{"id":"8c7fd8e5-4b48-420f-bad0-841908c941c8","arxiv_id":"2607.00256","paper_version":3,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"Linearization of hypergraph reaction-diffusion systems retains only exposure statistics, while packing effects control post-onset amplitudes, yielding a structural visibility hierarchy and dynamical graph surrogacy.","lead":"Reaction-diffusion systems on directed hypergraphs only 'see' first-tail-moment exposure at linear onset; higher co-occurrence (packing) appears only in nonlinear saturation. This explains when graph surrogates work and when higher-order structure actually changes patterns.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The reader's weakest-assumption diagnosis is accurate: the order-independent mean aggregator is what produces the exact first-tail-moment collapse. That choice is not a flaw in the logic; it is the modeling scope that makes the visibility hierarchy sharp. Within that scope the proofs are standard (center-manifold reduction plus Kronecker spectral decomposition) and the numerics are consistent. No stronger load-bearing concern surfaces, so the ACCEPT verdict stands.","tokens_in":31132,"tokens_out":436,"duration_ms":4942,"concrete_test":"Independently recompute the Jacobian blocks (39)–(42) for the three-node Example 5.6 under the stated mean aggregator; verify that L_tot is exactly the exposure matrix and that β_pack vanishes when ω_e = 0, matching the closed-form expression (96). Agreement confirms the reduction steps used for Theorems 4.5 and 5.9.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim holds under the paper's stated hypotheses. Theorem 4.5 (first-tail-moment reduction) follows directly once the aggregator is order-independent mean-plus-fixed-activation (Eq. 25) and the mixing matrix is shared across orders (Section 4.3.1): every partial derivative of ψ_k collapses to the same Dσ factor, so only the CHOM contractions of the Laplacian tensors survive. The packing decomposition (Theorem 5.9, Proposition 5.4) is then a straightforward projection of the multilinear maps onto the critical left eigenvector. The modeling restriction is the price of the clean hierarchy, but it is declared explicitly, the derivations are self-contained, and the small-system numerics (Tables 1–2, Figures 4–5) confirm both linear indistinguishability and packing-driven divergence. No hidden inconsistency or unsupported leap appears in the argument for the strongest claim.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The paper develops a theory of structural visibility for reaction-diffusion systems on directed B-hypergraphs near a codimension-one steady bifurcation. Under order-independent nonlinear mean aggregation (DeepSet-style mean followed by a fixed smooth activation σ and a shared mixing matrix M), the linearized dynamics about a homogeneous equilibrium depends on the hypergraph only through its zeroth and first tail moments (exposure). Exposure-equivalent hypergraphs therefore share identical dispersion relations, critical eigenspaces, and instability thresholds (Theorem 4.5, Corollary 4.9). Beyond onset, a weakly nonlinear center-manifold reduction yields a scalar amplitude equation whose quadratic and cubic coefficients decompose into exposure-driven and packing-driven parts (Proposition 5.4, Theorem 5.9). Packing effects are contractions of higher-order tail-moment tensors with the critical left and right eigenvectors; they control post-onset saturation, branch selection, and pattern morphology. The paper formalizes weak packing-equivalence, nonlinear distinguishability of exposure-equivalent systems, and dynamical graph surrogacy (when packing effects vanish). Small directed hypergraph numerics with Schnakenberg kinetics confirm linear indistinguishability and packing-driven divergence of saturated amplitudes.","tokens_in":31382,"tokens_out":1097,"duration_ms":8805,"significance":"If the results hold under the stated hypotheses, the paper supplies a precise, order-by-order account of which higher-order structural features are dynamically accessible near Turing onset. The exposure reduction explains why clique-expansion and graph surrogates often reproduce linear instability thresholds, while the packing decomposition identifies the precise obstruction to nonlinear surrogacy. The notions of weak packing-equivalence and dynamical graph surrogacy are clean and falsifiable. Strengths include explicit tensor contractions, Kronecker spectral lemmas, center-manifold derivations with proofs in the appendix, and fully specified small-system numerics (Tables 1–2, Figures 4–5, C.1) that match the predicted coefficients without fitted constants. The modeling restriction to order-independent mean aggregation is the price of the clean hierarchy, but it is declared explicitly and yields a self-contained theory that advances the structural understanding of higher-order dynamical systems.","major_comments":[{"comment":"Section 4.3.1 and Eq. (25): the first-tail-moment reduction (Theorem 4.5) and the subsequent packing hierarchy rest on the assumption that the aggregator is order-independent mean-plus-fixed-activation with a single mixing matrix M across all interaction orders. This is declared, but the manuscript should state more prominently (abstract/introduction and discussion) that other aggregators or order-dependent activations generically re-introduce higher moments already at linear order, so the visibility hierarchy is model-class-specific rather than universal for all multiway operators.","section":null},{"comment":"Section 6 and Tables 1–2: the numerical validation uses only N=6 directed hypergraphs. While sufficient to illustrate the coefficients, the claim that packing effects control post-onset morphology would be stronger with at least one larger or random ensemble example showing that the same exposure/packing distinction persists beyond hand-crafted small systems. This is not a correctness issue but a load-bearing support gap for the empirical claims.","section":null}],"minor_comments":[{"comment":"Abstract and introduction: 'co-occurence' should be 'co-occurrence' (also appears in the structural dictionary).","section":null},{"comment":"Eq. (2) and Remark 2.6: the 1/k! normalization is convenient; a brief note that the same hierarchy holds (up to combinatorial factors) for the unnormalized tensor would help readers comparing with other hypergraph Laplacian conventions.","section":null},{"comment":"Figure 3 caption: the visibility hierarchy is clear, but labeling the asymptotic orders (linear/quadratic/cubic) next to the arrows would make the figure self-contained.","section":null},{"comment":"Proposition 5.4 / Remark 5.5: the indirect cubic contribution γ_B inherits packing through the center-manifold correction u; a short sentence noting that this term is computed numerically in the tables (rather than left symbolic) would clarify reproducibility.","section":null},{"comment":"Section 2.6 (directification): the embedding is well-defined; a one-line remark that undirected results inherit the same exposure/packing statements via the B-embedding would tighten the undirected claim.","section":null},{"comment":"References: a few recent works on Turing patterns on hypergraphs/simplicial complexes (e.g., Muolo et al. already cited) could be cross-referenced more explicitly when discussing graph surrogacy success at linear order.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The paper is a solid contribution to math.DS / higher-order networks. The modeling restriction is the main scope limitation; once flagged clearly it does not undermine the theorems. Fit for a serious dynamical-systems or network-dynamics venue is good. No integrity or citation concerns."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The punchline is simple: under their mean-plus-fixed-activation coupling, linearization of the hypergraph RD system only sees first-tail-moment \"exposure,\" so exposure-equivalent hypergraphs are linearly indistinguishable; higher co-occurrence (\"packing\") only reappears after projection onto the critical mode and controls the amplitude coefficients. That separation, plus the notions of weak packing-equivalence and dynamical graph surrogacy, is the real contribution.\n\nWhat is new is the systematic visibility hierarchy. They define the tail-moment tensors, prove the first-tail-moment reduction (Thm 4.5), then show that the quadratic and cubic normal-form coefficients decompose into exposure-driven pieces plus packing effects obtained by contracting the higher moments with the critical left/right eigenvectors (Prop 5.4, Thm 5.9). The corollaries on nonlinear distinguishability and when a hypergraph admits a dynamical graph surrogate follow cleanly. The derivations use standard center-manifold reduction plus Kronecker structure; the appendix proofs are explicit. Small, fully specified directed hypergraphs confirm both linear indistinguishability and packing-driven divergence of saturated amplitudes. Citations cover the hypergraph-dynamics and classical Turing literature without obvious gaps.\n\nThe soft spot is the modeling choice itself: order-independent mean aggregation and a shared mixing matrix (Sec 4.3.1, Eq 25). That is what collapses every partial of ψ_k to the same Dσ factor and produces the clean hierarchy. Other aggregators would generically re-introduce higher moments already at linear order. They state the restriction clearly and do not over-claim; it is the price of the sharp statements, not a hidden flaw. Numerics are tiny and illustrative only; free parameters (Schnakenberg a,b, activation coefficients) are fixed by hand, but no circular fitting occurs. Codimension-one steady bifurcations only; Hopf and multimode cases are left open.\n\nThis is for people working on hypergraph dynamics, Turing patterns on networks, or anyone who wants a precise criterion for when clique expansions are valid near onset. The math is solid under the stated hypotheses, the central claim holds, and the paper deserves a serious referee. I would engage with it.","headline":"Clean, usable theory of which hypergraph statistics survive linearization vs. nonlinear saturation; modeling restriction is explicit and the math holds.","tokens_in":31932,"tokens_out":542,"would_cite":true,"duration_ms":6554,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37G10","37N25","05C65","35B36","34C23"],"pacs":[],"model":"grok-4.5","headline":"Linearization of reaction-diffusion systems on hypergraphs sees only first-moment exposure; higher-order co-occurrence reappears only as packing effects that shape nonlinear saturation.","keywords":["structural visibility","directed hypergraphs","dynamical graph surrogacy","reaction-diffusion systems on hypergraphs","diffusion-induced instability","packing effects","exposure equivalence","pattern formation"],"falsifier":"Build two exposure-equivalent hypergraphs with deliberately different packing effects, drive them through the same codimension-one Turing onset, and check whether their saturated modal amplitudes (or branch orientations) diverge exactly as predicted by the difference in β_pack and γ_pack while their linear thresholds remain identical.","tokens_in":32031,"feed_emoji":"🕸️","tokens_out":665,"duration_ms":8998,"temperature":0.7,"pith_summary":"Hypergraphs can encode rich multiway interactions, but this paper argues that dynamics do not automatically make all of that structure visible. Near a pattern-forming instability, the linearized operator depends on the hypergraph only through first-tail-moment statistics called exposure, so any two hypergraphs that share those statistics have identical dispersion relations and the same onset threshold. Beyond onset, a weakly nonlinear amplitude equation recovers only certain higher-order marginals of the adjacency tensor, contracted against the critical left and right eigenvectors and called packing effects. Exposure therefore governs linear onset, while packing effects control branch selection, saturated amplitude, and pattern morphology. The same decomposition also says when a genuine higher-order system can be replaced, near onset, by a pairwise graph surrogate: precisely when the packing effects vanish.","feed_headline":"Hypergraphs hide most structure from linear pattern onset","feed_subtitle":"Only first-moment exposure sets the threshold; packing effects reappear in nonlinear saturation","key_machinery":"The nonlinear structural decomposition theorem: amplitude coefficients factor as β = β_exp + β_pack and γ = γ_exp + γ_pack, where packing effects are the contractions of higher tail-moment tensors with the critical eigenvectors that alone enter the reduced dynamics.","core_discovery":"The paper establishes a structural visibility hierarchy for reaction-diffusion systems on directed B-hypergraphs with nonlinear mean aggregation. At linear order the Jacobian depends on the hypergraph only through exposure (zeroth and first tail moments), so exposure-equivalent hypergraphs are linearly indistinguishable. At quadratic and cubic order the reduced amplitude coefficients split into exposure-driven pieces plus packing effects obtained by contracting pair- and triple-packing tensors with the critical mode; those packing effects alone determine whether linearly identical systems remain distinguishable after onset and whether a graph surrogate can reproduce the reduced dynamics.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Most hypergraph structure vanishes under linear pattern onset","Exposure alone sets linear thresholds for hypergraph patterns","Packing effects surface only after nonlinear pattern saturation","Linearly equivalent hypergraphs diverge via packing post-onset","Higher-order structure filters through a dynamical visibility hierarchy"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The coupling must be the same mean-then-activate aggregator and the same mixing matrix for every interaction order; otherwise higher moments can already appear in the linearization.","fun_headline_variants_meta":{"raw":{"variants":["Most hypergraph structure vanishes under linear pattern onset","Exposure alone sets linear thresholds for hypergraph patterns","Packing effects surface only after nonlinear pattern saturation","Linearly equivalent hypergraphs diverge via packing post-onset","Higher-order structure filters through a dynamical visibility hierarchy"]},"model":"grok-4.5","effort":"low","cost_usd":0.006264,"raw_usage":{"total_tokens":1678,"prompt_tokens":855,"num_sources_used":0,"completion_tokens":76,"cost_in_usd_ticks":62640000,"prompt_tokens_details":{"text_tokens":855,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":747,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":855,"tokens_out":76,"duration_ms":5939,"temperature":1.0,"reasoning_tokens":747,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-12T09:38:22.071185+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Build two exposure-equivalent hypergraphs with deliberately different packing effects, drive them through the same codimension-one Turing onset, and check whether their saturated modal amplitudes (or branch orientations) diverge exactly as predicted by the difference in β_pack and γ_pack while their linear thresholds remain identical.","supporting_citations":[],"review_version":3}