{"id":"b1d34173-9ce4-4749-8908-841cb3e130a4","arxiv_id":"1908.02583","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper introduces dimension-agnostic simplicial degree measures and a multi-combinatorial Laplacian, then uses them to describe higher-order connectivity in 17 real-world networks.","lead":"This paper defines new higher-order adjacency and degree notions for simplices in networks, plus a generalized Laplacian whose entries compute them. It then applies these measures to 17 real-world datasets and reports that the new maximal simplicial degree distributions often differ from classical node degree distributions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Empirical 'for every dataset' claim is unsupported: Tables 4-5 omit 3 of 17 datasets (congress-bills, tags-stack-overflow, threads-stack-overflow) with no explanation.","rationale":"The reader's verdict is CONDITIONAL, and my stress-test does not move that verdict: the algebraic contribution appears defensible, while the empirical claims need additional support. The reader's weakest assumption focused on the 25-node truncation and visual power-law inspection; I agree those are real problems, but I found a more concrete and checkable gap: the paper claims results for all 17 datasets, yet the degree-distribution tables and figures report only 14. The three missing datasets are not marginal: congress-bills is the one where the truncation tail visibly fails to decay, and the two stack-overflow datasets are the largest in the collection. Without them, the headline empirical statements about 'every dataset' and about power-law closeness cannot be verified. The concrete test is straightforward: process the missing datasets or explicitly restrict the claim. Until then, the conditional verdict stands.","tokens_in":38147,"tokens_out":7894,"duration_ms":83125,"concrete_test":"Check whether the three missing datasets (congress-bills, tags-stack-overflow, threads-stack-overflow) can be processed with the same pipeline at the stated q_m and q_p. Obtain the facet lists from the cited Benson et al. repository, enumerate the q-simplices and compute deg*_U and deg* for each, and add the rows to Tables 4-5 and the log-log panels. If enumeration is infeasible for congress-bills at q=10, rerun at a feasible dimension (e.g., q=1 or q=2) and explicitly report the restriction; then re-test whether the power-law and surprising-difference conclusions hold on the full 17-dataset set.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The theoretical core (Theorems 1-4 and the multi-combinatorial Laplacian construction) appears internally coherent; my concern is with the empirical side of the central claim. The abstract and Section 5.2 assert conclusions 'for every analysed dataset' and for '17 real-world datasets', but Tables 4 and 5 list statistics for only 14 datasets. congress-bills, tags-stack-overflow, and threads-stack-overflow are absent from both tables and from the degree-distribution figures, and no section or footnote explains the omission. This matters for two reasons. First, congress-bills is the dataset for which the authors themselves note the 25-node cap produces a tail that does not tend to zero (Section 5.1, after Table 2), so omitting it removes the clearest case where the truncation artifact could dominate. Second, tags-stack-overflow and threads-stack-overflow are the two largest datasets in Table 1 (14.5M and 11.3M simplices), so their omission could systematically affect the claimed 'rich variety' and the power-law comparison. If the missing datasets were not processed because the maximal simplicial degree is computationally infeasible at the chosen q_m/q_p (e.g., congress-bills q=10 requires enumerating C(25,11) 10-simplices per 25-node facet), the claim must be restricted to 14 datasets and the infeasibility disclosed. If they were processed, the omission is a reporting error that prevents verification. Either way, the empirical half of the central claim is currently unsupported as stated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes new notions of higher-order lower, upper, and adjacency degrees for simplices of possibly different dimensions in a simplicial complex, together with a 'multi-combinatorial Laplacian' built from multi-parameter boundary operators. Theorem 1 states that the entries of this Laplacian matrix are exactly the new diagonal degrees and oriented off-diagonal adjacency degrees, while Theorems 2-4 give closed formulas for the generalized lower, upper, and adjacency degrees. The authors also apply the newly defined maximal upper and maximal simplicial degrees to 17 real-world datasets from Benson et al., claiming that the maximal upper simplicial degree distribution is closer to a power law with more pronounced decay, and that the maximal simplicial degree distribution is surprisingly different from the classical node degree distribution.","tokens_in":38439,"tokens_out":9610,"duration_ms":99493,"significance":"If the theoretical part is correct, it fills a genuine gap: existing simplicial degree notions mostly compare simplices of the same dimension, whereas this paper gives a unified formalism for arbitrary-dimensional comparisons and encodes several of the resulting degrees as entries of a single Laplacian-type matrix. The algebraic definitions are explicit, Theorem 1 is a direct and checkable computation, and the worked examples help the reader. The connection of formula (9) to the earlier 'q-simplex to facets degree' of Moore et al. is also useful. However, the advertised empirical findings are not currently supported: three of the seventeen datasets are missing from the degree-statistics tables and figures, and the 'closer to a power law' claim is based on visual inspection of log-log plots with no fitting or statistical comparison. The theoretical core is likely salvageable, but the empirical side of the central claims needs substantial reworking or restriction.","major_comments":[{"comment":"The abstract and Section 5.2 state conclusions for '17 real-world datasets', but Tables 4 and 5 report degree statistics for only 14 datasets, and Figures 7-12 cover only 13 datasets. congress-bills, tags-stack-overflow, and threads-stack-overflow are absent without any explanation. This matters because congress-bills is the dataset for which the authors themselves note that the 25-node cap produces a tail that does not tend to zero (Section 5.1, after Table 2), and tags-stack-overflow and threads-stack-overflow are the two largest datasets in Table 1 by number of simplices. The omission must be disclosed, and either the missing tables and figures must be provided or every '17 dataset' claim must be restricted to the 14 datasets actually analyzed.","section":"Section 5.1, Tables 4-5"},{"comment":"The statement that 'for every analysed dataset the degree distribution associated with the maximal upper simplicial degree follows a power-law distribution similar to that of the corresponding distributions associated with the classical node and node-to-facets degree, but with a more pronounced decay' is not supported by any quantitative analysis. The evidence is visual inspection of log-log plots: no exponents are estimated, no standard errors or confidence intervals are given, and no comparison is made to alternative distributions or to a fitted power-law baseline. The authors' own caveat in Section 5.2 that 'empirical data is not enough to properly fit real-world degree distributions' and the Section 5.1 warning that the 25-node cap may cause exaggerated tail cutoffs directly undermine the word 'prove' in the bullet list. Please replace the proof language with descriptive observations or provide formal fits and goodness-of-fit tests.","section":"Section 5.2, bullet list and degree-distribution discussion"},{"comment":"The generalized boundary operator ∂_{q,h} is not unambiguously defined because the summation is written as ∑_{j1,...,jh} without specifying the index set. If the sum ranges over ordered h-tuples, then each (q−h)-face is produced multiple times, which would contradict Example 5, where ∂_{2,2}(v_ijk) is computed as v_i − v_j + v_k with one term per removed vertex. This is load-bearing because the matrices B_{q,h} used in Theorems 1-4 are the matrix representations of this operator. The definition should be clarified by writing, for example, 1 ≤ j1 < ⋯ < jh ≤ q, and the sign convention should be stated consistently with that convention.","section":"Definition 15, Section 4.1"}],"minor_comments":[{"comment":"The statement of Theorem 3 does not require p ≥ q, but the formula uses h = p − q and h′ = p − q′, so if p < q the exponents are negative and the boundary matrices are undefined. Either add the assumption p ≥ q (and p ≥ q′) or state separately that deg_p^U(σ^(q)) = 0 when p < q.","section":"Theorem 3"},{"comment":"The displayed inequality γ* ≥ γ*_U ≥ γ_F ≥ γ is presented as if it were read off from the figures, but no degree exponents are estimated anywhere in the paper. This should be explicitly labeled as a tentative observation or removed until actual fits are performed.","section":"Section 5.2, parameter inequality"},{"comment":"The manuscript does not describe the algorithm used to compute the maximal upper and maximal simplicial degrees from the facet lists, nor does it provide code or pseudocode. Since the statistics in Tables 4 and 5 are central to the empirical claims, a brief algorithmic description or a link to code would substantially improve verifiability.","section":"Reproducibility"},{"comment":"There are several minor typographical issues, including 'ad thus' in Example 1 and 'simplical network' in the conclusions; a careful proofreading pass is recommended.","section":"Typos and notation"}],"recommendation":"major_revision","confidential_remarks":"The theoretical core of the paper (Sections 2-4) appears internally coherent and could be publishable after revision. The main obstacle is the empirical section: the '17 datasets' claim is not backed by the presented tables and figures, and the power-law and 'surprisingly different' claims rest on visual inspection rather than statistical evidence. I would be willing to reconsider a revised version that either completes the missing dataset analysis with appropriate caveats or restricts all claims to the datasets actually presented, and that replaces 'prove' with language justified by fits or explicit statistical tests."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the algebraic contribution is real and worth engaging; the empirical half is not yet in publishable shape.\n\nWhat is actually new: the paper defines p-lower and p-upper adjacency for simplices of different dimensions, strict variants, generalized degrees, and a (q,h,h')-multi-combinatorial Laplacian whose diagonal entries are those degrees. Theorems 2-4 give closed counting formulas through boundary matrices. This is a legitimate extension of the same-dimension adjacency and degree notions already in the literature, and the Laplacian unification is a nice step beyond the iterative matrix products in the work they cite. I checked the small examples and the appendix proof sketches; they are careful sign-and-count bookkeeping, and nothing looks circular. The claim that the Laplacian computes the degrees is derived from explicit definitions, not inferred from fitted data.\n\nThe soft spots are all on the empirical side. The abstract and Section 5 say the conclusions hold for 'every dataset' and for '17 real-world datasets,' but Tables 4 and 5 list only 14. I looked: congress-bills, tags-stack-overflow, and threads-stack-overflow are absent from both tables and from every degree-distribution figure, with no explanation. That matters because congress-bills is the dataset where the 25-node cap is visibly problematic (its facet-size tail does not decay at 25), and the other two are the largest datasets by far. Either the maximal simplicial degree was computationally infeasible for those datasets at the chosen q, or the rows were omitted by mistake; either way the 'every dataset' statements are unsupported as written. The power-law claim for the maximal upper simplicial degree rests on visual inspection of log-log plots, with no fits, no goodness-of-fit, and no generative model. The authors themselves concede that the real impact of the 25-node threshold is unverified and that data incompleteness may exaggerate tail cutoffs. That honesty is to their credit, but it undercuts the empirical conclusions as stated. No code or detailed computational protocol is released either.\n\nBottom line: send it to peer review because the theoretical part deserves serious scrutiny and could become a reference for higher-order network measures. A referee should require the empirical section to be redone: report all 17 datasets or explicitly restrict the claims to 14 and explain why three are missing, fit and test the distributions, and make the code available. If that revision happens, this will be a useful paper.","headline":"Solid algebraic core, shaky empirical wrapper: the simplicial-degree framework and the multi-combinatorial Laplacian are worth engaging, but the 'for every dataset' claims overrun the evidence as presented.","tokens_in":38980,"tokens_out":2408,"would_cite":false,"duration_ms":26481,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["55U10","62R40","91D30","05C82","82M99","82B43","05E45"],"pacs":["89.75.-k","89.75.Fb","89.75.Hc"],"model":"deepseek-v4-flash","headline":"A single Laplacian matrix computes all higher-order simplicial degrees.","keywords":["complex networks","simplicial complexes","higher-order degree","multi-combinatorial Laplacian","topological data analysis","network science","simplicial hubs","degree distributions"],"falsifier":"Compute the diagonal and off-diagonal entries of $L_{q,h,h'}$ by direct enumeration on a small oriented simplicial complex and compare them with the definitions of the higher-order degrees: any mismatch would falsify Theorem 1. Separately, on a dataset such as congress-bills, remove the 25-node cap (or raise it) and re-fit the maximal upper and maximal simplicial degree distributions; if the more-pronounced decay and the difference from the classical node degree distribution vanish, the empirical claims are falsified.","tokens_in":1648,"feed_emoji":"🔺","tokens_out":10474,"duration_ms":131010,"temperature":0.7,"pith_summary":"This paper argues that relevance in a network should be measured not only for nodes and edges but for simplices — filled triangles, tetrahedra, and higher-dimensional collaborative groups — and that the right way to do it is a family of degrees that let simplices of different dimensions be compared directly. It introduces higher-order lower, upper, and adjacency degrees for any pair of simplices, and packages them in a new multi-combinatorial Laplacian whose matrix entries are exactly those degrees: higher-order degrees on the diagonal, oriented adjacency degrees off it. Closed formulas are proved for all the generalized degrees, giving an explicit mechanism to compute quantities that earlier work only described as a searching-and-counting procedure. Applied to 17 real datasets, the paper finds that a maximal upper simplicial degree has a distribution closer to a power law with stronger decay than the classical node degree, while the maximal simplicial degree — which also counts the facets containing a simplex's strict faces — produces distributions generally surprisingly different from classical ones, with higher small-degree saturation or bell shapes.","feed_headline":"One matrix computes all higher-order simplicial degrees","feed_subtitle":"New simplicial degrees span dimensions and reveal hub patterns missing from ordinary graphs.","key_machinery":"The carrying object is the multi-parameter boundary operator $\\partial_{q,h}:C_q(K)\\to C_{q-h}(K)$, which removes $h$ vertices from an oriented $q$-simplex with the appropriate signs, together with its adjoint coboundary and the resulting multi-combinatorial Laplacian $\\Delta_{q,h,h'}=\\partial_{q+h,h}\\circ\\partial^*_{q+h,h}+\\partial^*_{q,h'}\\circ\\partial_{q,h'}$. It generalizes the graph and $q$-combinatorial Laplacians, and its matrix entries encode the new degrees; the strict $(h,p^*)$-upper degree is further related to the non-strict one by an inclusion-exclusion formula with binomial coefficients. The applied analysis rests on the maximal upper simplicial degree and the maximal simplicial degree of Definition 14, which count distinct facets containing a simplex and, for the latter, also distinct facets containing its strict faces.","core_discovery":"The central claim is Theorem 1: for an oriented simplicial complex, the $(i,j)$-th entry of the multi-combinatorial Laplacian matrix $L_{q,h,h'}=B_{q+h,h}B_{q+h,h}^t+B_{q,h'}^tB_{q,h'}$ is the higher-order upper degree $\\deg^{h,q+h}_U(\\sigma_i^{(q)})$ plus the lower count $\\binom{q+1}{q-h'+1}$ on the diagonal, and the sum of upper and lower oriented degrees off the diagonal. Theorems 2, 3, and 4 give closed formulas for the general $p$-lower degree, $p$-upper degree, and $p$-adjacency (and maximal $p$-adjacency) degrees in terms of products of boundary-matrix entries. Empirically the paper claims that for all 17 datasets the maximal upper simplicial degree distribution is closer to a power law with a more pronounced decay, and that the maximal simplicial degree distribution is generally surprisingly different from the classical node degree distribution, revealing fewer and smaller simplicial hubs.","pith_inferences":["If the 25-node cap mostly removes very rare large facets, the qualitative shape comparisons should survive; a direct test would rebuild one coauthorship or congress dataset without the cap and check the distributions.","Because the multi-parameter boundary operator does not square to zero, its kernel is not ordinary simplicial cohomology; the spectrum of the multi-Laplacian could nonetheless be tested as a new invariant, for instance by checking whether its low eigenvalues change under simplicial subdivision.","The paper's observation that higher-dimensional simplices show more random-looking degree distributions suggests a testable scaling law: fit the tail exponent $\\gamma^*$ as a function of simplex dimension and see whether it increases monotonically toward the random-network value."],"forward_implications":["The graph Laplacian and the $q$-combinatorial Laplacian are recovered as the $h=h'=1$ case, so the multi-combinatorial Laplacian is a genuine common generalization.","The 'q-simplex to facets degree' of earlier work, previously left as a searching-and-counting procedure, becomes an explicit sum of strict upper degrees.","For all 17 datasets, the maximal upper simplicial degree distribution is closer to a power law with a more pronounced decay, including datasets whose classical node degree is not scale-free.","The maximal simplicial degree distribution differs in general from the classical node degree distribution, with higher small-degree saturation or bell-shaped random-network behavior.","Datasets of the same type (coauthorship, email, tags, threads) exhibit similar higher-order connectivity patterns."],"supporting_citations":[{"why":"Supplies the 17 real-world datasets, already collected as simplices bounded to 25 nodes, used for all empirical distribution analyses.","marker":"[3]"},{"why":"Gives the $q$-combinatorial Laplacian matrix entries and the similarly/dissimilarly oriented conventions that Theorem 1 generalizes.","marker":"[20]"},{"why":"Defines the $q$-simplex to $(q+h)$-simplex degree and the $q$-simplex to facets degree that the new closed formulas compute as strict upper degrees.","marker":"[31]"},{"why":"Provides the node-to-facets degree and its observed classical degree distributions, which serve as the comparison baseline in the empirical section.","marker":"[35]"}],"fun_headline_variants":["One matrix computes all higher-order simplicial degrees","Simplicial Laplacian reveals higher-order network patterns","New degree measures capture multi-agent interactions","Maximal simplicial degree exposes different hub structure","Higher-order connectivity in 17 real-world networks"],"cache_read_input_tokens":41088,"weakest_assumption_plain":"The empirical conclusions depend on treating the 25-node-truncated, duplicate-free facet lists as faithful enough to compare distribution shapes, and on judging 'closer to a power law' by visual inspection of log-log plots; if either fails, the empirical claims lose support, although the algebraic theorems would stand.","fun_headline_variants_meta":{"raw":{"variants":["One matrix computes all higher-order simplicial degrees","Simplicial Laplacian reveals higher-order network patterns","New degree measures capture multi-agent interactions","Maximal simplicial degree exposes different hub structure","Higher-order connectivity in 17 real-world networks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000203,"raw_usage":{"total_tokens":1436,"prompt_tokens":1043,"completion_tokens":393,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":659,"completion_tokens_details":{"reasoning_tokens":334}},"tokens_in":659,"tokens_out":393,"duration_ms":4503,"temperature":1.0,"reasoning_tokens":334,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:36:24.948180+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the diagonal and off-diagonal entries of $L_{q,h,h'}$ by direct enumeration on a small oriented simplicial complex and compare them with the definitions of the higher-order degrees: any mismatch would falsify Theorem 1. Separately, on a dataset such as congress-bills, remove the 25-node cap (or raise it) and re-fit the maximal upper and maximal simplicial degree distributions; if the more-pronounced decay and the difference from the classical node degree distribution vanish, the empirical claims are falsified.","supporting_citations":[{"cited_title":"Moore, R.J","cited_arxiv_id":null,"evidence_quote":"Defines the $q$-simplex to $(q+h)$-simplex degree and the $q$-simplex to facets degree that the new closed formulas compute as strict upper degrees."}],"review_version":1}