{"id":"b27510d8-0f70-4223-b6ce-88d54520674f","arxiv_id":"2608.05427","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"An ILP for school consolidation is solved to optimality for Calabria's full public school network and is also cast as a constrained quadratic model for D-Wave's hybrid solver, which matched every classical optimum.","lead":"This paper builds an optimization model that decides which Italian public schools should merge with a larger nearby school, balancing travel time, enrollment limits, and protection of fragile municipalities. The authors test it on synthetic networks and on all public schools in Calabria, and they also run the model on D-Wave's hybrid quantum service.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Objective sign error: Eqs. (1)/(9) penalize same-municipality and same-track aggregations despite prose claiming rewards, so published optimal plans may not implement the stated policy.","rationale":"The reader's verdict identified the reward-penalty sign mismatch as one issue, but their weakest_assumption focused on missing travel-time and fragility data. I agree the data gap is real, but the sign error is more load-bearing because it is internal: even with perfect inputs, the objective function in Eqs. (1) and (9) penalizes the very consolidations the prose claims to reward. This undermines the central claim that the framework 'preserves the structural characteristics of the educational system' and 'promotes solutions that reduce organizational complexity.' The error is easily testable by changing the signs of the C5 and C6 terms and re-running the reported experiments; if the solutions change, the published tables do not correspond to the described decision-support tool. I therefore recommend keeping the reader's CONDITIONAL verdict, conditional on correcting the objective and regenerating all reported optimal plans. I do not recommend outright rejection because the ILP structure, constraints, and case-study setup appear coherent and the fix is straightforward; however, the current version is not internally consistent and should not be accepted without this correction and re-verification.","tokens_in":33424,"tokens_out":5339,"duration_ms":51005,"concrete_test":"Re-solve the 64 Calabria configurations with the objective terms corrected to implement rewards, e.g., replace +C5(1−β_ij)y_ij with −C5(1−β_ij)y_ij and +C6(1−θ_ij)y_ij with −C6(1−θ_ij)y_ij in Eqs. (1) and (9), keeping all other constraints and data identical. If any optimal objective value or hub assignment differs from Tables E.12/F.14, the published model does not match the stated policy, and the experimental results must be regenerated with the corrected objective.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is that the model produces optimal aggregation plans that preserve accessibility, capacity, and institutional compatibility. That claim fails as printed because the objective function contradicts the policy description. In Eq. (1) and its CFLP equivalent Eq. (9), the fifth and sixth terms are +C5(1−β_ij)y_ij and +C6(1−θ_ij)y_ij, with C5, C6 > 0. Since the model minimizes, these terms add cost when β_ij = 0 (same municipality) and when θ_ij = 0 (compatible track). The text in Section 4.1 explicitly says these terms 'reward aggregations between schools belonging to the same municipality' and 'reward aggregations between schools with compatible curricular structures.' A reward should appear with a negative sign in a minimization, not a positive one. Therefore, the optimizer is actually penalized for intra-municipal and track-compatible consolidations, and may prefer cross-municipal or track-incompatible aggregations whenever C5 or C6 exceed the corresponding penalty coefficients C1 or C2. The reported sensitivity results compound this confusion: Section 5.3.1.1 groups C6 with C1 as 'inter-municipality penalty coefficients,' while C6 is attached to (1−θ_ij), a curricular term. As a result, the 'optimal' aggregation plans in Tables E.12/E.13 and F.14 do not demonstrably implement the policy objectives claimed in the abstract and conclusions. This is an internal inconsistency, independent of the accuracy of the Calabria travel-time and fragility inputs.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an Integer Linear Programming (ILP) framework for school network reorganization, with two equivalent formulations (a baseline model and a capacitated facility location reformulation) that jointly incorporate provincial and municipal boundaries, institutional and curricular compatibility, territorial criticality, travel-time accessibility, capacity limits, and a global reorganization target. The authors introduce a synthetic benchmark generator, evaluate the classical solver Gurobi on synthetic instances of sizes 250–1000 and on a real dataset covering all public schools in Calabria, and reformulate the model as a Constrained Quadratic Model solved with D-Wave's hybrid quantum solver. The quantum solver is reported to reproduce the classical optimal objective in all tested configurations, with zero optimality gap and zero variance across repeated runs.","tokens_in":33815,"tokens_out":10484,"duration_ms":82711,"significance":"If the model and the reported solutions were correct, the paper would represent a substantive contribution to educational planning: it integrates an unusually rich set of policy constraints into a single optimization model, provides a reproducible synthetic instance generator, reports a complete regional case study, and gives an honest proof-of-concept assessment of hybrid quantum optimization in which the quantum solver is explicitly not claimed to outperform classical exact methods. The authors also deserve credit for the transparent comparison protocol and for making the real-world dataset available on request. However, the central claim that the model 'rewards' intra-municipal and track-compatible aggregations is contradicted by the printed objective function, and the sensitivity analysis misattributes the corresponding coefficients C5 and C6. These internal inconsistencies mean that the reported 'optimal' aggregation plans do not demonstrably implement the stated policy objectives and that the main conclusions of the computational study are not supported as written.","major_comments":[{"comment":"The objective function in Eq. (1) and its CFLP equivalent in Eq. (9) contain the terms +C5(1−β_ij)y_ij and +C6(1−θ_ij)y_ij, with C5, C6 > 0, under minimization. Since β_ij = 1 for cross-municipal pairs and θ_ij = 1 for track-incompatible pairs, these terms add positive cost exactly when the aggregations are intra-municipal (β = 0) and track-compatible (θ = 0). This is the opposite of the reward structure described in Section 4.1, where 'the fifth term rewards aggregations between schools belonging to the same municipality' and 'the sixth term rewards aggregations between schools with compatible curricular structures.' Consequently, the optimizer is penalized for the very consolidations the policy claims to favor, and the reported 'optimal' aggregation plans in Tables B.5, C.6, E.12, and F.14 do not demonstrably implement the stated policy objectives. The signs should be reversed (or the terms defined explicitly as negative rewards), and the entire computational campaign must be re-run under the corrected objective before the results can be interpreted.","section":"Section 4.1, Eq. (1); Section 4.2, Eq. (9)"},{"comment":"The sensitivity analysis repeatedly groups C1 and C6 as 'inter-municipality penalty coefficients' and C2 and C5 as 'curricular compatibility coefficients.' For example, the n=250 paragraph states that 'the strongest positive correlation is observed for the inter-municipality penalty coefficients C1 and C6,' and Section 6.2.1 reports that 'the geographical coefficients remain the principal drivers ... Pearson correlation coefficients equal to 0.57 for C6 and 0.35 for C1.' However, by the definitions in Section 4, C6 multiplies (1−θ_ij), a curricular-track term, and C5 multiplies (1−β_ij), a geographic term. The labels are therefore swapped, so the reported Pearson correlations do not measure what the text claims they measure. The conclusion that 'geographical coherence represents the primary determinant of the final aggregation strategy' is not supported by the presented statistics. The correlation analysis should be re-run with the correct mapping between coefficients and objective terms, and the associated discussion in Sections 5.3.1.2 and 6.2.1 should be revised accordingly.","section":"Section 5.3.1.1 and Section 6.2.1"},{"comment":"The accessibility constraints are effectively never binding in the synthetic study: Section 5.3.1.1 reports Pearson correlations for η_CI and η_USI that are numerically equal to zero for all instance sizes, and the thresholds are derived from the observed travel-time range by Eq. (22), which guarantees a large feasible arc set regardless of policy. For the real case, the paper fixes T_ICmax = 1200 s and T_USImax = 2400 s (Appendix E caption) but does not provide the road-network travel-time matrix, the criticality classification, or the enrollment thresholds used for eligibility filtering. As a result, the claim that the model 'preserves accessibility' in the Calabria case cannot be checked, and the reported optimal plans depend on unverifiable inputs. The authors should publish the real-case data or at least summary statistics (e.g., distributions of travel times on candidate arcs and the fraction of arcs satisfying each threshold) and should discuss whether the thresholds are binding in the optimal solutions.","section":"Section 5.2.2, Eq. (22); Section 5.3.1.1; Section 6.1, Appendix E"}],"minor_comments":[{"comment":"Section 3.6 states that the macro-level parameter defines the 'exact minimum proportion of schools that must remain active,' but constraint (5) is an upper bound (≤ γ n) on the number of active schools. The wording should be changed to 'maximum proportion' to match the mathematics and the subsequent interpretation of γ in Section 5.3.1.1.","section":"Section 3.6 and Section 4.1, constraint (5)"},{"comment":"The sentence listing application areas repeats 'packing problems' twice (see 'packing problems [10, 11], personnel scheduling [12, 13, 14, 15], packing problems [10, 16]'); one occurrence should be removed.","section":"Section 2"},{"comment":"The caption says 'maximum execution times were fixed to TICmax = 1200 s and TISmax = 2400 s'; these are maximum admissible travel times, not execution times, and the notation in the main text uses T_CI_max and T_USI_max.","section":"Appendix E, caption of Table E.12"},{"comment":"The statement that the real dataset 'does not include school institutions classified as being at risk of dimensioning' is unclear and appears to conflate the criticality-level-4 coefficient C4 with the eligibility concept; please clarify the intended meaning.","section":"Section 6.2.1"},{"comment":"The symbol N is used both for the full set of schools (Section 3) and for the eligible subset (Section 4, Table 2), which forces the reader to rely on font distinctions that are not visible in all contexts. A distinct symbol for the eligible subset would improve clarity.","section":"Section 4 and Table 2"}],"recommendation":"major_revision","confidential_remarks":"The sign error in the objective function is the central issue: because Eqs. (1) and (9) positively weight (1−β_ij) and (1−θ_ij), the printed model penalizes the very intra-municipal and track-compatible consolidations that the text claims are rewarded. A simple sign change would fix the formulation, but all reported optimal values, sensitivity analyses, and quantum comparisons would need to be recomputed, and the C5/C6 mislabeling in the interpretation would need to be corrected at the same time. I therefore recommend major revision rather than reject, because the underlying modeling idea is sound and the flaws are correctable within a revision cycle. Please also ask the authors to make the real-case travel-time and criticality data available for verification."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Useful applied OR paper, but the objective function as printed inverts the stated policy: intra-municipal and track-compatible aggregations are penalized, not rewarded. The sign error is load-bearing, and the sensitivity analysis misassigns C5 and C6. Fixable, but as submitted I would not trust the reported optimal plans.\n\nWhat is genuinely new: the ILP in Section 4 pulls together road-network travel times, territorial criticality, capacity, and institutional compatibility in one model for Italian school dimensioning, which is a real step beyond Bruno et al. (2016). The synthetic benchmark generator is a useful addition since no standard dataset exists, and the Calabria case study covers the complete network with 190 binary variables solved in well under 0.2 seconds. That runtime makes it genuinely usable as a planning tool. The literature review is solid, and the systematic sensitivity analysis is a plus, even if the interpretation of the coefficients is wrong.\n\nThe main problem is Eq. (1) and its CFLP twin Eq. (9). With beta_ij = 1 for cross-municipal pairs and theta_ij = 1 for incompatible tracks, the terms +C5(1-beta) and +C6(1-theta) add cost for same-municipality and track-compatible aggregations. The text says these terms \"reward\" those aggregations. In a minimization, that is the opposite. The reported optimal plans may therefore be biased toward cross-municipal and track-incompatible consolidations whenever C5 and C6 dominate. The sensitivity analysis compounds the problem: Section 5.3.1.1 groups C1 and C6 together as \"inter-municipality penalty coefficients\" and C2 and C5 together as \"curricular compatibility coefficients,\" but C6 is actually on the curricular term and C5 on the municipal term. Either the equations or the parameter labels are wrong, and the reader cannot tell which.\n\nMinor soft spots: the synthetic thresholds in Eq. (22) are derived from the observed travel-time range, and the correlation analysis confirms they are effectively non-binding, so the accessibility constraints are not really tested. The real-case travel times and fragility inputs are not shipped, only the school list. The D-Wave results show exactly zero variance and zero gap over five runs for all 64 cases; that needs a sentence explaining why (likely the instances are tiny), otherwise it reads as too good to be true.\n\nWho it is for: applied OR researchers and regional education authorities. The quantum part is a minor proof of concept. With the sign error fixed and the data shipped, this would be a solid applied paper. As it stands, the central claim fails on its own terms. A serious editor should still send it to peer review—the framework is novel and the flaw is correctable—but the reviewers should require a clean revision, not a minor edit.","headline":"Useful applied OR model for school dimensioning, but the objective function as printed rewards the opposite of what the text claims; needs a clean revision before its policy results are trusted.","tokens_in":34290,"tokens_out":5248,"would_cite":false,"duration_ms":44028,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that school network consolidation can be exactly optimized by one integer program that encodes travel-time, capacity, and institutional constraints, and that real regional instances solve in under 0.2 seconds.","keywords":["School network reorganization","School dimensioning","Integer Linear Programming","Hybrid quantum optimization","Facility Location Problem","Educational planning","Real-world case study","Territorial sustainability"],"falsifier":"Take the Calabria school network, recompute travel times from an independent road-network dataset, and re-solve one policy configuration; if the optimal aggregation plan changes or any chosen hub pair exceeds the stated thresholds, the claimed accessibility-aware optimality is not reproducible. For synthetic instances, set the travel-time thresholds externally rather than deriving them from the observed range and check whether the accessibility constraints bind; the paper's own near-zero correlations suggest they rarely do.","tokens_in":33225,"feed_emoji":"🏫","tokens_out":12790,"duration_ms":107200,"temperature":0.7,"pith_summary":"The paper is trying to establish that school network reorganization—deciding which autonomous schools stay open and which are absorbed into a hub—can be captured in one integer linear programming model. The model combines the constraints that govern Italian school dimensioning: aggregations cannot cross provinces or mix educational levels, curricular mismatches are penalized, fragile territories get lower enrollment thresholds, travel time over the road network is capped by school type, and hub capacity is bounded. If the model is right, regional authorities gain a decision-support tool that is exact and fast: the complete Calabria public school network was solved to optimality in under 0.2 seconds for every one of 64 policy configurations, and a hybrid quantum-classical solver reproduced those optima in all configurations. The practical point is that consolidation policy can be tested scenario by scenario, with accessibility and territorial equity made explicit rather than left to administrative judgment.","feed_headline":"One model finds optimal school merger plans in under a second","feed_subtitle":"It weighs travel time, capacity, and fragile-area protections; a quantum solver matches the classical result.","key_machinery":"The load-bearing object is the capacitated facility location formulation (Variant II) of the school aggregation problem, with binary hub variables $z_j$ and assignment variables $y_{ij}$ defined only on the feasible arc set $\\tilde{A}$. This set already requires same educational category and same province ($\\delta_{ij}+\\alpha_{ij}=2$), enrollment ordering ($s_i \\le s_j$), and travel time within the type-specific thresholds $T^{\\text{CI}}_{\\max}$ or $T^{\\text{USI}}_{\\max}$, so the compatibility constraints are pushed into the variable domain. The assignment constraints then force each school to exactly one configuration, residual capacities bound each hub, and the reorganization target $\\gamma$ caps the number of active hubs. The objective is a weighted sum of penalties for cross-municipal aggregation, curricular-track mismatch, and involvement of criticality-3 and criticality-4 schools, with rewards for intra-municipal and track-compatible aggregations.","core_discovery":"The paper's central claim is that school dimensioning can be solved as a single compact integer program that simultaneously honors administrative geography, institutional compatibility, territorial fragility, road-network accessibility, capacity limits, and the required proportion of remaining autonomous schools. The compact formulation embeds the hard compatibility rules in the domain of the assignment variables, so the solver only considers admissible aggregations, and it reaches proven optimality on every tested instance: 512 parameter combinations at each of four synthetic sizes up to 1,000 schools, and 64 policy configurations for the complete Calabria public school network, with average runtimes around 0.1 seconds and always below 0.2 seconds in Calabria. Reformulated as a constrained quadratic model, the same problem was submitted to a hybrid quantum-classical solver, which in every tested configuration reproduced the classical optimum with zero standard deviation and a zero optimality gap. The authors interpret the quantum experiments as validation of the reformulation and as a proof-of-concept benchmark, not as evidence of a quantum speed advantage.","pith_inferences":["The paper's travel-time constraints may not be seriously stressed in its own synthetic instances: thresholds are derived from the observed travel-time range (Equation 22), and the near-zero correlation between $\\eta$ values and objective values suggests the accessibility constraints rarely bind; an external, tight threshold would test them.","In the Calabria data, no school reaches criticality level 4 and Vibo Valentia never produces a hub or an aggregation, so the territorial-protection and cross-municipal penalty terms are partly dormant in the validation; adding a genuinely fragile and fragmented province would exercise those components.","Because the hybrid quantum service is a black box, the perfect zero-variance match cannot separate the annealer's own contribution from classical preprocessing; instrumenting that split would clarify what the quantum hardware is actually doing.","A natural next step beyond the paper is to replace the static enrollment and travel-time inputs with projected demographic and infrastructure scenarios, making the aggregation plan a time-indexed sequence rather than a single snapshot."],"forward_implications":["Authorities can rerun the optimization under many policy weights in seconds, which turns school dimensioning from a one-shot administrative act into a scenario-analysis exercise.","The aggregation plans are structurally stable: hub counts and cross-municipal aggregations change only modestly when penalty weights change, so different policy priorities do not force a completely different network.","The compact formulation scales to at least 1,000 schools with about 0.1 seconds average solution time, so it can serve as an exact baseline for future heuristics and for larger regions.","The constrained-quadratic reformulation is a faithful benchmark for quantum optimization: current hybrid hardware matches the classical optimum with zero gap, though not with a time advantage.","The synthetic instance generator supplies reproducible test cases for a problem that previously had no public benchmark dataset."],"supporting_citations":[{"why":"This work supplies the closest prior model of Italian school dimensioning through administrative mergers, which the paper extends with road-network accessibility, territorial fragility, and the full regulatory framework.","marker":"[35]"},{"why":"This work supplies the early location-and-size model for urban schools and establishes the optimization framing for school facility planning.","marker":"[22]"},{"why":"This work supplies the dynamic multi-period facility location model for school networks that the paper's capacitated location formulation follows.","marker":"[29]"},{"why":"This work supplies a p-median location-allocation redesign of a real secondary school network that serves as a reference for accessibility- and capacity-aware planning.","marker":"[30]"},{"why":"This work supplies a school consolidation model that explicitly balances accessibility and resource utilization, the same trade-off the paper encodes in its objective.","marker":"[37]"},{"why":"This work supplies a fairness-oriented closure-prioritization model for shrinking regions that motivates the territorial-fragility protections in the paper's objective.","marker":"[39]"},{"why":"This work supplies a consolidation model optimizing closures, assignments, and capacity utilization that the paper builds on for the aggregation structure.","marker":"[26]"},{"why":"This reference documents the exact ILP solver used as the classical benchmark against which all optimality gaps are measured.","marker":"[40]"}],"fun_headline_variants":["Quantum solver matches classical optimum for school mergers","Optimal school clustering in 0.1 seconds, even with quantum","School network optimization: classical and quantum agree on optimal plans","Fast optimal school mergers: one model, many constraints, quantum-verified","Merging schools optimally: sub-second classical, quantum confirms"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The Calabria results assume that the road-network travel times, territorial-criticality indices, and enrollment figures used as inputs are accurate and complete, and that the thresholds $T^{\\text{CI}}_{\\max}=1200$ s and $T^{\\text{USI}}_{\\max}=2400$ s are the right policy limits; if those inputs or thresholds are wrong, the reported optimal aggregation plans are not optimal for the actual region.","fun_headline_variants_meta":{"raw":{"variants":["Quantum solver matches classical optimum for school mergers","Optimal school clustering in 0.1 seconds, even with quantum","School network optimization: classical and quantum agree on optimal plans","Fast optimal school mergers: one model, many constraints, quantum-verified","Merging schools optimally: sub-second classical, quantum confirms"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000151,"raw_usage":{"total_tokens":1180,"prompt_tokens":904,"completion_tokens":276,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":520,"completion_tokens_details":{"reasoning_tokens":201}},"tokens_in":520,"tokens_out":276,"duration_ms":3155,"temperature":1.0,"reasoning_tokens":201,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T13:19:13.409504+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the Calabria school network, recompute travel times from an independent road-network dataset, and re-solve one policy configuration; if the optimal aggregation plan changes or any chosen hub pair exceeds the stated thresholds, the claimed accessibility-aware optimality is not reproducible. For synthetic instances, set the travel-time thresholds externally rather than deriving them from the observed range and check whether the accessibility constraints bind; the paper's own near-zero correlations suggest they rarely do.","supporting_citations":[{"cited_title":"Bruno, E","cited_arxiv_id":null,"evidence_quote":"This work supplies the closest prior model of Italian school dimensioning through administrative mergers, which the paper extends with road-network accessibility, territorial fragility, and the full regulatory framework."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work supplies the early location-and-size model for urban schools and establishes the optimization framing for school facility planning."},{"cited_title":"Antunes, D","cited_arxiv_id":null,"evidence_quote":"This work supplies the dynamic multi-period facility location model for school networks that the paper's capacitated location formulation follows."},{"cited_title":"Teixeira, A","cited_arxiv_id":null,"evidence_quote":"This work supplies a p-median location-allocation redesign of a real secondary school network that serves as a reference for accessibility- and capacity-aware planning."},{"cited_title":"Bhatnagar, N","cited_arxiv_id":null,"evidence_quote":"This work supplies a school consolidation model that explicitly balances accessibility and resource utilization, the same trade-off the paper encodes in its objective."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work supplies a fairness-oriented closure-prioritization model for shrinking regions that motivates the territorial-fragility protections in the paper's objective."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work supplies a consolidation model optimizing closures, assignments, and capacity utilization that the paper builds on for the aggregation structure."},{"cited_title":"URL:https: //www.gurobi.com/documentation/","cited_arxiv_id":null,"evidence_quote":"This reference documents the exact ILP solver used as the classical benchmark against which all optimality gaps are measured."}],"review_version":1}