{"id":"198ae413-a52d-4406-8b5c-9534d5785567","arxiv_id":"2501.11197","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":3.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"A D-Wave hybrid solver is applied to post-disaster road restoration on a 24-zone Sioux Falls network, with claimed speed and equity advantages that the paper's own equations do not support.","lead":"This paper applies D-Wave's hybrid quantum solver to plan which damaged roads to restore after a disaster, aiming to prioritize low-income neighborhoods. The authors report roughly 8.7-second runtimes versus over 600 seconds for a genetic algorithm, but the equity measure in their equations is a constant and cannot drive the prioritization they claim.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The equity term E in Eq. 7 is constant in the restoration variables C^1_a, so the stated objective cannot produce the µ-dependent equity allocations in Fig. 5b; the central equity claim is not derived.","rationale":"The reader's weakest_assumption is the same one I would flag: Eq. 7 is a Gini index over fixed zone incomes. The paper itself states that 'E represents inequity by utilizing the Gini coefficient, which considers income levels,' but it never ties E to the restoration variables. The later quadratic representation in Eqs. 21-22 introduces an undefined \\bar{W} and still omits C^1_a. Consequently, the µ-sweep in Fig. 5b and the entire Section IV-A discussion about shifting priorities cannot follow from the proposed objective. I considered whether another part of the model could carry the equity burden: the node categorization in Fig. 3b merely labels nodes, and no income-dependent coefficient appears in D, in the constraints, or in the budget. The GA fitness in Eq. 34 also includes E_j, but since E_j is constant across individuals, the GA likewise cannot select for equity. Thus the central novel contribution, integrating equity into quantum-based restoration, is not established. The runtime observation in Table IV is plausible, but it does not demonstrate quantum advantage or validate the framework, and without code or data it cannot be independently checked. No formal verification or reproducible artifact is supplied. The rejection rests on this internal inconsistency rather than on any contested benchmark or consensus disagreement. I would not escalate to accusations; the concrete check would settle whether the submitted objective is actually what was optimized.","tokens_in":19140,"tokens_out":3080,"duration_ms":31302,"concrete_test":"Run the stated optimization exactly as in Eq. 30 for the Sioux Falls instance at budget 300 with µ = 0.0 and µ = 1.0, using the same feasible set, link data, and D term. Since E is constant, any correct solver must return the same C^1_a vector for both µ values. Then compute restored capacity per income group from those two solutions; if the Figure 5b curves differ, the implementation did not optimize Eq. 30, and the equity mechanism in the paper is not the one described. As a secondary check, re-derive Eq. 21 from Eq. 7: replacing |I_r - I_s| by a quadratic changes the equity index and still leaves E with no dependence on any restoration variable.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that Q-RESTORE prioritizes low-income communities through equity-aware optimization. That requires the equity term in the minimization objective (Eq. 3) to depend on the restoration decisions C^1_a. As written, E = 1/(2N^2 \\bar{I}) \\sum_{r,s} |I_r - I_s| (Eq. 7), and the later 'quadratic equivalent' E = 1/(2N^2 \\bar{W} \\bar{I}) \\sum_{r,s} u_{r,s} with u_{r,s} = (I_r - I_s)^2 (Eqs. 21-22), involves only fixed zone incomes. The CQM Hamiltonian H(x) in Eq. 30 therefore contains (1-µ) times a constant. For a fixed budget, the feasible set is independent of µ, so the argmin over C^1_a is identical for all µ; only the objective value shifts. Figure 5b, which shows restored capacities by income group changing with µ from 0.0 to 1.0, is incompatible with the stated model. The low-income priority reported in Fig. 5a may reflect network topology or the D term rather than any equity mechanism, unless an unstated income-dependent term was actually optimized. This is an internal inconsistency in the central construction, not merely a disagreement with the field's consensus. Missing code and data, and the unexplained µ = 0.2 versus µ = 0.5 in the Table III caption, compound the problem, but the constant-equity issue is decisive.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Q-RESTORE proposes a hybrid quantum-classical optimization framework for post-disaster transportation network restoration, using a bi-objective objective that weights a recovery deficiency index D against a Gini-based equity measure E. The model is tested on a 24-zone Sioux Falls network with up to 25 damaged links, comparing D-Wave's hybrid solver with a genetic algorithm across budget levels. The paper claims that the quantum solver runs in about 8.7 seconds versus over 600 seconds for the GA, and that the equity term causes restoration to prioritize low-income neighborhoods, with low-income communities served first and higher-income areas receiving more capacity only as budgets grow.","tokens_in":19414,"tokens_out":5900,"duration_ms":57840,"significance":"If the central claims were correct, the paper would contribute a useful demonstration of hybrid quantum solvers for a socially relevant infrastructure restoration problem. The manuscript deserves credit for stating its model explicitly and for using literature-based BPR constants and stated income thresholds rather than fitting parameters to a desired outcome; the runtime comparison is also a concrete, falsifiable claim. However, the equity mechanism that motivates the entire study is not present in the model as written: the equity term is constant in the restoration variables, so it cannot produce the reported mu-dependent allocations. The significance of the paper therefore depends on a reformulation and a complete rerun of the experiments, rather than on local corrections.","major_comments":[{"comment":"The equity term E is constant in the decision variables. Eq. (7) defines E = 1/(2N^2 \\bar I) \\sum_{r,s} |I_r - I_s|, and Eq. (21) replaces the absolute value with u_{r,s} = (I_r - I_s)^2; neither expression contains C_a^1. Substituting into Eq. (3) gives R = \\mu D + (1-\\mu) E_0 with E_0 a fixed number for the given zoning, so the term (1-\\mu)E_0 shifts the objective value but cannot change the argmin over C_a^1 at a fixed budget B, because neither the objective coefficient of C_a^1 nor the feasible set depends on \\mu. Figure 5b, which reports different restored capacities per income group as \\mu goes from 0.0 to 1.0, is therefore incompatible with the stated model. The low-income prioritization shown in Figure 5a, if observed, must come from the mobility term D or from network topology, not from the equity term as defined. This invalidates the paper's central claim that the framework targets the connectivity needs of different income communities through an equity-aware objective.","section":"Section III-A, Eqs. (7) and (21); Figure 5b"},{"comment":"The reported experimental setting is internally inconsistent. The text states that \"a single value of \\mu = 0.2 is used in the resilience measure\" while Table III is captioned \"RECOVERY CAPACITIES ... FOR \\mu = 0.5 ACROSS VARIOUS BUDGET LEVELS,\" and Figure 5b uses \\mu = 0.0, 0.25, 0.75, and 1.0. The reader cannot determine which objective was actually optimized for the headline results, and the discrepancy compounds the issue raised about the equity term.","section":"Section IV-A and Table III caption"},{"comment":"The \"Maximum recovery capacity\" entries in Table III do not match the capacities of the same link numbers in Table II. For example, link 7 is listed with maximum recovery 6.81 instead of 46.81, link 10 with 9.82 instead of 9.04, link 43 with 7.02 instead of 27.02, and link 45 with 4.42 instead of 9.64 (the value 4.42 appears in Table II for link 57). If these are post-disaster residual capacities, the paper does not state how they were computed; if they are not, several of the numerical results, including the equity percentages and the sum totals, are based on data other than the stated network parameters.","section":"Tables II and III"},{"comment":"The experiments are not reproducible from the manuscript: no code, data files, D-Wave solver parameters (such as \\lambda_1, \\lambda_2, number of reads, or time limit), or random seeds are provided. This matters because both the hybrid solver and the GA are stochastic, and the central quantitative claims (restoration values, runtime, and equity percentages) depend on these choices. The paper should at minimum report the full set of hyperparameters and, ideally, make the code and data available.","section":"Section IV and Section III-C"}],"minor_comments":[{"comment":"The penalty formula is written as \\rho = (c_a - B) \\times \\rho when the restoration cost exceeds the budget, which is self-referential and dimensionally inconsistent; a distinct symbol and a max(0,\\cdot) or absolute-value expression are needed.","section":"Section III-D, Eq. (35)"},{"comment":"The budget is defined in units of veh/h and is then treated as a capacity sum in Eq. (18); the equivalence between monetary budget and capacity units is stated only in prose and should be made precise in the model.","section":"Section III-B and Table I"},{"comment":"The symbol p is used both for the crossover point and for a random probability in the mutation operator, and the index i is used for individuals while appearing elsewhere in quantum-mechanical contexts; the notation should be disambiguated.","section":"Section III-D, Eqs. (37)-(39)"},{"comment":"The figure captions and text refer to low-, average-, and high-income nodes, but the figures do not identify node numbers or link numbers, making it difficult to verify the claimed income-based allocation pattern from the figures alone.","section":"Figures 4 and 5"}],"recommendation":"reject","confidential_remarks":"The manuscript appears to be an early preprint rather than a journal-ready submission. The central equity claim is contradicted by the paper's own equations, the experimental parameters are described inconsistently, and the reported tables contain mismatches with the stated network data. These are internal correctness problems, not disagreements with community consensus, and they cannot be repaired by local edits; the equity term would need to be reformulated as a function of the restoration decisions and the experiments would need to be rerun. I see no indication of bad faith, but the current manuscript is not publishable in its present form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The central equity claim in Q-RESTORE doesn't hold up: the Gini term E in Eq. (7) is a constant computed from fixed zone incomes, so the µ sweep in Fig. 5b cannot come from the stated objective. The paper's own equations rule out the headline result. That is the thing to know before reading.\n\nWhat the paper does well: it's a clean application of D-Wave's hybrid CQM solver to the Sioux Falls restoration problem, with a standard bi-level formulation (BPR travel times, user-equilibrium lower level, budget/capacity constraints) and a reasonable statement of the GA baseline. The runtime observation (~8.7s vs ~665s) is plausible for a commercial hybrid solver on a 24-node benchmark, and the authors are honest that this is a hybrid, not pure quantum annealing. There's no parameter fitting to a target result; BPR constants come from the literature and income thresholds are stated assumptions.\n\nThe soft spots are bigger than the strengths. The equity mechanism is load-bearing and it's not derived. E depends only on I_r and I_s, not on the restoration variables C_a, so for a fixed budget the optimal plan is identical for all µ. Figure 5b contradicts Eq. (7). There are also mechanical errors: the text says µ = 0.2 while Table III's caption says µ = 0.5; the capacity tables are internally inconsistent (links 10, 7, 43); and no code or data are released. The GA comparison is under-specified (no convergence criteria, no seed info) and the GA underutilizes the budget (72.38 of 75), which weakens the \"quantum beats GA\" claim. None of these are fatal on their own, but the constant-equity issue is.\n\nThe authors are not being deceptive; the writing is clear and the framework is presented in good faith. But the central contribution — equity-aware prioritization through the weighted objective — doesn't exist as written. This is a routine application of a commercial solver to a known benchmark, not a novel framework.\n\nBottom line: don't send this to peer review in its current form. The authors need to fix the equity term (make it depend on restored capacities or on connectivity to low-income zones), rerun the experiments, and release the data. If they do that, there could be a modest contribution on the quantum-computing-for-restoration side. For now, this is a cautionary example for a reading group, not a citable result.","headline":"The equity term in Q-RESTORE is constant in the restoration variables, so the paper's own equations contradict the central equity claim; the framework needs a rewrite, not a referee.","tokens_in":20025,"tokens_out":2867,"would_cite":false,"duration_ms":28662,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Q-RESTORE claims a hybrid quantum solver can plan post-disaster road repairs in about 8.7 seconds while restoring low-income areas first.","keywords":["Q-RESTORE","quantum annealing","hybrid quantum solver","transportation network restoration","equity","Gini coefficient","genetic algorithm","post-disaster recovery"],"falsifier":"Re-run the Q-RESTORE optimization with all zone incomes set equal, or with the income labels randomly permuted; a genuinely equity-sensitive objective must change which links are restored, while the objective as written in Eq. (7) yields identical restoration plans. Alternatively, evaluate $E$ from Eq. (7) for two different feasible restoration solutions and observe that the value is unchanged, which would contradict the reported dependence of allocations on $\\mu$.","tokens_in":18875,"feed_emoji":"🛣️","tokens_out":10212,"duration_ms":86486,"temperature":0.7,"pith_summary":"Q-RESTORE is a proposed framework for post-disaster transportation network restoration that casts link-recovery choices as a constrained quadratic optimization problem and solves it with a hybrid quantum-classical annealer. The paper claims this solver produces restoration plans in roughly 8.7 seconds regardless of budget size, where a genetic algorithm needs more than 600 seconds, and that the plans consistently give more restored capacity to links serving low-income neighborhoods. The intended payoff is practical: emergency managers could recompute equitable recovery plans in near real time as damage reports change, instead of waiting on slow population-based search. A sympathetic reading is that quantum hybrid optimization is being positioned as a decision-support tool that can put equity on the same objective function as traffic efficiency.","feed_headline":"Hybrid quantum solver plans road repairs in 8.7 seconds","feed_subtitle":"A quantum-classical optimizer beats a genetic algorithm on post-disaster recovery and puts low-income areas first.","key_machinery":"The carrying object is the Hamiltonian of a constrained quadratic model, $H(x) = \\mu D + (1-\\mu) E + \\lambda_1(\\sum_a M_a(C^1_a) - B)^2 + \\lambda_2 \\sum_a \\max(0, C^0_a + C^1_a - C_a)^2$, in which $D$ is a recovery-deficiency index built from BPR travel times and $E$ is a Gini-coefficient equity term over zone incomes. The hybrid solver anneals a transverse-field Hamiltonian that interpolates from a uniform superposition to this problem Hamiltonian, with classical post-processing enforcing constraints. The genetic algorithm comparison uses the same fitness pieces $R_j = \\mu D_j + (1-\\mu) E_j + \\rho_j$ with tournament selection, one-point crossover, and budget-preserving mutation, so the two solvers are claimed to differ only in how the same objective is searched.","core_discovery":"On the paper's own terms, the central discovery is that the hybrid quantum solver, applied to a bi-objective restoration problem on the Sioux Falls network, maximizes recovery per dollar while putting equity first: links in low-income zones receive the largest restored capacities at every tested budget, and only after those are served do average- and high-income links receive capacity. The reported runtimes are 8.75, 8.75, 8.75 and 8.75 seconds for budgets of 75, 150, 225 and 300, against 665.0, 665.9, 677.7 and 674.7 seconds for the genetic algorithm. The paper interprets the flat runtime and the rising low-income share as evidence that a hybrid quantum annealer can handle the constraint-heavy restoration search faster than evolutionary search and with an equity-oriented allocation policy.","pith_inferences":["The equity term $E$ in Eq. (7) depends only on fixed zone incomes $I_r$ and $I_s$, not on the restored capacities $C^1_a$; if that is the objective the solver actually received, changing $\\mu$ could not change the optimal link choices, so the reported $\\mu$-sweep likely requires a different or augmented equity formulation.","A direct test of the equity mechanism is to permute the income labels on the 24 zones and rerun the optimization: a genuinely equity-driven objective must shift restored capacities accordingly, while the constant-$E$ version would leave them unchanged.","The speed comparison is against a single GA configuration with population 50, mutation rate 0.1, and tournament size 3; a tuned or warm-started GA might close much of the 600-second gap, so the headline result should be read as 'this hybrid solver beats this GA setup,' not as a general quantum-classical advantage.","For deployment, the useful output is not one plan but a frontier: re-running the solver with several $\\mu$ values traces the trade-off between travel-time recovery and equity, and each frontier point is cheap enough to show to decision makers in real time."],"forward_implications":["If the 8.7-second runtime is representative, restoration plans can be re-optimized on the fly as new damage assessments arrive, shrinking the gap between data collection and decision.","Because runtime stays flat across budgets from 75 to 300, the paper's claim implies that the hybrid solver's cost is dominated by fixed annealing and post-processing overhead, not by the number of links restored.","An equity-first allocation rule follows directly: spend early budget on low-income-serving links, then extend to average- and high-income links as the budget grows, which is a concrete policy recipe for recovery funds.","The GA comparison implies that population-based evolutionary search is the relevant classical bottleneck, so future classical baselines should be judged on the same budget-utilization and equity criteria, not only solution time.","If the framework scales to larger cities, the solver still needs a fixed demand matrix and link-capacity inputs, meaning the practical constraint is data update frequency rather than optimization compute time."],"supporting_citations":[{"why":"Supplies the fixed travel demand for the Sioux Falls test network, the data on which all restoration runs are evaluated.","marker":"[84]"},{"why":"Provides link capacities and free-flow travel times for the Sioux Falls network used in the experiment.","marker":"[85]"},{"why":"Defines the restoration optimization variant whose mobility and accessibility measures the paper adopts for its objective.","marker":"[86]"},{"why":"Earlier demonstration of quantum computing for transport network design problems that the paper extends with equity constraints.","marker":"[27]"},{"why":"Describes the control system and annealing hardware behind the hybrid solver's reported 8.7-second runs.","marker":"[87]"},{"why":"Gives the processor topology and embedding tooling the hybrid solver uses to map the CQM onto the quantum unit.","marker":"[88]"},{"why":"Supplies the adiabatic annealing argument that the ground state of the problem Hamiltonian corresponds to the optimal restoration plan.","marker":"[89]"}],"fun_headline_variants":["Quantum hybrid solver rebuilds roads in 8.7 sec","8.7-second quantum plan for fair road repair","Quantum solver beats genetics on equity-aware road fixes","Hybrid quantum method restores roads fast, fair","Quantum-driven restoration prioritizes low-income zones"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the equity term in the objective actually determines which links get restored; as written, $E = \\frac{1}{2N^2\\bar{I}}\\sum_{r,s}|I_r - I_s|$ depends only on neighborhood incomes, not on the restoration variables, so it cannot by itself produce the reported low-income-first allocation.","fun_headline_variants_meta":{"raw":{"variants":["Quantum hybrid solver rebuilds roads in 8.7 sec","8.7-second quantum plan for fair road repair","Quantum solver beats genetics on equity-aware road fixes","Hybrid quantum method restores roads fast, fair","Quantum-driven restoration prioritizes low-income zones"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000176,"raw_usage":{"total_tokens":1277,"prompt_tokens":922,"completion_tokens":355,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":280}},"tokens_in":538,"tokens_out":355,"duration_ms":3294,"temperature":1.0,"reasoning_tokens":280,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T18:32:40.237249+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the Q-RESTORE optimization with all zone incomes set equal, or with the income labels randomly permuted; a genuinely equity-sensitive objective must change which links are restored, while the objective as written in Eq. (7) yields identical restoration plans. Alternatively, evaluate $E$ from Eq. (7) for two different feasible restoration solutions and observe that the value is unchanged, which would contradict the reported dependence of allocations on $\\mu$.","supporting_citations":[{"cited_title":"An optimal schedule for urban road network repair based on the greedy algorithm,","cited_arxiv_id":null,"evidence_quote":"Supplies the fixed travel demand for the Sioux Falls test network, the data on which all restoration runs are evaluated."},{"cited_title":"A modified active set algorithm for transportation discrete network design bi-level problem,","cited_arxiv_id":null,"evidence_quote":"Provides link capacities and free-flow travel times for the Sioux Falls network used in the experiment."},{"cited_title":"Transportation infrastructure restoration opti- mization considering mobility and accessibility in resilience measures,","cited_arxiv_id":null,"evidence_quote":"Defines the restoration optimization variant whose mobility and accessibility measures the paper adopts for its objective."},{"cited_title":"Quantum computing for transport network design problems,","cited_arxiv_id":null,"evidence_quote":"Earlier demonstration of quantum computing for transport network design problems that the paper extends with equity constraints."},{"cited_title":"A scalable control system for a superconducting adiabatic quantum optimization processor,","cited_arxiv_id":null,"evidence_quote":"Describes the control system and annealing hardware behind the hybrid solver's reported 8.7-second runs."},{"cited_title":"Quantum annealing-infused microgrids formation: Distribution system restoration and resilience enhancement,","cited_arxiv_id":null,"evidence_quote":"Supplies the adiabatic annealing argument that the ground state of the problem Hamiltonian corresponds to the optimal restoration plan."}],"review_version":1}