randomization draws a treatment assignment from a constrained set of balanced allocations, and randomization-based inference relies on that set's distribution. We study what happens when an approximate sampler replaces the exact target. On one fixed, reproducible benchmark built from the 16-county Dickinson cluster-randomization example, we define an exact accepted set of 1,287 of 12,870 balanced allocations, obtained by exact rational scoring with a canonical index tie-break. We certify sampler outputs by total variation (TV) distance to this target, using full-statevector probabilities where available. On the anchor, ideal statevector results show that a fixed-step amplification control retains raw leakage of 0.00144, while two fixed-depth QAOA headline runs remain far from the target (raw TV 0.833 and 0.948). On four small synthetic instances, the energy-selected XY-QAOA outputs also remain far from the target (raw TV 0.59 to 0.72). The penalty control places 51% of its raw mass on wrong-weight strings. Its randomization-test p-value is undefined without an extension rule; we report an envelope from two fully defined extensions. For any fixed outcome realization and fully defined randomization-test event, the difference between the p-value under a sampler and under the exact target is bounded by TV. In a simulation anchored on the observed covariates with simulated outcomes, the exact reference holds sharp-null rejection near 5%, a biased classical control inflates type-I error, and the two statistics respond differently to the same distortion. The study demonstrates a certification workflow and its inferential consequences on one empirical-anchor benchmark, with a synthetic extension that is pre-specified, hash-recorded, and frozen before the relevant execution; it was not externally preregistered. Primary inference in the extension is limited to equal-size Gaussian outcomes; binary, heterogeneous, and unequal-size analyses are sensitivity-only. The study does not establish quantum advantage, hardware performance, noise robustness, or clinical benefit.
Calibrated support observations can constrain a complete quantum instrument without identifying its action by unrestricted tomography. We study what those observations guarantee when the noise law is not supplied. On a fixed four-dimensional incidence-record experiment, we define the set of all complete two-outcome completely positive instruments compatible with the exact raw rejection and outcome probabilities, retaining the classical flags. Exact ideal data identify the target, but an explicit family of complete instruments reproduces the nonzero-data category laws of a low-error contamination device while admitting coherent error of square-root order. A target-ray mass certificate and a common complete-dilation comparison give a universal upper, while an exactly feasible rational instrument supplies a stronger lower witness. At the diagnostic parameter r = 1/300000000, the unhalved target-relative worst-case radius lies between 0.000258065115532250 and 0.000951566242566. The result separates avoidable inequality conservatism from information absent from the measured data. Fixed-functional and fixed-map optimizations have separately certified scopes; they do not solve the global radius. The conclusions are conditional on the supplied carrier, calibration and common-device model. They do not establish an apparatus, device-independent certification, a finite-sample power guarantee or complete formal verification. Selected supporting declarations have one recorded successful Lean 4.19.0 macOS execution and declaration-specific source, type and axiom review; this is not an end-to-end formalization.
Abstract Purpose : Poor feasible sampling in full-space Penalty-X quantum approximate optimization algorithm (QAOA) can reflect representation-induced dilution, classical optimizer inadequacy, objective misalignment, or limited ansatz capacity. We separate these explanations on a controlled routing benchmark. Methods : Exact-statevector experiments use 140 resource-constrained shortest-path tasks and a scale-controlled Hamiltonian. Exact depth embeddings diagnose optimizer inadequacy. A frozen discovery/held-out design compares mean-energy training (O0), an exact-feasibility capacity control (O2), and conditional value at risk (CVaR)-0.10 (O3), with post-hoc depth, optimizer, tail-fraction, and finite-shot checks. Results : O2 showed that the fixed ansatz could reach feasible mass not induced by O0. On 84 held-out tasks from 15 graphs, O3 improved graph-level log-feasibility gain over O0 by +0.3547 decades and was non-inferior to O2: the O3–O2 effect was −0.0086, with lower bound −0.0360 above the frozen −0.10 margin. Certified nested failures occurred under COBYLA, Nelder–Mead, and SLSQP. Fixed-endpoint sampling was stable, whereas both finite-shot retraining intervals crossed zero. Conclusion : Optimizer inadequacy and objective misalignment are distinct under the tested diagnostics, and CVaR approaches the mechanistic O2 capacity control on held-out graphs. The classically tractable study supports no quantum-advantage or extrapolative scaling claim and does not test hardware noise.
When does a noisy quantum sampler yield an end-to-end polynomial-time optimization algorithm with performance guarantees? Building on finite-depth and finite-shot guarantees for Constraint-Enhanced QAOA, we show that inverse-polynomial ideal probability on the optimal set, together with independent sampling, polynomial-time feasibility repair, and scoring, produces an exact-hit fully polynomial randomized approximation scheme, which we call an FPRASq.
This guarantee survives device noise within an instance-dependent window. For effective circuit depth linear in the product of layer count and problem size, preserving an inverse-depth fraction of the ideal optimal mass increases the required shot complexity by one power of the problem size. Beyond this window, deterministic repair guarantees feasibility and provides an instance-dependent approximation guarantee whenever the induced objective inflation is controlled.
The resulting NP-HQ algorithm fits the Chen-Cotler-Huang-Li oracle model. On any NP-hard kernel-admissible promise family, reproducing its inverse-polynomial optimal overlap with a polynomial-time classical sampler would imply that NP is contained in BPP, even with identical repair and perfect access to the constraint structure. Thus, the separation lies in generating the sampling distribution.
We further introduce Heavy-Hitter QAOA, which preserves these conditional guarantees while reducing the retained candidate set and classical post-processing cost by one power of the problem size. Hardware experiments on IBM Eagle r3 processors cover instances with up to one hundred logical variables and match or improve every tested QOptlib reference tour.
Exact identification of a target instrument need not imply a linear stability bound for nearby observations. We study fixed finite-dimensional complete instruments with rank-one target Choi blocks, all outcomes retained, a fixed diagnostic contamination path with strictly positive branch noise, and an all-device bound on Choi mass outside the target rays. Under these hypotheses, linear prescribed-target certification is equivalent to injectivity of the joint observation and completeness map on the physical target-tangent space. An invisible nonzero tangent yields an exactly feasible positive instrument pair with square-root separation; a separate target-ray condition distinguishes sharp square-root behavior from constant-order ambiguity. For two prescribed three-basis readouts on a four-dimensional incidence-record benchmark, we obtain linear and square-root compatible-instrument radii, respectively. We then distinguish arbitrary probability residuals from simultaneous physical preparation and readout errors. Positivity and normalization give a support-amplitude estimate under which both protocols have linear calibration-only sensitivity, although their responses to contamination and unrestricted probability error differ. A fixed-sample application uses additional exactly trusted references to acquire operator-error bounds and obtain a simultaneous full-instrument confidence statement. A separate named-source power calculation reaches the prescribed error threshold with a conservative sufficient allocation of approximately 6.12 × 10^21 trials. The results concern a fixed laboratory-frame target and supplied common-device and sampling assumptions; they establish neither practical efficiency nor device-independent calibration. Written proofs are supported by exact finite certificates and a returned kernel-checked Lean subset; the full argument is not formalized end to end.
The observations available for a quantum instrument need not guarantee the accuracy of the device that generated them. On a fixed four-dimensional incidence-record benchmark, the paper compares two readout augmentations retaining six mixed preparations, both instrument outcomes, all attempts and shared completeness. Each adds three four-outcome bases and eighteen preparation/readout pairs. At r = 1/300000000, the 123 design admits a complete compatible instrument with unhalved prescribed-target diamond error at least 0.000103738856511, above the prospective 10^(-4) threshold. Replacing one basis gives the 124 design and a uniform upper 396319/8640000000 = 0.0000458702546296296... for every compatible complete instrument. The diagnostic generates data; its noise law and a Choi-rank bound are not imposed on the unknown device. The added basis effects observe ten Hermitian output directions, while original destination tests supply the leakage control. Five triples in the fixed finite menu are sufficient, one has a complete counterexample, and four remain unresolved. The paper claims neither a general measurement minimum, optimal radius, experimental certification, finite-sample power, calibrated hardware nor a general small-noise rate classification. Exact instrument identification remains obstructed on the unchanged preparations. Scientific cutoff: AUG04, 15 October 2026; AUG05 remains unexecuted. The companion contains editable manuscript sources, complete exact scientific arrays, the unchanged standalone reader, and accepted returned Mac Lean evidence for a supporting subset: 83 supporting theorems, 37 definitions and one abbreviation, with all 121 source/type/axiom correspondences. The complete-instrument theorems are not formalized end to end. Public execution views have documented path-only redactions; this preparation does not add a kernel run or external human peer review.
This registration describes a preregistered confirmatory study of heterogeneous pre-decoder conditioning in a rotated surface-code quantum error-correction workflow. Its purpose is to test whether, at a fixed total conditioning budget, heterogeneously distributed upstream conditioning produces lower logical error than budget-matched uniform conditioning, whether a single frozen conditioned representation remains effective across an architecture-diverse downstream decoder set, and whether performance varies nontrivially across a preregistered ordered target:non-target ratio grid. The study is conducted under frozen distances, frozen operating points, frozen seed and shot schedules, preregistered pooled metrics, and locked success criteria so that any supported effect can be interpreted as evidence about the architecture under test rather than post hoc tuning. The design contains four preregistered arms: untreated control, budget-matched uniform conditioning, budget-matched differential conditioning, and an ordered ratio-sweep arm with fixed total burden. A central feature of the protocol is that the conditioned representation is formed once upstream and then passed unchanged to each confirmatory decoder family, with no decoder-specific retuning, remapping, or auxiliary adaptation permitted after formation. The minimum confirmatory decoder set is minimum-weight perfect matching, Union-Find, and BP+OSD, selected to evaluate transfer across structurally different decoder families. Expected outcomes are defined in confirmatory terms. The primary pooled causal endpoint tests whether budget-matched differential conditioning outperforms budget-matched uniform conditioning, while the primary pooled transfer endpoint tests whether the same frozen conditioned representation improves performance relative to untreated control across decoder families. Additional preregistered questions address tunability through the ordered ratio response and persistence of favorable pooled effects across more than one primary code distance. A strong confirmatory outcome requires favorable pooled results with two-sided 95% bootstrap confidence intervals excluding 1 in the favorable direction, plus distance persistence at two or more primary distances; mixed, null, and thesis-changing outcomes are also explicitly predefined. This registration is limited to architecture-level evaluation and reproducible public inference. It does not claim that the frozen partition rule is uniquely optimal, that any specific targeting heuristic is the sole source of benefit, that threshold or asymptotic scaling is established, or that the exact coupling law, score function, or targeting recipe is disclosed. The confirmatory claim is bounded to whether budget-matched heterogeneous upstream conditioning improves logical performance and transfers across decoder families when the upstream representation is held fixed.