{"id":"dfd5ec0c-12b1-461a-bbc5-f2b2d28b2b3d","arxiv_id":"2505.12608","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"A flow-based spatial contiguity constraint is encoded as DQM and QUBO quantum annealing models with a hybrid seed scheme, but no experiments are run.","lead":"This paper translates the flow-based spatial contiguity constraint from classical regionalization into DQM and QUBO formulations intended for D-Wave quantum annealers. It adds a hybrid seed-based scheme for larger maps, but contains no experiments and leaves several constraint details unresolved.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The root-identity constraint is never enforced, so the stated flow equations admit invalid p-region solutions with empty regions.","rationale":"The reader's verdict correctly flags the root mechanism, but the precise failure is more severe than 'the equations become inconsistent.' Because the predetermined roots are not constrained to their own regions, the flow equations remain consistent in a degenerate way: a root whose region is empty behaves exactly like a non-root node, so an all-in-one-region assignment can satisfy every stated penalty. This is a formal correctness gap in the central claim that the model ensures spatial contiguity for p-regions. Adding explicit constraints d_{r(k)}=k would close the empty-region loophole, but it would also expose the second, still-existing limitation: the model solves a rooted variant of p-regions, since any feasible partition must contain the preselected roots. The absence of hardware runs or simulations is a separate evidentiary weakness, but the under-specified root constraint is sufficient to reject the paper's strongest claim as written. The theorem itself is reasonable under the intended reading, and the classical flow argument is sound, which is why the concern is precise rather than a wholesale dismissal.","tokens_in":12823,"tokens_out":14084,"duration_ms":161743,"concrete_test":"Set up the Section 6 QUBO for the path graph 1-2-3 with p=2, r(1)=1, r(2)=2, and positive lambda_1 through lambda_4. Assign x_11=x_21=x_31=1 (all areas to region 1), x_i2=0, encode f12=2 and f23=1 using binary expansions with enough bits, set u_ij,k=x_ik*x_jk, and evaluate every penalty term in the complete QUBO. If all penalty terms are zero, the formulation accepts an empty region 2 and therefore does not solve the p-regions problem as defined.","verdict_should_be":"REJECT","load_bearing_attack":"Section 5 predetermines a root r(k) for each region and states parenthetically 'so that d_{r(k)}=k', but neither the Complete DQM Formulation nor the Complete QUBO Formulation in Section 6 contains a constraint enforcing d_{r(k)}=k. This omission is not benign: the flow-conservation equations remain globally consistent even when a region is empty. In the QUBO, the set I\\{r(d_1),...,r(d_p)} excludes every predetermined root from the non-root penalty, while a root whose own region is empty is covered by the lambda_4 penalty for that region, which requires outflow-inflow = -1, exactly the equation of a non-root node. Concretely, take the three-node path 1-2-3 with p=2 and roots r(1)=1, r(2)=2; assign all areas to region 1 and set f12=2, f23=1. Then every lambda_1 through lambda_4 penalty is zero, so the model accepts the invalid partition in which region 2 is empty, violating Definition 3.1. The contiguity theorem may hold for nonempty regions containing their preselected roots, but the model as written does not ensure a valid p-region partition.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a DQM and a QUBO formulation of the p-regions problem with an explicit spatial-contiguity constraint, using a network-flow model with one predetermined root (source) per region. It claims to be the first quantum model that ensures spatial contiguity among neighboring spatial entities, and it sketches a hybrid quantum-classical seeding approach for larger instances. The central technical content is Theorem 5.1, which states that if flow conservation holds in each region, then the region is connected, and the subsequent DQM/QUBO encodings of the flow constraints, assignment variables, and penalties.","tokens_in":13094,"tokens_out":5307,"duration_ms":53364,"significance":"If the model were correct, the paper would fill a real gap: no previous work encodes spatial contiguity for regionalization in a form directly mappable to D-Wave quantum annealers. The flow-based encoding restricted to adjacency edges is a sensible variable-reduction idea, and the paper correctly identifies Shirabe's flow model as the appropriate starting point. However, the current formulation does not enforce the root-assignment condition, and the QUBO penalty set is not well defined because it depends on the decision variables. These are load-bearing issues: the model as written admits invalid p-region partitions with empty regions, so the central claim of ensuring spatial contiguity for the p-regions problem is not supported. The paper also provides no experimental validation or resource estimates despite asserting direct hardware solvability in the abstract.","major_comments":[{"comment":"","section":"Section 5, 'Complete DQM Formulation' and Section 6, 'Complete QUBO Formulation'"},{"comment":"","section":"Section 6, 'Complete QUBO Formulation'"},{"comment":"","section":"Section 5, 'Intra-Region Flow Constraint' and 'Complete DQM Formulation'"},{"comment":"","section":"Abstract and Section 8"}],"minor_comments":[{"comment":"","section":"Section 5, 'Complete DQM Formulation'"},{"comment":"","section":"Throughout"},{"comment":"","section":"Section 3, Definition 3.1"},{"comment":"","section":"Section 4, Figure 2"},{"comment":"","section":"Section 7"}],"recommendation":"major_revision","confidential_remarks":"This is a workshop-style paper that has been accepted at a SIGMOD workshop; it is being submitted to a journal with the same content. The main concern is that the central formulation is not correct as written, but the errors appear fixable by adding root-assignment constraints and rewriting the QUBO penalty set without depending on decision variables. The lack of any experimental validation or resource estimates is also substantial for a paper claiming direct hardware solvability. Given the workshop provenance, the authors might prefer to add experiments and rework the formulation before a journal submission."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper has a real idea, but the central model as written doesn't enforce what it claims. The root-identity constraint d_{r(k)}=k is stated in passing in Section 5 and then never added to either the DQM or the QUBO objective. The stress-test example is right: on the three-node path with p=2, assign all nodes to region 1 and set f12=2, f23=1; every lambda_1 through lambda_4 penalty is zero, so the model accepts a partition where region 2 is empty. That is a load-bearing gap, not a cosmetic one.\n\nWhat is genuinely new deserves credit: the DQM/QUBO encoding of Shirabe's flow model with flow variables restricted to adjacency edges, the binary expansion of flow variables, and the seeding-based hybrid decomposition. Those encoding details are not in the cited literature, and restricting flows to neighborhood edges is a sensible way to cut qubit count. The flow conservation theorem is standard and correctly stated for fixed roots. The paper also cites the relevant classical work on Shibabe and Duque, so the lineage is clear.\n\nThe soft spots go beyond the root gap. There are no hardware runs or simulations anywhere; Section 8 says only that empirical validation is ongoing. Predetermining roots changes the p-regions problem, which has no roots in its original definition, so the model solves a rooted variant unless seed selection is globally optimal. The set notation I\\{r(d_1),...,r(d_p)} in the QUBO is ambiguous, which is another sign that the constraint layer is under-specified. Penalty coefficients are said to be 'suitably high' with no guidance on setting them.\n\nNone of this kills the underlying idea—a corrected model needs explicit root-assignment penalties (e.g., add a term forcing x_{r(k),k}=1) and at least a small annealer demonstration. But as written, the mathematical claim that the model ensures contiguity for the p-regions problem is not supported. This reads like a workshop paper, and it is one (Q-Data 2025).\n\nMy recommendation: a serious referee should see it, not because it is near publishable, but because the flaw is concrete, fixable, and worth documenting. If the authors add the missing constraints and run a small test, this could be a useful contribution to the quantum-spatial optimization literature. I'd send it to review with a major-revision decision. Your reader's REJECT is fair for the current arXiv version; I'd soften that only slightly because the core idea is salvageable.","headline":"The contiguity encoding has a real idea, but the root-assignment constraint is never enforced, so invalid partitions with empty regions have zero penalty—this needs fixing before the paper claims to solve p-regions.","tokens_in":13612,"tokens_out":2832,"would_cite":false,"duration_ms":27369,"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":"The paper claims the first quantum model that enforces spatial contiguity among neighboring spatial entities, using flow variables on adjacency edges in DQM and QUBO forms plus a hybrid seed-based extension for larger maps.","keywords":["quantum annealing","spatial contiguity","p-regions problem","network flow model","discrete quadratic model","QUBO","hybrid quantum-classical optimization","spatial regionalization"],"falsifier":"Take a path of four areas with $p=2$ and fix the roots at areas 1 and 2. The connected partition $\\{\\{1,2\\},\\{3,4\\}\\}$, which places both roots in one region, cannot be represented by the model even if it minimizes the dissimilarity objective; constructing the DQM/QUBO for this instance and comparing its ground-state objective with the unrestricted optimum therefore settles whether the rooted restriction changes the problem.","tokens_in":12619,"feed_emoji":"🗺️","tokens_out":9095,"duration_ms":90290,"temperature":0.7,"pith_summary":"The paper sets out to make spatial contiguity—the requirement that all areas in a region form one connected piece—expressible as an optimization model that quantum annealers can accept. The key move is to encode contiguity through network flow: each region has a preselected source area, every other area in the region receives one unit of flow from that source, and flow is allowed only between neighboring areas. This flow constraint is delivered in two forms, a Discrete Quadratic Model (DQM) for small instances and a Quadratic Unconstrained Binary Optimization (QUBO) formulation for binary hardware, together with a hybrid seeding scheme that grows regions classically and uses quantum optimization to move boundary areas. Because flow variables exist only on adjacency edges, the model avoids the quadratic variable blow-up of pairwise formulations. If the formulation is right, the same flow gadget can enforce contiguity in any spatial optimization problem, not just p-regions.","feed_headline":"First quantum model enforces spatial contiguity of regions","feed_subtitle":"A flow-based DQM/QUBO encodes connected p-regions using only adjacency edges, with a hybrid route for large maps.","key_machinery":"The load-bearing object is a directed flow network defined on the adjacency graph of the spatial areas. Each area $i$ is a node labeled by its region $d_i$; each directed edge that corresponds to a shared border carries a flow $f_{ij}\\ge 0$, so the number of flow variables equals the number of adjacency edges rather than all pairs of areas. The constraint $u_{ij}=|d_i-d_j|$ switches flow off between different regions, and the inflow-minus-outflow equation, with value $1$ at every non-root and $1-|R_k|$ at the region's root, forces every area to be reachable from the root. This is what turns a global topological property (connectedness) into local linear equations, and the paper's variable-reduction argument is what makes the DQM/QUBO encoding feasible on current hardware.","core_discovery":"The paper claims to present the first quantum model that ensures spatial contiguity constraints among neighboring spatial entities, using the p-regions problem as its testbed. With assignment variables $d_i\\in\\{1,\\dots,p\\}$, the model adds nonnegative flow variables $f_{ij}$ on adjacency edges only, and gates them by $u_{ij}=|d_i-d_j|$ so that flow cannot cross region borders. For each region $k$, a root area $r(k)$ is fixed in advance with $d_{r(k)}=k$; every non-root area in the region must satisfy $\\sum_{j\\in N(i)}f_{ji}-\\sum_{j\\in N(i)}f_{ij}=1$, and the root must supply $|R_k|-1$ units. The paper proves (Theorem 5.1) that these conservation equations force each region to be connected, and argues that the resulting DQM is small enough to run directly on quantum annealing hardware for small datasets, while a hybrid seed-growing scheme extends the approach to larger maps by restricting quantum optimization to movable border areas.","pith_inferences":["Beyond the paper's claims, the rooted formulation means the model actually solves a rooted variant of p-regions: a contiguous partition is representable only when each region contains exactly one preselected root. Extending to the original problem means optimizing root choice jointly, unless one can prove some optimal partition contains the scattered seeds.","Beyond the paper's claims, the assertion that the DQM is solvable directly on quantum annealing hardware is not backed by a hardware run or simulation in the paper, so whether annealers find ground states reliably for these energy landscapes remains an open empirical question, likely sensitive to penalty weights.","Beyond the paper's claims, a natural testable extension is to make the roots themselves discrete decision variables; on sparse maps the added assignment dimension may still fit within current qubit limits and would remove the rooted restriction."],"forward_implications":["Any region-based optimization that needs contiguity—electoral districting, forest planning, site selection, wildlife corridors—can in principle use the same flow gadget, since the paper presents the model as general rather than p-regions-specific.","Variable count grows with the number of shared borders, so maps with sparse spatial adjacency produce smaller DQM/QUBO instances and are better suited to current quantum hardware.","Small p-region instances can be encoded directly as a DQM and submitted to quantum annealing hardware; larger maps can be split by seed selection, classical region growing, and a quantum pass over movable border areas.","When the flow conservation equations are satisfied, each region is provably connected (Theorem 5.1), so contiguity is guaranteed by the model rather than checked post hoc."],"supporting_citations":[{"why":"Supplies the network-flow contiguity mechanism that the paper extends from a single region to multiple regions.","marker":"[37]"},{"why":"Defines the p-regions problem and the mixed-integer baseline that motivates reducing variable count.","marker":"[13]"},{"why":"Provides the Discrete Quadratic Model framework the small-scale formulation targets.","marker":"[18]"},{"why":"Supplies the QUBO framework used for the binary encoding of the same flow constraints.","marker":"[16]"},{"why":"Motivates the hybrid quantum-classical solver architecture used for larger maps.","marker":"[10]"},{"why":"Establishes the seeding-based regionalization approach that the hybrid scheme builds on.","marker":"[28]"},{"why":"Identifies articulation areas so the hybrid step can filter movable border areas without breaking contiguity.","marker":"[39]"}],"fun_headline_variants":["First quantum model enforces spatial contiguity","Flow-based DQM enforces contiguity in p-regions","Hybrid quantum-classical route to contiguous regions","Quantum flow model guarantees connected spatial regions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The model fixes one source area per region in advance, but the original p-regions problem has no such sources, so the contiguity guarantee applies only to the rooted variant and the result quality depends on those preselected sources being compatible with a good partition.","fun_headline_variants_meta":{"raw":{"variants":["First quantum model enforces spatial contiguity","Flow-based DQM enforces contiguity in p-regions","Hybrid quantum-classical route to contiguous regions","Quantum flow model guarantees connected spatial regions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001092,"raw_usage":{"total_tokens":4526,"prompt_tokens":876,"completion_tokens":3650,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":492,"completion_tokens_details":{"reasoning_tokens":3589}},"tokens_in":492,"tokens_out":3650,"duration_ms":27544,"temperature":1.0,"reasoning_tokens":3589,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:30:37.071281+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a path of four areas with $p=2$ and fix the roots at areas 1 and 2. The connected partition $\\{\\{1,2\\},\\{3,4\\}\\}$, which places both roots in one region, cannot be represented by the model even if it minimizes the dissimilarity objective; constructing the DQM/QUBO for this instance and comparing its ground-state objective with the unrestricted optimum therefore settles whether the rooted restriction changes the problem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the network-flow contiguity mechanism that the paper extends from a single region to multiple regions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Discrete Quadratic Model framework the small-scale formulation targets."},{"cited_title":"2022.A Hybrid Solver for Constrained Quadratic Models","cited_arxiv_id":null,"evidence_quote":"Motivates the hybrid quantum-classical solver architecture used for larger maps."}],"review_version":1}