Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Purification complexity in evaporating random circuits: a Page-transition benchmark

Ryosuke Shiroshita

Published Sep 7, 2026• 0 citations
Publisher

Abstract

We prove early preparation and entropy bounds for an explicit evaporating random-circuit model. Independent Haar two-qubit gates scramble an active register before each emission, and emitted qubits never interact again. Preparation complexity allows arbitrary one- and two-qubit gates and any finite number of initially pure ancillas, at fixed trace-distance error. We specialize the established two-copy purity method to permanent deletion and bound its evolution by a moving hyperbolic-cosine comparison profile. The first scrambling round gives a uniform relative bound on expected subsystem purity; positivity preserves this bound through every subsequent emission, even if later scrambling rounds are shortened or omitted. For the stipulated schedule, radiation before a logarithmic window around half emission has a linear preparation upper bound with an explicit failure bound. The entropy of every radiation prefix differs from its dimension ceiling by at most a logarithmic term, simultaneously with high probability. Numerical recurrence calculations, exact subset identities and small-system state-vector simulations provide reproducible checks. These results settle the early-preparation and entropy clauses of the benchmark. They do not establish its post-Page complexity lower bound at the accumulated gate-count scale. We identify the additional distributional information that such a lower bound requires and explain why the entropy Page law alone is insufficient. This preprint was written by exactory.ai (https://www.exactory.ai), an AI research system. The human author, Shiroshita, Ryosuke, authorized this deposit and is responsible for it.

DOI
10.5281/zenodo.22647764

Related papers

Pre-Registered Finite-Size Predictions for Page-Curve Experiments on Near-Term Quantum Hardware

Ben Holland

The Page curve, the rise and fall of radiation entropy during unitary evaporation, is now measurable on quantum simulators: trapped-ion and superconducting platforms routinely prepare scrambled states of 10–50 qubits and estimate subsystem Rényi entropies by randomized measurements. Semiclassical gravity, including the quantum-extremal-surface and replica-wormhole machinery that derives the Page curve, is constructed in the G → 0 limit and makes no prediction at finite qubit number. We state a set of exact finite-N laws for a concrete evaporation protocol — the location and determinism of the fine-grained transition, its first-order (whole-island) character, a quantized entanglement barrier with a per-qubit mechanism identity, the barrier's monotone response to scrambling depth, and a one-particle "mode books" formula that predicts the transition time under partial (graybody) emission — together with the complete experimental design: register sizes, gate budgets (~84–348 two-qubit gates at 12–24 qubits), a one-line noise mitigation certified to 3× typical hardware error rates, a measurement plan in which no register larger than max(t, N_I − t + 1) qubits is ever measured, and a total shot budget of order 10⁵. Every prediction is stated with the condition that falsifies it. All numbers were fixed before any experiment; the reproduction package included with this deposit reproduces every claim, including the exact hardware-numbers table, in under a minute.

•Aug 9, 2026•0 citations

Linear cross-entropy benchmarking with Clifford circuits

Jianxin Chen, Dawei Ding, Cupjin Huang +1 more

With the advent of quantum processors exceeding 100 qubits and the high engineering complexities involved, there is a need for holistically benchmarking the processor to have quality assurance. Linear cross-entropy benchmarking (XEB) has been used extensively for systems with 50 or more qubits but is fundamentally limited in scale due to the exponentially large computational resources required for classical simulation. In this work we propose conducting linear XEB with random Clifford circuits of constant to logarithmic depth, a scheme we call Clifford XEB. Since Clifford circuits can be simulated in polynomial time, Clifford XEB can be scaled to much larger systems. To validate this claim, we run numerical simulations for the classes of Clifford circuits we propose with noise and observe exponential decays. When noise levels are low, the decay rates are well correlated with the noise of each cycle assuming a multiplicative error accumulation, i.e., where the fidelities of individual gates multiply. We perform simulations of systems up to 1225 qubits, where the classical processing task can be easily dealt with by a workstation. Furthermore, using the theoretical guarantees in Chen et al. [PRX Quantum 3, 030320 (2022)], we prove that Clifford XEB with our proposed Clifford circuits must yield exponential decays under a general error model for sufficiently low errors. Our theoretical results explain some of the phenomena observed in the simulations and shed light on the behavior of general linear XEB experiments.

arXiv:2206.08293•Physical Review A•Nov 20, 2023•15 citations

Exact balanced subsystem designs for Haar-purity fluctuations

Ryosuke Shiroshita

We determine the minimum complex-Haar variance of the average purity over exactly n distinct balanced bipartitions of an n-party pure state, identifying complementary subsystems. For each n in {6,8,10,12,14,16}, the same explicit families are globally optimal for every integer equal local dimension d >= 2. The established fixed-pair Haar moment expresses the variance as a positive affine function of an even correlation cost. A parity floor settles n congruent to 2 modulo 4. For n congruent to 0 modulo 4, an established quadratic supersaturated-design bound combines with a strict scalar minorant, proved uniformly in d, to give the nonlinear optimum. The equality conditions characterize the entire minimizing set independently of dimension. We give exact score-gap identities, a direct sixteen-site grid construction, and an obstruction to attaining its bound by adjoining a column to a complete balanced orthogonal core. We also derive the exact mean variance of uniformly sampled families without replacement from the complement quotient. At fixed n>6, the relative gain has order d^(-2) with explicit coefficients; at n=6 every family is optimal. Explicit coordinates and a standalone exact-arithmetic checker accompany the proof. The results concern fluctuations across random states and provide finite benchmarks for selecting subsystems. Correspondence: Ryosuke.Shiroshita@exactory.ai This preprint was written by exactory.ai (https://www.exactory.ai), an AI research system. The human author, Shiroshita, Ryosuke, reviewed the full content and is responsible for it.

•Sep 15, 2026•0 citations

Hardness and Complexity Transition of Noisy Random Circuit Sampling

Byeongseon Go, Changhun Oh, Hyunseok Jeong

Random circuit sampling (RCS) is a leading candidate for demonstrating quantum advantage, supported by strong complexity-theoretic evidence of hardness in the ideal setting and by rapid experimental progress to date. In practice, however, noise is unavoidable, and a central problem is to identify the noise-strength boundary between classically simulable and classically hard regimes. In this work, we establish an architecture-general hardness bound for this boundary for the standard local depolarizing noise of strength $γ$. Assuming the standard average-case #P-hardness conjecture for ideal RCS, we show that, for any circuit architecture satisfying this conjecture, noisy RCS on the same architecture remains hard to simulate classically within any inverse-polynomial total variation distance whenever $γ=O(\log n/(nd))$ for $n$-qubit circuits of depth $d$, unless the polynomial hierarchy collapses. Crucially, noisy-RCS hardness follows without any additional conjectural or architecture-specific assumption beyond those already entering the ideal-RCS hardness framework. Our proof combines a low-degree polynomial extrapolation with a monotonicity reduction showing that efficient classical simulation at one depolarizing noise strength implies efficient simulation at every larger strength. Together, these ingredients transfer the standard ideal-RCS hardness conjecture to sampling hardness at a prespecified noise strength. Finally, combining the convergence-to-uniformity result of Dalzell et al. [Commun. Math. Phys. 405, 78 (2024)] with our monotonicity reduction yields efficient classical simulation for $γ=ω(\log n/(nd))$ on layered, regularly connected architectures. Thus, wherever the two architectural settings overlap, this identifies $γ=Θ(\log n/(nd))$ as the asymptotic complexity-transition scale.

quant-pharXiv:2607.20804•Jul 24, 2026•2 citations

Basis covariance and fluctuation diagnostics for measurements in closed quantum gravity

Ryosuke Shiroshita

Proposals for observations in closed quantum gravity must specify how unresolved microscopic alternatives become measurement outcomes. We compare a recent independent-sum prescription with the same authors' earlier diagonal sum. A three-dimensional calibration changes a normalized probability from 2/3 to 1/2 under an unresolved basis sign change, while the projector probability remains 3/5. We propagate the associated Gram repair through the state functional and identify its selected-ray postselection. We then extend the earlier finite Gaussian microscopic model to separable anisotropic covariance and arbitrary predetermined finite recorded experiments. Its derived weight covariances give a total variation bound without assuming independent outcomes. Bayes conditioning on successful preparation yields an explicit microsector law and a uniform conditional-history bound that vanishes as the environmental effective rank grows. A pair of successive noncommuting measurements shows that two instruments with the same first outcome probabilities retain a nonzero history discrepancy in this limit. The results apply to the specified random-state ensemble and clarify what normalization and environmental concentration can establish. They do not derive a gravitational action, gauge-covariant apparatus, or the full reconstruction required by the motivating Grand Challenge. This preprint was written by exactory.ai (https://www.exactory.ai), an AI research system. The human author, Shiroshita, Ryosuke, authorized its preparation and publication and is responsible for it.

•Sep 7, 2026•0 citations

Multi-qubit controlled gate synthesis without T-count overhead in the small-error limit

Soichiro Yamazaki, Seiseki Akibue, Kaoru Sano

We study the Clifford+T synthesis of quantum multiplexers $U=\bigoplus_{i=1}^{2^n}U_i$: an $n$-qubit control register selects which single-qubit gate $U_i\in\mathrm{SU}(2)$ acts on the target qubit. For each block $U_i$, consider the minimum even T-count needed to approximate it within diamond distance $\varepsilon$ without ancillae, and let $m$ be the largest of these costs. For an arbitrary multiplexer $U$, we construct an approximation of $U$ within the same distance using at most $m+O(\sqrt{2^n(m+1)})$ $T$ gates. With one control qubit, the cost is $m+O(1)$ and no ancillae are required. We also prove a lower bound that allows the Clifford+T circuit to use any number of clean ancillae initialized to zero and returned exactly to zero. For fixed $n$ and independent Haar-random blocks, the minimum T-count among these circuits lies between $(3-δ)\log_2(1/\varepsilon)$ and $(3+δ)\log_2(1/\varepsilon)$ for every fixed $δ>0$, except with probability $O(\varepsilon^{c_{n,δ}})$ as $\varepsilon\rightarrow0$. The optimal leading coefficient is therefore the same as for a single Haar-random $\mathrm{SU}(2)$ gate. In the lower-bound proof, unitarity forces pairs of off-diagonal blocks describing transitions in opposite directions to satisfy cancellation relations with residuals of order $\varepsilon^2$. We combine these relations with the arithmetic of Clifford+T circuits to bound the Haar measure of targets compatible with each pair of off-diagonal blocks. As an application, we obtain ancilla-free approximations of Haar-random $\mathrm{SU}(4)$ gates within diamond distance $\varepsilon$, using at most $(9+δ)\log_2(1/\varepsilon)$ $T$ gates for every fixed $δ>0$, except with probability $O(\varepsilon^{c_δ})$ with $c_δ>0$.

quant-pharXiv:2603.14202•Mar 17, 2026•1 citation
We prove early preparation and entropy bounds for an explicit evaporating random-circuit model. Independent Haar two-qubit gates scramble an active register before each emission, and emitted qubits never interact again. Preparation complexity allows arbitrary one- and two-qubit gates and any finite number of initially pure ancillas, at fixed trace-distance error. We specialize the established two-copy purity method to permanent deletion and bound its evolution by a moving hyperbolic-cosine comparison profile. The first scrambling round gives a uniform relative bound on expected subsystem purity; positivity preserves this bound through every subsequent emission, even if later scrambling rounds are shortened or omitted. For the stipulated schedule, radiation before a logarithmic window around half emission has a linear preparation upper bound with an explicit failure bound. The entropy of every radiation prefix differs from its dimension ceiling by at most a logarithmic term, simultaneously with high probability. Numerical recurrence calculations, exact subset identities and small-system state-vector simulations provide reproducible checks. These results settle the early-preparation and entropy clauses of the benchmark. They do not establish its post-Page complexity lower bound at the accumulated gate-count scale. We identify the additional distributional information that such a lower bound requires and explain why the entropy Page law alone is insufficient. This preprint was written by exactory.ai (https://www.exactory.ai), an AI research system. The human author, Shiroshita, Ryosuke, authorized this deposit and is responsible for it.
The Page curve, the rise and fall of radiation entropy during unitary evaporation, is now measurable on quantum simulators: trapped-ion and superconducting platforms routinely prepare scrambled states of 10–50 qubits and estimate subsystem Rényi entropies by randomized measurements. Semiclassical gravity, including the quantum-extremal-surface and replica-wormhole machinery that derives the Page curve, is constructed in the G → 0 limit and makes no prediction at finite qubit number. We state a set of exact finite-N laws for a concrete evaporation protocol — the location and determinism of the fine-grained transition, its first-order (whole-island) character, a quantized entanglement barrier with a per-qubit mechanism identity, the barrier's monotone response to scrambling depth, and a one-particle "mode books" formula that predicts the transition time under partial (graybody) emission — together with the complete experimental design: register sizes, gate budgets (~84–348 two-qubit gates at 12–24 qubits), a one-line noise mitigation certified to 3× typical hardware error rates, a measurement plan in which no register larger than max(t, N_I − t + 1) qubits is ever measured, and a total shot budget of order 10⁵. Every prediction is stated with the condition that falsifies it. All numbers were fixed before any experiment; the reproduction package included with this deposit reproduces every claim, including the exact hardware-numbers table, in under a minute.
With the advent of quantum processors exceeding 100 qubits and the high engineering complexities involved, there is a need for holistically benchmarking the processor to have quality assurance. Linear cross-entropy benchmarking (XEB) has been used extensively for systems with 50 or more qubits but is fundamentally limited in scale due to the exponentially large computational resources required for classical simulation. In this work we propose conducting linear XEB with random Clifford circuits of constant to logarithmic depth, a scheme we call Clifford XEB. Since Clifford circuits can be simulated in polynomial time, Clifford XEB can be scaled to much larger systems. To validate this claim, we run numerical simulations for the classes of Clifford circuits we propose with noise and observe exponential decays. When noise levels are low, the decay rates are well correlated with the noise of each cycle assuming a multiplicative error accumulation, i.e., where the fidelities of individual gates multiply. We perform simulations of systems up to 1225 qubits, where the classical processing task can be easily dealt with by a workstation. Furthermore, using the theoretical guarantees in Chen et al. [PRX Quantum 3, 030320 (2022)], we prove that Clifford XEB with our proposed Clifford circuits must yield exponential decays under a general error model for sufficiently low errors. Our theoretical results explain some of the phenomena observed in the simulations and shed light on the behavior of general linear XEB experiments.
We determine the minimum complex-Haar variance of the average purity over exactly n distinct balanced bipartitions of an n-party pure state, identifying complementary subsystems. For each n in {6,8,10,12,14,16}, the same explicit families are globally optimal for every integer equal local dimension d >= 2. The established fixed-pair Haar moment expresses the variance as a positive affine function of an even correlation cost. A parity floor settles n congruent to 2 modulo 4. For n congruent to 0 modulo 4, an established quadratic supersaturated-design bound combines with a strict scalar minorant, proved uniformly in d, to give the nonlinear optimum. The equality conditions characterize the entire minimizing set independently of dimension. We give exact score-gap identities, a direct sixteen-site grid construction, and an obstruction to attaining its bound by adjoining a column to a complete balanced orthogonal core. We also derive the exact mean variance of uniformly sampled families without replacement from the complement quotient. At fixed n>6, the relative gain has order d^(-2) with explicit coefficients; at n=6 every family is optimal. Explicit coordinates and a standalone exact-arithmetic checker accompany the proof. The results concern fluctuations across random states and provide finite benchmarks for selecting subsystems. Correspondence: Ryosuke.Shiroshita@exactory.ai This preprint was written by exactory.ai (https://www.exactory.ai), an AI research system. The human author, Shiroshita, Ryosuke, reviewed the full content and is responsible for it.
Random circuit sampling (RCS) is a leading candidate for demonstrating quantum advantage, supported by strong complexity-theoretic evidence of hardness in the ideal setting and by rapid experimental progress to date. In practice, however, noise is unavoidable, and a central problem is to identify the noise-strength boundary between classically simulable and classically hard regimes. In this work, we establish an architecture-general hardness bound for this boundary for the standard local depolarizing noise of strength γγγ. Assuming the standard average-case #P-hardness conjecture for ideal RCS, we show that, for any circuit architecture satisfying this conjecture, noisy RCS on the same architecture remains hard to simulate classically within any inverse-polynomial total variation distance whenever γ=O(log⁡n/(nd))γ=O(\log n/(nd))γ=O(logn/(nd)) for nnn-qubit circuits of depth ddd, unless the polynomial hierarchy collapses. Crucially, noisy-RCS hardness follows without any additional conjectural or architecture-specific assumption beyond those already entering the ideal-RCS hardness framework. Our proof combines a low-degree polynomial extrapolation with a monotonicity reduction showing that efficient classical simulation at one depolarizing noise strength implies efficient simulation at every larger strength. Together, these ingredients transfer the standard ideal-RCS hardness conjecture to sampling hardness at a prespecified noise strength. Finally, combining the convergence-to-uniformity result of Dalzell et al. [Commun. Math. Phys. 405, 78 (2024)] with our monotonicity reduction yields efficient classical simulation for γ=ω(log⁡n/(nd))γ=ω(\log n/(nd))γ=ω(logn/(nd)) on layered, regularly connected architectures. Thus, wherever the two architectural settings overlap, this identifies γ=Θ(log⁡n/(nd))γ=Θ(\log n/(nd))γ=Θ(logn/(nd)) as the asymptotic complexity-transition scale.
Proposals for observations in closed quantum gravity must specify how unresolved microscopic alternatives become measurement outcomes. We compare a recent independent-sum prescription with the same authors' earlier diagonal sum. A three-dimensional calibration changes a normalized probability from 2/3 to 1/2 under an unresolved basis sign change, while the projector probability remains 3/5. We propagate the associated Gram repair through the state functional and identify its selected-ray postselection. We then extend the earlier finite Gaussian microscopic model to separable anisotropic covariance and arbitrary predetermined finite recorded experiments. Its derived weight covariances give a total variation bound without assuming independent outcomes. Bayes conditioning on successful preparation yields an explicit microsector law and a uniform conditional-history bound that vanishes as the environmental effective rank grows. A pair of successive noncommuting measurements shows that two instruments with the same first outcome probabilities retain a nonzero history discrepancy in this limit. The results apply to the specified random-state ensemble and clarify what normalization and environmental concentration can establish. They do not derive a gravitational action, gauge-covariant apparatus, or the full reconstruction required by the motivating Grand Challenge. This preprint was written by exactory.ai (https://www.exactory.ai), an AI research system. The human author, Shiroshita, Ryosuke, authorized its preparation and publication and is responsible for it.
We study the Clifford+T synthesis of quantum multiplexers U=⨁i=12nUiU=\bigoplus_{i=1}^{2^n}U_iU=⨁i=12n​Ui​: an nnn-qubit control register selects which single-qubit gate Ui∈SU(2)U_i\in\mathrm{SU}(2)Ui​∈SU(2) acts on the target qubit. For each block UiU_iUi​, consider the minimum even T-count needed to approximate it within diamond distance ε\varepsilonε without ancillae, and let mmm be the largest of these costs. For an arbitrary multiplexer UUU, we construct an approximation of UUU within the same distance using at most m+O(2n(m+1))m+O(\sqrt{2^n(m+1)})m+O(2n(m+1)​) TTT gates. With one control qubit, the cost is m+O(1)m+O(1)m+O(1) and no ancillae are required. We also prove a lower bound that allows the Clifford+T circuit to use any number of clean ancillae initialized to zero and returned exactly to zero. For fixed nnn and independent Haar-random blocks, the minimum T-count among these circuits lies between (3−δ)log⁡2(1/ε)(3-δ)\log_2(1/\varepsilon)(3−δ)log2​(1/ε) and (3+δ)log⁡2(1/ε)(3+δ)\log_2(1/\varepsilon)(3+δ)log2​(1/ε) for every fixed δ>0δ>0δ>0, except with probability O(εcn,δ)O(\varepsilon^{c_{n,δ}})O(εcn,δ​) as ε→0\varepsilon\rightarrow0ε→0. The optimal leading coefficient is therefore the same as for a single Haar-random SU(2)\mathrm{SU}(2)SU(2) gate. In the lower-bound proof, unitarity forces pairs of off-diagonal blocks describing transitions in opposite directions to satisfy cancellation relations with residuals of order ε2\varepsilon^2ε2. We combine these relations with the arithmetic of Clifford+T circuits to bound the Haar measure of targets compatible with each pair of off-diagonal blocks. As an application, we obtain ancilla-free approximations of Haar-random SU(4)\mathrm{SU}(4)SU(4) gates within diamond distance ε\varepsilonε, using at most (9+δ)log⁡2(1/ε)(9+δ)\log_2(1/\varepsilon)(9+δ)log2​(1/ε) TTT gates for every fixed δ>0δ>0δ>0, except with probability O(εcδ)O(\varepsilon^{c_δ})O(εcδ​) with cδ>0c_δ>0cδ​>0.