{"id":"715ec88c-344e-4920-81e1-290f7e9bdc08","arxiv_id":"2511.22090","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"high","formal_verification":"none","parameter_count":6,"one_line_summary":"Quantum Monte Carlo mean estimation is plugged into Bayesian optimization for fuselage gap minimization, and simulated comparisons claim query savings over classical BO.","lead":"A team applies a quantum-Monte-Carlo-based Bayesian optimization framework to choosing actuator forces that reduce gaps when joining aircraft fuselage sections, claiming far fewer samples than classical Bayesian optimization. The potential payoff is cheaper, faster quality control in aerospace assembly, but the quantum speedup is assumed through an oracle rather than demonstrated on hardware.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim of quantum advantage rests on an assumed oracle O (Eq. 6); no construction or cost model is given, so the query-count comparison in Sec. 4 is unsupported.","rationale":"The reader's weakest assumption—that the quantum oracle O can be constructed and that a query to O is a legitimate substitute for one classical evaluation—is exactly the load-bearing point. The paper provides no circuit, state-preparation, or gate-level analysis for O, and its own conclusion admits that all experiments were classical simulations. This is not a disagreement with quantum Monte Carlo theory; it is an unsupported bridge between a theoretical oracle model and the experimental comparison. The reader's verdict of REJECT with moderate confidence is appropriate, and my concern does not alter that verdict. No additional independent concern (e.g., about the regret bound adaptation) is needed to reach the same conclusion.","tokens_in":14035,"tokens_out":2127,"duration_ms":23483,"concrete_test":"Specify a complete quantum circuit oracle O for the linear surrogate model in Sec. 4.2.1 (including state preparation of the Gaussian noise and coherent encoding of y(ω)), with gate counts and qubit counts. Then compute the total quantum cost per QMC estimate (oracle calls × circuit depth) and compare it with the classical BO query count in Fig. 5 for the same precision ε_s. If the total quantum cost exceeds the classical BO query count, the claimed query advantage disappears; if the oracle relies on QRAM or HHL, include those costs in the comparison.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—fewer queries to achieve optimal solutions—depends entirely on the quantum oracle O in Eq. (6), which coherently encodes the distribution of fuselage shapes and their deviations. Lemma 1 (Sec. 3.3.3) gives the QMC query complexity relative to O, but the paper never constructs O for the FEA/surrogate environment described in Sec. 4.2.1, nor analyzes the cost of preparing it. The experiments in Sec. 4 are classical simulations that count theoretical QMC queries to O as if each query were equivalent to one classical FEA/surrogate evaluation. This equivalence is not justified: encoding P(ω) and y(ω) for the linear surrogate model would require either coherent access to the displacement matrix U and the noise distribution (e.g., via QRAM or quantum linear algebra), whose end-to-end cost is not quantified, or a circuit-level implementation with unknown constant factors C1, C2. The paper's own conclusion admits all experiments were performed on classical simulators and deployment on real quantum hardware is future work. Without a concrete oracle construction and a complexity model that includes oracle-preparation cost, the query comparison in Figs. 4–6 is not a fair accounting of total computational cost, and the claimed quantum advantage in sample efficiency is not demonstrated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a Quantum Bayesian Optimization (QBO) framework for minimizing dimensional gaps in fuselage assembly. Force-controlled actuators are optimized under noisy measurements; a Gaussian-process surrogate with random Fourier features and a UCB acquisition function selects force combinations, and quantum Monte Carlo (QMC) is used to estimate the noisy mean response, assuming a quantum oracle O that encodes the distribution of possible final shapes and deviations. The authors cite prior regret bounds for QBO and present comparisons against classical BO in simulated fuselage assembly cases, reporting faster convergence and lower MAE for QBO under two noise levels. The paper's advertised central claim is improved sample efficiency from quantum mean estimation.","tokens_in":14417,"tokens_out":4294,"duration_ms":41975,"significance":"If substantiated, the claimed sample-efficiency improvement would be practically valuable because FEA and physical measurements in fuselage assembly are expensive, and a demonstration of quantum advantage in a manufacturing design problem would be a notable first. The paper is clear about the application, problem formulation, and noise sources, and it usefully imports the QBO regret framework from the quantum bandit literature. However, the central evidence is currently not convincing: the quantum advantage rests entirely on an assumed oracle that is never constructed or costed, and the experimental query accounting is ambiguous. The manuscript's strengths are in the application framing and in the integration of existing QBO theory, not in a validated implementation or a fair query comparison.","major_comments":[{"comment":"The quantum oracle O is assumed but never constructed. For the linear surrogate model Y = F_D·U + ε used in Sec. 4.2.1, encoding P(ω) and y(ω) coherently would require a concrete data-loading or quantum-linear-algebra procedure; no circuit, QRAM model, or end-to-end complexity estimate is provided. Lemma 1 gives query complexity only relative to O, so all QMC query counts are counts of an abstract oracle whose relationship to a classical surrogate/FEA evaluation is not established. This is load-bearing because the abstract and Sec. 4 claims of 'fewer queries' compare these oracle queries with classical environment evaluations.","section":"Eq. (6), Sec. 3.3.2, Sec. 4.2.1"},{"comment":"Query/iteration accounting is ambiguous. Step 6 of Algorithm 1 runs QMC(O, ηϵ_s, δ), which by Lemma 1 consumes N_ϵ = O(1/ϵ·log(1/δ)) (or the bounded-variance version) queries to O per stage. However, Sec. 1 states that samples, queries, and iterations are interchangeable, and all experiments plot cumulative regret and running-minimum MAE against 'iterations' (Figs. 4-7). If one iteration is one QMC run, the total number of oracle queries is Σ_s N_s, which can greatly exceed T; if the x-axis is oracle queries, then the number of evaluated force combinations is much smaller than T and the cumulative regret defined in Eq. (7) is not the quantity plotted. The conclusion that QBO achieves optimal solutions with fewer queries is therefore not supported by the experiments as reported.","section":"Algorithm 1, Sec. 3.3.4, Sec. 1"},{"comment":"The regret bound R_T = O(poly log T) is imported from Dai et al. (2023) and is not derived in this paper. More importantly, it is not shown that the bound applies under the oracle-query accounting used here. The stage-wise oracle query cost in Algorithm 1 is not reconciled with the T appearing in Eq. (7), so the theoretical result is not connected to the experimental iterations or to the claimed query counts.","section":"Lemma 2, Sec. 3.3.4"},{"comment":"The conclusion admits that all experiments were performed on classical simulators, with deployment on real quantum hardware left to future work. There is no implementation of the oracle or of QMC on quantum hardware. The numerical comparisons in Figs. 4-6 should therefore be described as classical simulations of an assumed quantum subroutine, not as experimental demonstrations of quantum advantage. This does not invalidate the idea, but it does not support the abstract's claim of demonstrated advantage.","section":"Sec. 5"}],"minor_comments":[{"comment":"The title 'Quantum Safe-Set Bayesian Optimization' does not match the manuscript body, which describes only 'Quantum Bayesian Optimization' and contains no safe-set concept. This should be reconciled.","section":"Title/Sec. 1"},{"comment":"Typo: 'O(ploy logT)' should be 'O(poly log T)'.","section":"Sec. 3.3.4"},{"comment":"The discussion of Fig. 5(a) says larger confidence intervals for QBO imply a 'wider search space' and 'more different optimal solutions'; this reasoning is not quantified. If multiple optima are claimed, the number or diversity of distinct solutions should be reported.","section":"Sec. 5, Sec. 4.3.2"},{"comment":"The reference 'Lutz et al. (2024)' has an incomplete author string ('J. C. and'). Some equations, e.g., Eq. (3), use vector u without a subscript for measurement point i, which is notationally inconsistent with the surrounding definitions.","section":"References"},{"comment":"The box plot description mentions an 'orange median line' while the figure legend uses QBO (orange) and classic BO (blue); the text should specify which color corresponds to which method in the box plots.","section":"Fig. 6"}],"recommendation":"reject","confidential_remarks":"The manuscript's central claim of a quantum advantage is not supported because the quantum oracle is assumed rather than constructed or costed, and the query/iteration accounting makes the experimental comparisons difficult to interpret. These are load-bearing issues rather than local presentation problems. The application is timely and the use of existing QBO theory is appropriate, but the current evidence is essentially a classical simulation that assumes the conclusion. A revision could potentially address these issues with a concrete oracle construction or at least a cost model and corrected query accounting, but that would be substantial new work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful part of this paper is the application: it takes the QBO machinery from Dai et al. (2023), adds the weighted GP from Deng et al. (2022), and runs a case study on actuator force optimization for fuselage assembly. That specific demonstration is new, and the manufacturing problem is well motivated. The writing is clear, the related work is relevant, and the paper does not oversell the fact that all experiments are classical simulations—the conclusion admits this. As an engineering case-study write-up, it is a reasonable contribution.\n\nThe soft spot is the one the stress-test flagged. The entire sample-efficiency advantage rests on the quantum oracle O in Eq. (6). The paper defines it abstractly, imports Lemma 1 from Montanaro, and then counts QMC queries as if each query were equivalent to one classical FEA/surrogate evaluation. But no construction of O for the linear surrogate is given, and the cost of preparing O is not analyzed. That makes the query-count comparisons in Figs. 4–6 an arithmetic consequence of a favorable assumption rather than a demonstrated speedup. The gap between claiming quantum advantage and claiming a framework that could exhibit that advantage once an oracle is built is not minor; it is the main claim of the paper.\n\nThere is also a more prosaic accounting problem. Section 2 states that samples, queries, and iterations are used interchangeably. Algorithm 1 consumes N_epsilon queries per stage, while the experiments report iterations. So the reader cannot tell how many actual oracle calls are being compared. That ambiguity should have been caught before submission.\n\nThe theoretical regret bounds are imported from prior work, which is fine—they cite properly—but they inherit all the assumptions of that prior work, including the oracle. The free parameters are standard for BO, and the RFF approximation is standard. No fabricated entities beyond the oracle itself.\n\nWho gets value from this paper? Someone working on quantum Bayesian optimization who wants a concrete application domain and a clear write-up of how QBO would be plugged in. For a manufacturing audience, the takeaway should be framed as a proposal, not a demonstrated advantage.\n\nFor peer review: this deserves a serious referee, not a desk reject. The application is real, the writing is honest, and the gap between the oracle assumption and the empirical claims is exactly what a good reviewer should force the authors to address. But the central claim, as stated, is not supported. I would expect major revision, with the authors either constructing the oracle for the linear surrogate or substantially softening the abstract and results.","headline":"A coherent application of QBO to fuselage assembly, but the central claim of fewer queries is unsupported because the quantum oracle is assumed, not constructed.","tokens_in":14830,"tokens_out":1454,"would_cite":false,"duration_ms":16149,"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 quantum version of Monte Carlo estimation lets Bayesian optimization find optimal fuselage actuator forces with far fewer simulator queries, the paper claims.","keywords":["Quantum Bayesian Optimization","Fuselage assembly","Sample efficiency","Quantum Monte Carlo","Bayesian optimization","Shape control","Gaussian processes","Manufacturing optimization"],"falsifier":"Implement the eight-actuator fuselage experiment replacing the quantum Monte Carlo subroutine with a classical Monte Carlo estimator that uses the same number of queries, keeping all other settings identical; if the classical version achieves the same or better final MAE and cumulative regret, the quantum speedup claim is falsified. Alternatively, provide an explicit quantum circuit for the oracle and measure the actual number of physical gate operations required to encode one surrogate evaluation, and compare that to the cost of one classical FEA call.","tokens_in":13963,"feed_emoji":"🔧","tokens_out":2690,"duration_ms":27773,"temperature":0.7,"pith_summary":"This paper tries to establish that replacing classical Monte Carlo mean estimation with quantum Monte Carlo inside a Bayesian optimization loop dramatically reduces the number of simulator evaluations needed to find optimal actuator forces in aircraft fuselage assembly. The motivation is that fuselage shape adjustment is expensive to evaluate and noisy in practice, and the paper argues that quantum mean estimation achieves the same accuracy with quadratically fewer queries. If true, this would make shape-control optimization far more sample-efficient, reducing the cost of quality improvement in manufacturing. The paper reports experiments, all run on a classical simulator, showing lower cumulative regret and lower final dimensional error for the quantum approach compared to classical Bayesian optimization under two noise levels.","feed_headline":"Quantum Monte Carlo finds fuselage fixes with fewer queries","feed_subtitle":"By encoding shape deviations into a quantum oracle, the method matches classical fit quality at a fraction of the simulator budget.","key_machinery":"The load-bearing component is the quantum oracle O defined by O:|0> -> sum_omega sqrt(P(omega)) |omega>|y(omega)>, which coherently encodes the probability distribution over possible fuselage shapes and their associated deviations for a given actuator force setting. Quantum Monte Carlo (QMC), the subroutine that estimates the mean of this distribution, provides the theoretical quadratic speedup in sample complexity. The surrounding machinery is a weighted Gaussian Process surrogate with random Fourier features and an Upper Confidence Bound acquisition function, which guides the selection of force combinations to evaluate using the oracle.","core_discovery":"The central claim is that a Quantum Bayesian Optimization (QBO) framework, which uses a quantum oracle to encode the distribution of possible fuselage shape deviations and quantum Monte Carlo to estimate the mean response, reaches optimal actuator force configurations with significantly fewer queries than classical Bayesian optimization. The theoretical basis is that QMC estimates a distribution mean to accuracy epsilon with O(1/epsilon) oracle queries, whereas classical Monte Carlo needs O(1/epsilon^2); the paper further invokes a regret bound of O(poly log T) for QBO versus the classical lower bound of Omega(sqrt(T)). In simulations of two-actuator and eight-actuator fuselage shape control","pith_inferences":["The paper's query-count comparison implicitly treats each quantum oracle query as the same cost as one classical simulator evaluation, but no construction or circuit depth for the oracle is provided, so the wall-clock cost of realizing the quantum advantage remains untested.","The claimed speedup is purely about sample complexity; if the oracle requires substantial quantum resources to encode an FEA or surrogate model, the practical benefit could vanish, especially on near-term quantum hardware.","The method could be tested on other expensive noisy black-box problems beyond fuselage assembly, such as design of composite structures or process parameter tuning, to see whether the sample-efficiency advantage persists.","A direct empirical falsification would be to rerun the same experiments with classical Monte Carlo mean estimation at equal query counts; if performance matches QBO, the quantum oracle is not providing the claimed benefit."],"forward_implications":["If the central claim holds, fuselage assembly quality control can achieve the same dimensional accuracy with substantially fewer expensive simulator or physical evaluations, making broader design-space exploration feasible.","The theoretical regret bound suggests the approach could scale to higher-dimensional actuator placement problems where classical BO becomes prohibitively sample-hungry.","The method's reliance on a surrogate linear model (rather than full FEA) means it can be paired with any fast approximate simulator, potentially extending to other noisy manufacturing optimization tasks.","The paper's reported convergence within a few hundred to a few thousand queries, if reproducible, would enable near-real-time optimization for each new fuselage initial condition.","The future direction of incorporating stress constraints such as the Tsai-Wu failure criterion would make the optimized force settings physically realizable, strengthening the practical claim."],"fun_headline_variants":["Quantum optimization squeezes fuselage assembly errors","Fewer queries, tighter fuselage seams via quantum BO","Quantum Bayesian method cuts simulator calls for fuselage fit","Quantum oracle tightens fuselage fit with fewer samples","Quantum algorithm speeds precise fuselage assembly"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The central claim collapses if a quantum oracle that encodes the distribution of fuselage deviations for a given force setting cannot be constructed in practice at a cost comparable to one classical simulator evaluation, since all the query-count savings rest on the existence and validity of that oracle.","fun_headline_variants_meta":{"raw":{"variants":["Quantum optimization squeezes fuselage assembly errors","Fewer queries, tighter fuselage seams via quantum BO","Quantum Bayesian method cuts simulator calls for fuselage fit","Quantum oracle tightens fuselage fit with fewer samples","Quantum algorithm speeds precise fuselage assembly"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000348,"raw_usage":{"total_tokens":1763,"prompt_tokens":790,"completion_tokens":973,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":534,"completion_tokens_details":{"reasoning_tokens":901}},"tokens_in":534,"tokens_out":973,"duration_ms":6861,"temperature":1.0,"reasoning_tokens":901,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T19:50:40.572244+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Implement the eight-actuator fuselage experiment replacing the quantum Monte Carlo subroutine with a classical Monte Carlo estimator that uses the same number of queries, keeping all other settings identical; if the classical version achieves the same or better final MAE and cumulative regret, the quantum speedup claim is falsified. Alternatively, provide an explicit quantum circuit for the oracle and measure the actual number of physical gate operations required to encode one surrogate evaluation, and compare that to the cost of one classical FEA call.","supporting_citations":[],"review_version":1}