{"id":"5782ebae-bc27-47a3-87bb-4f82f2c7eb95","arxiv_id":"2505.10012","paper_version":1,"verdict":"UNVERDICTED","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Quantum CAE is proposed as a framework that integrates quantum computing into engineering design automation, illustrated through combinatorial optimization case studies and AI-generated quantum circuits.","lead":"A perspective paper proposes 'Quantum CAE,' a framework that applies quantum algorithms to simulation, optimization, and machine learning within engineering design, drawing parallels with computer-aided engineering. It argues that current quantum annealers can already provide Level 3 automation benefits, with AI agents designing quantum circuits to push toward higher automation levels.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'tangible benefits' claim for Quantum CAE rests on case studies with no classical baseline; the load-bearing assumption that quantum annealing outperforms classical optimization in these black-box loops is unverified.","rationale":"The reader's weakest_assumption is exactly the load-bearing concern: the case studies lack classical baselines and time-to-solution metrics, so the claim that quantum annealing integrated into BOCS/FMQA yields design-relevant solutions more efficiently than classical optimization is unverified. My stress-test confirms this and finds no other internal inconsistency that would change the verdict. The paper is a perspective/review, and the verdict UNVERDICTED is appropriate because the central empirical claim depends on external references that are not reproduced. A concrete check—re-running one case study with a classical solver under identical evaluation budgets—would either support or undermine the 'tangible benefits' assertion. Therefore the reader's verdict does not need to change.","tokens_in":10083,"tokens_out":3079,"duration_ms":31964,"concrete_test":"Reproduce the Okada et al. noise-filter optimization (ref. [32], IEEE Access 2023) using the same QUBO formulation and cost function, but replace the quantum annealer with a classical simulated annealing or tabu search solver, using the same number of cost-function evaluations. Compare the best cost achieved and the number of evaluations to first feasible solution. If the classical solver matches or beats the quantum annealer, the paper's claim of 'tangible benefits' from quantum annealing is unsupported by its cited evidence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim (abstract; Section VI) is that Quantum CAE delivers 'tangible benefits' in discrete-variable design problems at current small scales. In Section IV, this is supported only by two case studies, Matsumori et al. [31] and Okada et al. [32], which are said to show that BOCS and FMQA 'efficiently identify' good solutions. No comparison is made to classical solvers, no time-to-solution or wall-clock data are reported, and no optimality gaps are given. Figure 3 shows only that the cost decreases with more searches for the noise filter problem; it does not establish that quantum annealing provides any advantage over, say, simulated annealing or tabu search on the same QUBO. The reader's weakest_assumption correctly identifies this: the claimed benefit rests entirely on self-cited prior work, not reproduced here. If classical heuristics achieve comparable or better solutions with the same evaluation budget, the central claim loses its empirical basis. This is not a critique of the perspective format per se, but the abstract and summary state the benefit as fact, not as a hypothesis, so the missing baseline is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This perspective paper introduces \"Quantum CAE,\" a framework that applies quantum algorithms for simulation, optimization, and machine learning within computer-aided engineering and scientific automation. The author draws parallels between scientific discovery workflows and CAE design automation, argues that current technology enables Level 3 automation under well-defined conditions, and illustrates the framework with two case studies of discrete-variable design optimization using BOCS and FMQA with quantum annealers and Ising solvers (automotive mounting-point placement and noise-filter design). The paper then discusses AI agents that autonomously generate quantum circuits and sketches a roadmap toward higher automation levels. The central empirical claim, stated in the abstract and Section VI, is that Quantum CAE can deliver tangible benefits at current small scales, supported primarily by the two case studies and references to the author's own prior work.","tokens_in":10337,"tokens_out":3489,"duration_ms":35069,"significance":"If the practical-benefit claim were convincingly demonstrated, this paper would constitute a valuable synthesis for near-term quantum computing in industrial design, giving engineers a concrete framework for applying quantum optimization. The conceptual mapping between scientific automation and CAE is clear and potentially useful, and the paper usefully assembles a broad literature on quantum-assisted design, including work by others. However, the manuscript is a perspective that presents no new technical results, and its load-bearing empirical claims rest on a small set of self-cited, non-benchmarked case studies. The acknowledged exponential worst-case overhead for classical data I/O also sits in tension with the paper's feasibility tone. The significance is therefore primarily programmatic rather than evidentiary; the paper could serve as a roadmap if the claims are appropriately reframed as hypotheses and the limitations are more fully acknowledged.","major_comments":[{"comment":"The central claim that BOCS and FMQA 'efficiently identify' good solutions and that Quantum CAE delivers 'tangible benefits even at small scales' (Section VI) is not supported by the evidence presented. The two case studies are described only qualitatively: no comparison to classical optimization baselines (e.g., simulated annealing, tabu search, or classical Bayesian optimization), no time-to-solution or wall-clock data, and no optimality gaps are reported. Figure 3 only shows a cost function decreasing with the number of searches, which any reasonable optimization method would exhibit; it does not demonstrate any quantum-specific advantage. Because the abstract and summary assert the benefit as fact rather than as a hypothesis, this missing baseline is load-bearing for the paper's main claim.","section":"Section IV, case studies [31] and [32], and Figure 3"},{"comment":"The paper acknowledges that encoding classical data into quantum states and extracting information back has exponential computational cost in the worst case, yet the abstract and Section VI claim that current technology enables practical, tangible benefits. The manuscript does not reconcile this tension for the specific discrete-variable problems in the case studies, nor does it explain how the black-box optimization loop avoids or mitigates this overhead. Without such an analysis, the practical-feasibility claim is not internally consistent, and the paper should either quantify the overhead for the presented problems or soften the claim to a proposal that still requires this issue to be addressed.","section":"Section III, paragraph on challenges"},{"comment":"The evidence for Quantum CAE's value is heavily dominated by prior papers authored or coauthored by the current author (refs [26], [31], [32], [46], [49], [53], and [55]). These works are cited as demonstrations, but they are not reproduced, not independently validated, and not benchmarked against classical solvers in this manuscript. For a perspective, citing one's own prior work is acceptable if the claims are framed as promising directions with appropriate caveats; however, the current text states the outcomes as established results (e.g., 'These case studies highlight the effectiveness of quantum annealing as well as Ising solvers in enhancing automation frameworks'). The author should reframe these as preliminary or illustrative and explicitly identify the need for independent, systematic benchmarks.","section":"General, evidence base"}],"minor_comments":[{"comment":"The name is typeset as 'Erd˝ os,' which is likely a character-encoding error; it should be 'Erdős'.","section":"Section I, paragraph on Paul Erdős"},{"comment":"Reference [12] contains an anomalous author entry 'T. Google, and A. I. Language,' which appears to be a corrupted or misformatted citation and should be corrected.","section":"References, ref [12]"},{"comment":"The name 'Al´an Aspuru-Guzuk' should be spelled 'Alán Aspuru-Guzik'.","section":"Acknowledgments"},{"comment":"The y-axis is labeled 'y' while the caption refers to the 'cost function'; the notation should be unified, and the axes should be described consistently.","section":"Figure 3 caption"},{"comment":"The sentence 'These methods rely on discrete design variables and evaluation approaches, such as simulations, to assess product characteristics' is redundant with the preceding discussion and could be removed or rewritten for clarity.","section":"Section IV, opening paragraph"},{"comment":"The six-level JST taxonomy is referenced but not fully defined; readers unfamiliar with the original report may not understand what distinguishes Level 3 from Levels 4 and 5. A brief definition or a pointer to the original report would improve accessibility.","section":"Section II, JST automation levels"}],"recommendation":"major_revision","confidential_remarks":"The paper is a perspective that makes fairly strong practical-feasibility claims on the basis of a small number of self-cited case studies. The self-citation pattern is notable: at least seven of the key supporting references are authored or coauthored by the author, and they are used as evidence without independent validation. I recommend that the editor require the author to substantially temper the central claim to a hypothesis or provide independent benchmarks, whichever is more appropriate for the venue. The paper might be better aligned with a venue that explicitly welcomes forward-looking perspective pieces; its current framing risks overstating the established evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick read for you. This is a perspective, not a research paper. The only genuinely new thing is the name 'Quantum CAE' and the mapping of JST/SAE automation levels onto engineering design workflows. That framing has some use for industrial R&D groups deciding where to try quantum optimization, and the paper is clearly written and honest about the worst-case exponential cost of encoding classical data into quantum states. It does not, however, introduce a single new result, and the central claim that Quantum CAE delivers 'tangible benefits even at small scales' is not actually supported in the text.\n\nThe support for that claim is the two case studies in Section IV: Matsumori et al. and Okada et al., both co-authored by this author. The paper tells us BOCS and FMQA 'efficiently derived' good solutions, but there is no comparison to a classical solver, no wall-clock data, no optimality gap, and no reproducibility statement. Figure 3 only shows a cost curve decreasing with number of searches, which any reasonable heuristic does. So the load-bearing evidence is self-cited and unbenchmarked. That is a real soft spot, not a manufactured one.\n\nThe paper's other weakness is the citation pattern. Of the case studies and algorithm demonstrations that matter, a large fraction are the author's own prior work ([26], [31], [32], [46], [49], [53], [55]). None are independently reproduced in the paper. For a perspective that is not necessarily disqualifying, but it makes the 'tangible benefits' wording feel like advocacy rather than analysis.\n\nWhat the paper does well: it writes the abstraction level correctly. The discussion of simulation/optimization/ML as the three legs, and the link to black-box optimization with QUBO, is accurate and accessible. It also gives a realistic picture of the current state: small problems, Ising solvers, and the importance of conventional algorithm development when quantum hardware is small. If your group is thinking about whether to try quantum annealers in a design loop, this is a reasonable starting bibliography.\n\nVerdict: this is a perspective that deserves a serious referee, but the referee should require the benefits claim to be hedged to 'potential' and should ask for at least one sentence acknowledging that no classical baseline comparison is made. As is, I wouldn't cite it for any technical claim, but I might cite it for the Quantum CAE framing if I wrote an industrial roadmap. Reading group: maybe, if you want a short discussion of how quantum computing gets positioned in engineering.","headline":"A clearly written perspective whose only genuinely new contribution is the label 'Quantum CAE'; the central 'tangible benefits' claim rests on self-cited case studies with no classical baseline.","tokens_in":10798,"tokens_out":2252,"would_cite":false,"duration_ms":20969,"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":"This perspective argues that Quantum CAE—using quantum algorithms for simulation, optimization, and machine learning inside engineering design—can deliver tangible benefits in discrete-variable design problems at current small scales.","keywords":["Quantum CAE","quantum annealing","black-box optimization","QUBO","scientific automation","design automation","digital scientists","AI4Science"],"falsifier":"Run BOCS and FMQA on the same electronic-board or noise-filter design problems with the quantum annealer replaced by a classical optimizer such as simulated annealing, keeping the surrogate model and evaluation budget identical; if the classical version matches or beats the quantum-in-the-loop version on solution quality and wall-clock time, the paper's claim of tangible near-term benefit is not supported.","tokens_in":9884,"feed_emoji":"⚛️","tokens_out":6137,"duration_ms":58807,"temperature":0.7,"pith_summary":"The paper argues that scientific discovery and engineering design run on the same three-step loop—generate a hypothesis, test it by experiment or simulation, fold the findings into knowledge—and that quantum computing can accelerate each step. It introduces Quantum CAE as the umbrella term for this insertion, with quantum optimization as the most mature ingredient. The central practical claim is that current small-scale quantum annealers and Ising solvers, embedded in black-box optimization methods, already produce useful solutions to discrete-variable design problems that are NP-hard. The paper presents two industrial-style case studies as evidence and maps this capability onto Level 3 scientific automation, meaning full autonomy under well-defined conditions with humans supervising.","feed_headline":"Small-scale quantum annealers already earn their keep in design","feed_subtitle":"A framework called Quantum CAE plugs quantum optimization into engineering loops, claiming Level 3 automation is reachable now.","key_machinery":"The load-bearing machinery is the black-box optimization cycle woven into CAE: an optimizer proposes candidate designs (hypothesis generation), a simulator evaluates them (experiment), and a machine-learning model absorbs the data and improves the next proposal (knowledge integration). Discrete design variables are encoded as QUBO problems, and quantum annealing or Ising solvers perform the optimization step. BOCS (Bayesian Optimization of Combinatorial Structures, a posterior-sampling approach) and FMQA (Factorization Machine Quantum Annealing, a point-estimate approach) are the two instantiations that carry the case studies. The framework-level object, Quantum CAE, names the full integration of quantum simulation, quantum machine learning, and quantum optimization, with the long-term goal of exchanging quantum states directly between tasks.","core_discovery":"The central claim is that Quantum CAE can deliver tangible benefits at small scale for discrete-variable engineering design problems by treating optimization as the hypothesis-generation step in an automated design loop. The paper identifies quantum annealing and Ising solvers as practical tools for this step, combined with learned surrogate models such as factorization machines or posterior-sampling models. Two case studies—mounting-point placement on an electronic control board and printed-circuit-pattern design for a noise filter—are offered as demonstrations that the approach finds feasible, near-optimal designs within limited iterations, with the noise-filter solution resembling a topology-optimization result. The paper further claims that this level of integration corresponds to Level 3 automation, and that the remaining path to Levels 4 and 5 requires teams of specialized AI agents, including agents that design quantum circuits for QUBO problems. On the simulation side, it argues quantum algorithms for equations like radiative transfer are emerging, but the main near-term evidence is optimization.","pith_inferences":["The paper leaves the size of the benefit unquantified: its two case studies do not report classical baselines or time-to-solution, so the fair reading is that the claim is feasibility, not demonstrated superiority; a direct comparison against classical heuristics on the same problems would isolate the quantum contribution.","A testable extension would run FMQA and BOCS on the same design problems with the quantum step replaced by simulated annealing, holding the surrogate model fixed; matching performance would suggest the benefit comes from the black-box loop rather than the quantum sampler.","The Quantum CAE framing implies that quantum advantage may show up first as better end-to-end design outcomes rather than as a stand-alone algorithmic speedup, which is a different benchmark from the one usually used in quantum computing studies.","The same loop could transfer to other discrete domains, such as pharmaceutical candidate selection or materials design, wherever a cheap simulator can evaluate a QUBO-encoded proposal."],"forward_implications":["Discrete-variable design problems that are NP-hard can in principle be tackled in current engineering practice with quantum annealers or Ising solvers inside a black-box loop, without waiting for fault-tolerant hardware.","Reaching Level 3 scientific automation—full autonomy under well-defined conditions—is presented as feasible now for computationally closed fields such as structural and circuit design.","If larger annealers arrive, the bottleneck shifts from optimization to building the prediction model, so classical algorithm development remains a priority in the short term.","Higher automation levels will depend on specialized AI agents that can design and validate quantum circuits, such as generators for QUBO-solving circuits, rather than on a single general-purpose algorithm.","Successful Quantum CAE would reduce prototype and development cycles by replacing physical experiments with quantum-assisted simulation and automated design refinement."],"supporting_citations":[{"why":"Introduces BOCS, the probabilistic modeling method used as the black-box optimization approach in the case studies.","marker":"[29]"},{"why":"Introduces FMQA, pairing factorization machines with quantum annealing, the second optimization method tested.","marker":"[30]"},{"why":"Reports the electronic control board mounting-point optimization case study used as evidence that Quantum CAE delivers design-relevant solutions.","marker":"[31]"},{"why":"Reports the noise-filter design optimization case study and the iterations-to-solution data used to show feasible patterns emerge.","marker":"[32]"},{"why":"Supplies the topology-optimization result that the paper compares visually with the black-box optimization solution.","marker":"[33]"},{"why":"Provides a quantum circuit algorithm for radiative transfer, an example of the simulation-side ingredient Quantum CAE requires.","marker":"[49]"},{"why":"Describes an AI agent that generates quantum circuits for QUBO problems, supporting the claim that specialized agents will drive higher automation levels.","marker":"[55]"}],"fun_headline_variants":["Quantum annealers plug into design loops at small scale","Quantum CAE: optimization as hypothesis generation","Small-scale quantum optimization earns Level 3 automation","Quantum CAE shows small-scale annealers deliver in design"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claimed practical benefit rests on the assumption that quantum annealing embedded in black-box optimization finds design-relevant solutions at least as efficiently as classical optimization; the cited case studies do not include classical baseline comparisons or time-to-solution metrics, so this is asserted rather than demonstrated here.","fun_headline_variants_meta":{"raw":{"variants":["Quantum annealers plug into design loops at small scale","Quantum CAE: optimization as hypothesis generation","Small-scale quantum optimization earns Level 3 automation","Quantum CAE shows small-scale annealers deliver in design"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000696,"raw_usage":{"total_tokens":3099,"prompt_tokens":852,"completion_tokens":2247,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":468,"completion_tokens_details":{"reasoning_tokens":2186}},"tokens_in":468,"tokens_out":2247,"duration_ms":15462,"temperature":1.0,"reasoning_tokens":2186,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:18:22.522389+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run BOCS and FMQA on the same electronic-board or noise-filter design problems with the quantum annealer replaced by a classical optimizer such as simulated annealing, keeping the surrogate model and evaluation budget identical; if the classical version matches or beats the quantum-in-the-loop version on solution quality and wall-clock time, the paper's claim of tangible near-term benefit is not supported.","supporting_citations":[{"cited_title":"Kadowaki, Enhancing quantum annealing in digi- tal–analog quantum computing, APL Quantum 1, 26101 (2024)","cited_arxiv_id":null,"evidence_quote":"Introduces BOCS, the probabilistic modeling method used as the black-box optimization approach in the case studies."},{"cited_title":"Gilliam, S","cited_arxiv_id":null,"evidence_quote":"Introduces FMQA, pairing factorization machines with quantum annealing, the second optimization method tested."},{"cited_title":"Marsh and J","cited_arxiv_id":null,"evidence_quote":"Reports the electronic control board mounting-point optimization case study used as evidence that Quantum CAE delivers design-relevant solutions."},{"cited_title":"Baptista and M","cited_arxiv_id":null,"evidence_quote":"Reports the noise-filter design optimization case study and the iterations-to-solution data used to show feasible patterns emerge."},{"cited_title":"Kadowaki and M","cited_arxiv_id":null,"evidence_quote":"Provides a quantum circuit algorithm for radiative transfer, an example of the simulation-side ingredient Quantum CAE requires."},{"cited_title":"Layden, G","cited_arxiv_id":null,"evidence_quote":"Describes an AI agent that generates quantum circuits for QUBO problems, supporting the claim that specialized agents will drive higher automation levels."}],"review_version":1}