{"id":"8ee740a5-3022-4008-bea0-82ff1f1ef75f","arxiv_id":"2505.04285","paper_version":1,"verdict":"UNVERDICTED","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A review of the Russian Quantum Center's 2020-2024 quantum software roadmap, summarizing algorithms, emulators, error correction, and cloud execution, with no new results.","lead":"An industrial research team reviews its own five-year quantum computing software program, covering algorithms, emulators, error correction, and a cloud platform. It is useful as a status report on Russia's state-funded quantum computing efforts, not as a source of new scientific results.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central cloud-execution claim is asserted without verifiable data; cited references may only document direct hardware runs, not cloud-mediated ones.","rationale":"The reader's weakest_assumption focuses on the fpQAOA fixed-angles hypothesis in Section II A. That is a real concern: the transferability of near-optimal parameters is an empirical conjecture cited from [45-47], and no generalization guarantee is given. However, it is not the most load-bearing issue for the paper's central claim. The abstract's headline result is the execution of quantum algorithms using a cloud-based quantum computing platform, and the body of the paper supports this only with an unsupported assertion. The paper itself lays out the central claim in the abstract and Section VII A, but provides no data, logs, or references that specifically demonstrate cloud-mediated execution as opposed to direct hardware access. Because the review's distinctive contribution is the integrated software stack and remote access platform, the cloud-execution claim is the load-bearing one. My concern does not refute the paper; it identifies a missing-support condition. The reader's verdict of UNVERDICTED remains appropriate, and I recommend no change to the verdict.","tokens_in":44035,"tokens_out":4265,"duration_ms":43676,"concrete_test":"Check Refs. [38], [39], [55], [120], and [156] for explicit statements that the experimental runs were performed through the described cloud-based quantum computing platform (e.g., submitted via its API or web interface). In parallel, request from the authors a sample of platform job identifiers or execution logs for the 'at least 1000 launches per year since 2021,' including timestamps, input QASM files, target processor IDs, and output bitstring statistics. If no cited reference describes cloud-mediated execution and no logs are provided, the abstract's headline claim should be downgraded to unverified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract and Section VII present as the headline result the execution of quantum algorithms using a cloud-based quantum computing platform. Section VII states that at least 1000 experimental launches per year have been performed since 2021 and that algorithms were 'experimentally executed' on real ion, neutral-atom, superconducting, and photonic processors via the platform. However, the paper provides no execution logs, output distributions, job counts with dates/processors, or other verifiable artifacts. The only evidence offered is the assertion itself and a schematic figure (Figs. 3 and 5). The cited works [38,39,55,120,156] report algorithms run on trapped-ion (and superconducting in [38]) hardware, but none of the in-text descriptions indicate that those runs were submitted through the described cloud platform as distinct from direct laboratory control. If the runs were direct hardware experiments, then the distinguishing achievement claimed in the abstract—execution via the cloud platform—remains unsupported. This is a missing-support problem for the central claim, not a disagreement with the field consensus.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript is a review of a software stack developed under the Russian Quantum Center / Rosatom quantum-computing roadmap (2020–2024). It surveys quantum algorithms (QAOA variants for linear systems, differential equations, and factoring; quantum chemistry algorithms; Hamiltonian simulation and signal-processing approaches), monitoring and benchmarking tools, compilation and optimization methods including qudit implementations, error suppression and correction, quantum processor emulators, and a cloud platform for quantum computing. The abstract and Section VII identify as the headline result the execution of quantum algorithms on real ion, neutral-atom, superconducting, and photonic processors through the developed cloud platform. The body is written in Russian apart from the English abstract, and the paper includes appendices listing implemented algorithms, error-correction methods, and emulators.","tokens_in":44245,"tokens_out":3562,"duration_ms":40357,"significance":"If its central claims are substantiated, the paper documents a functioning multi-hardware quantum software ecosystem with cloud access, which would be of interest to the quantum computing community as a national-roadmap case study. The manuscript has useful strengths: it provides a concrete catalog of algorithms and emulators, gives algorithmic details for the emulators (e.g., the Clifford-Clifford boson sampling procedure in Section VI.C), and presents runtime comparisons with Qiskit, Strawberry Fields, and Perceval. However, the headline cloud-execution claim is currently supported only by assertions and schematic figures, not by verifiable execution data, and several technical performance claims are presented without proof or with circular validation. The significance of the review is therefore conditional on the evidence being supplied.","major_comments":[{"comment":"The central claim that quantum algorithms were executed on real processors via the cloud platform is not supported by verifiable evidence. The text states that at least 1000 experimental launches per year have been performed since 2021 and lists ion, neutral-atom, superconducting, and photonic processors, but no execution logs, job counts, dates, or output data are given. The cited references [38,39,55,120,156] describe runs on trapped-ion and other hardware, yet none of the in-text descriptions indicate that those runs were submitted through the described cloud platform as distinct from direct hardware laboratorial control. The authors should either provide verifiable platform artifacts (e.g., anonymized job records or a public endpoint) or explicitly reformulate the claim to what the cited references actually support.","section":"§VII, §VII.A"},{"comment":"The claimed guaranteed residual bound for the fpQAOA-based linear-systems solver is stated without proof. After the QUBO formulation in Eq. (4), the text asserts that the algorithm produces an answer with residual not exceeding t√n 2^{1−k}, where t is an unspecified norm bound, n is the dimension, and k is the number of bits, but no derivation or reference to a derivation is provided. Since this is presented as a formal guarantee for a key algorithm, the paper must either supply the proof or clearly state where it appears.","section":"§II.A (linear-systems algorithm)"},{"comment":"The emulator-to-device validation is circular because the noise parameters were fitted to maximize the fidelity between emulator and device outputs. Agreement on the same data used for fitting cannot be reported as evidence that the emulator adequately reflects the physical device. This is especially problematic in Table VIII, where the noisy boson-sampling emulator gives fidelities of 46.54% and 28.05% for 2 and 3 input photons, substantially worse than the ideal emulator values; this contradicts the narrative that the fitted noisy model captures the device. The authors should report error bars, use a held-out set of circuits, and discuss the Table VIII discrepancy explicitly.","section":"§VI.D, Tables V–VIII"},{"comment":"The fixed-angles hypothesis for QAOA is load-bearing for the claimed efficiency of the linear-systems, differential-equation, and factoring algorithms, but the paper provides no quantitative cross-validation protocol or generalization analysis for the universal angles. The statement that the hypothesis was 'confirmed' for the verification classes is not accompanied by the size of the training set, the number of test instances, or statistical significance measures. Either provide this evidence or explicitly label the transferability assumption as an open empirical conjecture on which the reported speedups depend.","section":"§II.A (fpQAOA fixed-angles hypothesis)"}],"minor_comments":[{"comment":"The body of the paper is written in Russian while the abstract is in English; if the target venue is an international journal, a full English version is required.","section":"Language"},{"comment":"There is a typo in the Russian text: 'ползволяет' should be 'позволяет'; also the residual bound 't√n21−k' is typeset in a way that makes the exponent of 2 unclear.","section":"§II.A"},{"comment":"The text contains typos such as 'кторая' and 'соответвуют'; these should be corrected throughout.","section":"§IV"},{"comment":"The fidelity values in Tables V–VIII are reported without error bars or the number of experimental shots, which makes it difficult to assess whether differences between emulator and device are statistically meaningful.","section":"Tables V–VIII"},{"comment":"The reference list mixes published papers and arXiv preprints without consistently noting the published versions; please standardize the citation format.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is essentially a self-review of the authors' own roadmap output, and much of the evidence consists of citations to their own prior papers. That is not disqualifying for a review article, but it makes the lack of independent verifiable artifacts for the cloud-execution headline more serious. If the authors can provide execution logs or clearly delimit the claim to direct hardware runs rather than cloud-mediated runs, the paper would be much more defensible."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nThis is a useful roadmap review of five years of quantum software work under the Rosatom roadmap. If you want a compact map of what one team built—QAOA variants, VQE for molecules, qudit Toffoli decompositions, a single-ancilla QEC code, emulators for ion and photonic hardware, and a cloud access platform—this is a good starting point. The emulator benchmark tables (II–IV) and device-agreement tables (V–VIII) are concrete, and the compilation section gives real gate counts before and after optimization. The authors clearly label the paper as a review and mostly refer to their prior work.\n\nThe main soft spot is the abstract's headline claim: \"execution of quantum algorithms using a cloud-based quantum computing platform.\" Section VII repeats it but gives no execution logs, no job counts with dates and processor types, no output distributions, nothing that lets a reader verify runs went through the platform rather than direct laboratory control. The cited references describe hardware experiments, but they don't obviously establish the cloud-mediated workflow. That is a missing-support problem for the central advertised result.\n\nSmaller issues: the fixed-angle fpQAOA hypothesis is load-bearing for several algorithm descriptions and is cited to prior work rather than argued here—fine for a review, but it remains an empirical conjecture. The emulator agreement numbers are calibration, not prediction, since noise parameters were fitted to maximize fidelity; no error bars are given. A few technical claims, like the guaranteed residual bound for the linear-systems algorithm, are stated without proof, though the original papers are cited.\n\nThe audience is people tracking Russia's quantum software roadmap or looking for a bibliography on qudit compilation and emulator design. It doesn't change scientific understanding, but it is a useful inventory. A serious referee should be sent this, with the expectation that they demand either verification of the cloud claim or a rewrite that demotes it from headline to aspiration. The rest of the material is substantial enough to deserve referee time.\n\nRecommendation: invite revision. Ask for concrete support for the cloud-execution claim—logs, dates, processor names, platform-specific documentation—and require that the emulator agreement tables be labeled as fitted calibration. If the cloud claim is dropped or qualified, this is publishable as a scoped review.","headline":"Useful roadmap review of a national QC software program; headline cloud-execution claim is asserted without verifiable evidence.","tokens_in":44857,"tokens_out":3453,"would_cite":false,"duration_ms":34332,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A multi-year quantum computing roadmap produced a complete software stack and ran real algorithms on physical processors through a single cloud platform.","keywords":["quantum algorithms","cloud quantum computing","QAOA","fixed-angle conjecture","qudits","quantum emulators","quantum error correction","quantum benchmarking"],"falsifier":"Take a problem class the review does not cover, such as Max-Cut on graphs with unbounded degree or an ill-conditioned linear system, fix the angles by the same training procedure, then measure the success probability on unseen instances against random-angle baselines; a failure to beat random angles would falsify the fixed-angle transfer assumption that underlies the QAOA speedups.","tokens_in":43879,"feed_emoji":"⚛️","tokens_out":5942,"duration_ms":56942,"temperature":0.7,"pith_summary":"This review synthesizes the software half of a large national quantum-computing effort from 2020 to 2024 and argues that the pieces now form a functioning vertical stack: quantum algorithms, circuit compilers and qudit decompositions, error-suppression and correction methods, processor emulators, benchmarking tools, and a cloud platform that exposes algorithms and hardware to users. The headline result is that real quantum algorithms—including Grover search, Bernstein-Vazirani, variational quantum eigensolver runs, QAOA-based optimization, and machine-learning classification—were executed on physical processors through that cloud platform. The review also reports that the cloud platform connects more distinct hardware platforms (trapped ions, neutral atoms, superconducting circuits, photonics) than most comparable services. If the claims hold, they show that a coordinated software program can make heterogeneous quantum processors usable as a single remote computing resource, not just as laboratory demonstrations.","feed_headline":"Cloud platform ran quantum algorithms on four processor types","feed_subtitle":"The roadmap delivered algorithms, compilers, emulators, and error suppression, then validated them on real hardware.","key_machinery":"The integrating object is the cloud-based quantum computing platform, which wraps emulators, compilers, benchmarking, and physical processors behind a common interface and is the reason algorithms can be called 'executed' rather than simulated. The main algorithmic workhorse is QAOA with fixed parameters (fpQAOA): universal angles are precomputed once on a training sample of problem instances, then reused without classical optimization, which is what makes the linear-system, differential-equation, and factoring-related algorithms practical on noisy hardware. A second mechanism is qudit compilation, which uses the extra levels of $d$-level systems as ancillas or embedded qubits to cut the number of noisy two-body gates in Toffoli decompositions from $O(N^2)$ to $2N-3$.","core_discovery":"The central claim, stated in the paper's own terms, is that the roadmap delivered a complete, tested software ecosystem rather than isolated tools. The authors treat the execution of quantum algorithms through the cloud-based quantum computing platform as the key achievement: algorithms developed in-house were run on a trapped-ion processor, a neutral-atom processor, and in test mode on additional trapped-ion, superconducting, and photonic processors, with emulators filling in where hardware was not available. Supporting results include the demonstration that fixed-angle QAOA variants solve linear systems, differential equations, and factoring-related optimization without per-instance parameter optimization; variational algorithms reaching chemical accuracy on small molecules; qudit-based compilations that reduce multi-qubit gates to $2N-3$ entangling operations; and a single-ancilla error-correction scheme that improves stored-state fidelity under simulated noise.","pith_inferences":["Editorial inference: the fixed-angle approach, if it generalizes beyond the tested classes, would turn QAOA into a 'compile-once, run-many' pattern, but the current evidence is empirical and the review does not characterize when transfer fails.","Editorial inference: the qudit Toffoli decomposition's claimed $2N-3$ entangling gates suggests that mapping qubit circuits onto qudit processors could become a standard compilation step on any platform with accessible higher levels.","Editorial inference: the cloud platform's multi-architecture access makes direct hardware benchmarking feasible; a natural next test is submitting identical benchmarking circuits through the same interface to each connected processor and publishing the resulting fidelity and quantum-volume measurements."],"forward_implications":["If the fixed-angle transfer hypothesis holds broadly, QAOA-family algorithms can be deployed with essentially no quantum-classical optimization loop, removing a main practical bottleneck of variational approaches.","Qudit compilation lowers the gate count of multi-qubit primitives enough that algorithms like Grover and generalized Toffoli become usable on current noisy hardware; the review reports higher accuracy for the qutrit realization on a trapped-ion processor.","The single-ancilla error-correction scheme lets repetition-code experiments run on devices that lack many qubits, and simulations show fidelity gains for both Markovian and non-Markovian noise models.","Emulators calibrated against real processors reproduce device output with high fidelity once noise parameters are tuned, so they can stand in for hardware during algorithm development.","A single cloud interface can support at least a thousand experimental runs per year across multiple hardware types, making cross-platform algorithm comparison routine."],"supporting_citations":[{"why":"Defines the QAOA family that the fpQAOA variants build on.","marker":"[40]"},{"why":"First statement of the fixed-angle concentration hypothesis for QAOA.","marker":"[45]"},{"why":"Formulates the fixed-angle conjecture for regular Max-Cut graphs.","marker":"[47]"},{"why":"Supplies the training-set heuristic used to compute the universal angles.","marker":"[49]"},{"why":"Provides the baseline emulator for performance comparisons.","marker":"[104]"},{"why":"Describes the qubit-to-qudit transpiler used to run compiled circuits.","marker":"[124]"},{"why":"Presents the single-ancilla stabilizer code whose fidelity gains are reported.","marker":"[35]"},{"why":"Supplies the Cartan-decomposition approach used for fixed-depth Hamiltonian simulation.","marker":"[90]"},{"why":"Used as the external comparison point for the cloud platform's breadth.","marker":"[154]"}],"fun_headline_variants":["Quantum roadmap delivers full software stack, cloud-validated","Cloud quantum platform runs algorithms from complete software suite","Quantum software ecosystem tested on real processors via cloud","Fixed-angle QAOA solves problems without per-instance tuning","Single-ancilla error correction improves stored-state fidelity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported speedups of the fpQAOA-based algorithms stand on the empirical assumption that a single fixed set of angles found on training instances works nearly as well on any unseen instance of the same problem class; the review offers no proof or broad failure analysis for this transfer.","fun_headline_variants_meta":{"raw":{"variants":["Quantum roadmap delivers full software stack, cloud-validated","Cloud quantum platform runs algorithms from complete software suite","Quantum software ecosystem tested on real processors via cloud","Fixed-angle QAOA solves problems without per-instance tuning","Single-ancilla error correction improves stored-state fidelity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00133,"raw_usage":{"total_tokens":5351,"prompt_tokens":828,"completion_tokens":4523,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":444,"completion_tokens_details":{"reasoning_tokens":4447}},"tokens_in":444,"tokens_out":4523,"duration_ms":33873,"temperature":1.0,"reasoning_tokens":4447,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:32:28.947124+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a problem class the review does not cover, such as Max-Cut on graphs with unbounded degree or an ill-conditioned linear system, fix the angles by the same training procedure, then measure the success probability on unseen instances against random-angle baselines; a failure to beat random angles would falsify the fixed-angle transfer assumption that underlies the QAOA speedups.","supporting_citations":[],"review_version":1}