{"id":"6e1bca01-7dbb-4848-b138-4e165d55427b","arxiv_id":"2611.16362","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For each k, a cyclotomic Cohen-Macaulay h-polynomial of multiplicity 8k that is not a complete intersection exists exactly for k outside {1,2,3,6}, primes at least 7, and 3 times primes at least 17.","lead":"The paper proves that a product of cyclotomic polynomials of multiplicity 8k can be the h-polynomial of a Cohen-Macaulay ring without being a complete intersection exactly when k is not 1, 2, 3, 6, a prime at least 7, or 3 times a prime at least 17. It also reduces, for any fixed a, the question for infinitely many prime multiplicities ap to a finite check, a template that may simplify other cases in combinatorial commutative algebra.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"Nonexistence half of Theorem 1.3 rests entirely on the unshipped finite searches in Lemma A.1(ii)-(iii); a missed or misclassified candidate would overturn the classification.","rationale":"The reader's weakest assumption and mine coincide. The existence direction of Theorem 1.3 is supported by explicit constructions whose coefficient sequences can be checked directly, and the reduction lemmas in Sections 3 and 4 appear internally consistent. However, the only-if direction has no independent argument for the infinite excluded families: every family of the form 8p for p>=11, 24p for p>=29, and the exceptional 24p cases reduces to a finite search that is asserted but not shipped. This is a genuine correctness risk rather than a mere style issue, because Theorem 1.3 is a complete classification and the nonexistence claims are precisely the unverified part. If the independent reimplementation reproduces the empty intersection, the central claim becomes as secure as the rest of the proof and the paper could be accepted; until then, the conditional verdict is appropriate. I therefore recommend no change to the reader's verdict.","tokens_in":19150,"tokens_out":19274,"duration_ms":126734,"concrete_test":"Independently reimplement the Appendix A enumeration, e.g. in SageMath: for each a in {8,24} and each divisor ell of a with c = a/ell, enumerate all exponent vectors (e_m) such that sum e_m*phi(m) <= c-1 and prod Phi_m(1)^{e_m} = c, using m <= 2*phi(m)^2 for the finite factor range. For a=8, test C(t)*Psi_11(t); for a=24, test C(t)*Psi_29(t); and for Lemma A.1(iii), test C(t)*Psi_p(t) for p=17,19,23. Verify that no candidate has an M-sequence coefficient sequence and a negative b_n under Lemma 2.2. Publishing the full candidate list, or a certificate for the empty intersection, would settle whether the concern lands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The only-if direction of Theorem 1.3 is not established by a closed-form argument. For k=p with p>=11, the sole exclusion is Theorem 1.2 with a=8 together with Lemma A.1(ii); for k=3p with p>=29, the sole exclusion is Theorem 1.2 with a=24 together with Lemma A.1(ii); and the cases p=17,19,23 rest on Lemma 4.8 plus Lemma A.1(iii). The reduction chain itself appears coherent, but the terminal step in each case is the Appendix A assertion 'No candidate meets both required conditions.' The appendix states the enumeration bounds (deg C(t) <= a/ell - 1, phi(m) <= c-1, and m <= 2*phi(m)^2) but ships no code, no logs, and no candidate lists, and it does not document the exact cutoff used for the Lemma 2.2 b_n checks or the implementation of Macaulay's inequalities. Since the infinite nonexistence families of Theorem 1.3 collapse onto this finite search, a single missed or misclassified candidate would falsify the classification: for example, a surviving C(t) for a=8, ell=1 would produce a CM-realizable non-type-CI product of multiplicity 88, directly contradicting the theorem for k=11. I found no internal inconsistency in the reduction chain; the vulnerable premise is precisely the exhaustiveness and correctness of the Appendix A computation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper studies which products of cyclotomic polynomials can occur as h-polynomials of standard graded Cohen-Macaulay algebras without being of complete intersection type. The main result is a complete characterization for multiplicities of the form 8k: such a polynomial exists if and only if k is not in the set N = {1,2,3,6} ∪ {p: p≥7 prime} ∪ {3p: p≥17 prime}. The proof combines a reduction theorem (Theorem 1.2) that limits the nonexistence problem for infinitely many prime parameters to a finite search, with explicit constructions for the existence cases. The finite searches are reported in Appendix A.","tokens_in":19547,"tokens_out":3743,"duration_ms":22377,"significance":"If correct, the paper provides a complete classification for a nontrivial infinite family of multiplicities, which is a significant advance over the previous work [7] that handled the non-Cohen-Macaulay case. The reduction theorem (Theorem 1.2) is a clever technique that may be applicable to other values of a. The constructions are explicit and the nonexistence arguments are systematic. The paper also gives partial results for other multiplicities and discusses Gorenstein realizability.","major_comments":[{"comment":"The enumeration for n ≤ 99 is described as: 'Enumerate these finite sequences, retain those satisfying Macaulay's inequalities, and factor the corresponding polynomials over Q. Apply Lemma 2.2 whenever all irreducible factors are cyclotomic.' This is also not reproducible without code or a detailed list. However, the n ≤ 99 case is less load-bearing because the main theorem concerns 8k, and the small n are only used for the excluded k in N (k=1,2,3,6,7) and for the partial results in Section 5. The paper could be accepted without this if the main result's nonexistence for k≥11 is fully verifiable, but currently it is not.","section":"Appendix A, Lemma A.1(i)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The bound m ≤ 2φ(m)^2 is used to limit the search. This bound is not proved or referenced in the paper. The authors state 'm ≤ 2φ(m)^2 ≤ 2(c-1)^2'. For c up to 24 (when a=24, ℓ=1 gives c=24), this gives m ≤ 2*23^2 = 1058, which is manageable but still large. The bound m ≤ 2φ(m)^2 is not standard; the standard bound is φ(m) ≥ √(m/2) for m>6, which gives m ≤ 2φ(m)^2. But this is not stated. The paper should either provide a reference or a short proof. More importantly, the enumeration of all products of cyclotomic polynomials with degree ≤ c-1 and value at 1 = c is nontrivial because the exponents e_m can be large. The description 'the degree bound also limits the exponents' is insufficient. For example, Φ_2(t) has degree 1 and value 2, so to achieve c=24, one could have up to 24 factors of Φ_2, giving degree 24, which exceeds c-1=23. So the degree bound does limit exponents, but the enumeration of all combinations of cyclotomic polynomials with degree ≤ c-1 and product value c is a finite but nontrivial combinatorial search. The paper does not describe how this search was implemented or how exhaustiveness was ensured. This is a critical gap.","section":"Appendix A, Lemma A.1(ii)-(iii)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 4.8 for p=17,19,23 contains a step that I find unclear. The authors write: 'If 2s > R, then minimality of s implies h_i ≥ i+1 for 0 ≤ i ≤ ⌊R/2⌋. Since A(t)Φ_p(t) is palindromic, R ≥ p+23, and p+23 is even, we have 24p = ∑ h_i ≥ (⌊R/2⌋+1)^2 ≥ ((p+25)/2)^2 > 24p, a contradiction.' The inequality ∑ h_i ≥ (⌊R/2⌋+1)^2 assumes that h_i ≥ i+1 for all i up to ⌊R/2⌋, and then uses palindromicity to double the sum? Actually, the sum from 0 to R is at least 2∑_{i=0}^{⌊R/2⌋} (i+1) - (if R even, the middle term counted once). The expression (⌊R/2⌋+1)^2 is the sum of the first (⌊R/2⌋+1) odd numbers? Let's check: ∑_{i=0}^{m} (i+1) = (m+1)(m+2)/2. Twice that minus the middle gives roughly (m+1)(m+2) - something. The claimed bound (⌊R/2⌋+1)^2 is not obviously correct. For R=40, ⌊R/2⌋=20, (21)^2=441, while 2∑_{i=0}^{20} (i+1)=2*231=462. So the bound is actually weaker, so the inequality might still hold, but the derivation is sloppy. More importantly, the authors then claim 'R ≥ p+23, and p+23 is even, we have ... ≥ ((p+25)/2)^2 > 24p'. For p=17, p+23=40, (p+25)/2=21, 21^2=441, 24p=408, so 441>408, contradiction. For p=19, p+23=42, (p+25)/2=22, 22^2=484, 24p=456, 484>456. For p=23, p+23=46, (p+25)/2=24, 24^2=576, 24p=552, 576>552. So the numerical check works, but the derivation of the bound should be clarified. This is a minor issue in a major comment because the lemma is load-bearing for the p=17,19,23 cases, but the argument can be fixed with a clearer inequality.","section":"§4.4, Lemma 4.8","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"In the same proof, the authors write: 'Suppose otherwise, and put P_j = ∑_{i=0}^j a_i, with P_j = 0 for j<0. For 0≤j≤d, the coefficient formula for h_j implies P_j = h_j + P_{j-p}. Since h_j ≥ 0, induction on j shows that P_j ≥ h_j ≥ 0. Splitting the coefficients of A(t) into three consecutive blocks and using palindromicity, we obtain 24 = A(1) = ∑_{i=0}^{s-p-1} a_i + ∑_{i=s-p}^{s-1} a_i + ∑_{i=s}^d a_i = P_{s-p-1} + h_{s-1} + P_{d-s} ≥ s + h_{d-s}.' The step 'P_{d-s} ≥ h_{d-s}' is claimed but not justified. The induction argument gives P_j ≥ h_j for all j, but here we have P_{d-s} where d-s could be less than p? Actually, the induction works for all j, so P_{d-s} ≥ h_{d-s} is valid. Then the authors claim 'h_{d-s} ≥ min{d-s+1, s}'. This is because if d-s ≤ s, then by minimality of s, h_{d-s} ≥ d-s+1; if d-s > s, then h_{d-s} ≥ h_s = s (since the sequence is nonincreasing after s). This is plausible but should be stated explicitly. The overall argument is sound but the presentation is compressed. This is a minor clarity issue, not a fatal error.","section":"§4.4, Lemma 4.8","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The definition of N in (1) uses 'p: p≥7 prime' and '3p: p≥17 prime'. For clarity, note that 3p for p=17,19,23 are included, but these are also covered by the separate treatment in §4.4. This is fine.","section":"§1, Introduction","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The definition of b_n(f(t)) uses e_{nr}. It might be helpful to note that the sum is over r≥1 such that nr is an index where e_{nr} is defined (i.e., e_m=0 for m not in the support).","section":"§2.1, Lemma 2.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The reference [7, Proposition 3.2] is cited for the bound deg f(t) ≤ f(1)-1. The paper also notes that this follows from the absence of internal zeros for CM-realizable polynomials. This is fine.","section":"§2.2, Lemma 2.5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof says 'If d=0, then A(t)=1; if d=1, then A(t)=1+t. The assertion follows in both cases.' It would be clearer to explicitly state that for d=0, deg=0 ≤ a-1 holds because a≥1; for d=1, deg=1 ≤ a-1 requires a≥2, which holds because A(1)=a and A(t) is a product of cyclotomic polynomials with degree 1, so a=2. This is fine.","section":"§3, Lemma 3.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof uses the inequality 2∑_{i=0}^{p-1} (i+1) = p(p+1) > ap. This requires a < p+1, which holds because p > a. Good.","section":"§3, Lemma 3.3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The notation 'α = p+r-s' appears. It might be helpful to note that α is the smallest exponent in W(t) when W(t) ≠ 0.","section":"§3, Lemma 3.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The final inequality 'a = qs + C(1) > qs + α = p + r + (q-1)s ≥ p > a' uses that α = p+r-s, so qs+α = qs + p + r - s = p + r + (q-1)s. The last inequality uses (q-1)s ≥ 0 and r ≥ 0, so p+r+(q-1)s ≥ p. This is correct.","section":"§3, Lemma 3.5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The coefficient sequence of S_0(t) is given as (1,4,6,5,5,6,4,1). It would be helpful to note that this is palindromic and has length 8, so degree 7. The check of Macaulay's inequalities is described briefly. This is acceptable.","section":"§4.1, Proposition 4.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof for k≥7 says 'the only new increases are from 12 to 15 and from 15 to 16'. It would be clearer to state the coefficient sequence explicitly for general k. The description is sufficient.","section":"§4.2, Theorem 4.3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The verification of Macaulay's inequalities for F_{p,q}(t) is quite involved. The paper provides a table for h_7, h_8 and then uses Lemma 2.6 for 8 ≤ i ≤ p-2. The bounds for the remaining indices p-1 to p+5 are given in a table. This is thorough but dense. A reader might benefit from a brief explanation of how the coefficient formula leads to the bounds. However, the level of detail is acceptable for a research paper.","section":"§4.3, Proposition 4.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The three polynomials Q_7, Q_11, Q_13 are given. It might be helpful to note that these correspond to k=21,33,39 respectively, as used in the proof of Theorem 1.3.","section":"§4.5, Proposition 4.11","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper states 'In this subsection, the ground field is Q.' This is important because the g-theorem and Adiprasito's result are over Q. The paper should clarify that the Gorenstein realizability results are over Q, while the CM realizability results are over any field (as per Theorem 2.4). This is stated but could be emphasized.","section":"§5.4, Gorenstein realizability","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The description of the finite verification for (i) says 'Enumerate these finite sequences, retain those satisfying Macaulay's inequalities, and factor the corresponding polynomials over Q.' Factoring polynomials over Q from coefficient sequences is not straightforward: given a coefficient sequence, one must construct a polynomial and factor it. The paper does not explain how this is done or how one ensures that all irreducible factors are cyclotomic. This is a gap, but since the main result does not depend on the n≤99 enumeration (except for small k), it is not load-bearing for the main theorem. However, for completeness, the authors should provide more detail or code.","section":"Appendix A, Lemma A.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper uses the term 'CM-realizable' to mean 'is the h-polynomial of a standard graded Cohen-Macaulay algebra'. This is defined in the introduction. The equivalence with M-sequences is given in Theorem 2.4. This is clear.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper uses 'type CI' to mean a product of Ψ_m(t) polynomials. This is defined in Definition 2.1. The characterization via b_n in Lemma 2.2 is useful.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Reference [7] is listed as 'arXiv:2511.07937, 2025'. This is a preprint. The paper cites it for the degree bound in Lemma 2.5 and for the non-Cohen-Macaulay classification. This is acceptable as long as the preprint is publicly available. However, the paper should note that [7] is a preprint, not yet published.","section":"References","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"The honest headline: Theorem 1.3 is a genuine classification and Theorem 1.2 is a genuinely reusable structural result, but the only-if direction rests at its terminal step on a finite search that is described, not shipped. Treat the classification as conditional until that computation is reproducible.\n\nWhat is actually new: Higashitani-Ueyama handled multiplicities 1,4,6 and primes, and gave non-CM realizations for composites. The Cohen-Macaulay version changes the answer, and the full characterization at 8k (forbidden set {1,2,3,6} plus primes at least 7 plus 3p for primes p at least 17) is not in the literature. Theorem 1.2 is the core: for fixed a, nonexistence at multiplicity ap over all sufficiently large primes reduces to finitely many pairs (ell, C(t)). That is a reusable tool.\n\nThe existence side is explicit and checkable. Theorems 4.2, 4.3, 4.6 and Proposition 4.11 give concrete products, and Lemma 2.7 provides a clean non-type-CI certificate that gets reused honestly throughout. I walked the Section 3 chain, including the flat-interval argument in Lemma 3.3 and the W(t)=0 elimination in Lemma 3.5, and found no false step. Section 5 scopes the 2k/4k situation honestly; the authors state those limitations themselves, and I do not read them as hidden defects.\n\nWhere it is soft: the only-if half of Theorem 1.3 collapses at its last step onto Lemma A.1(ii)-(iii), the assertion that \"no candidate meets both required conditions\" for a in {8,24} and for p = 17,19,23. The appendix gives the enumeration bounds (phi(m) <= c-1, m <= 2(c-1)^2, degree limit on exponents), so a determined reader could reimplement the search. But no code, no logs, no candidate lists are shipped. A single missed or misclassified candidate overturns the result: for a=8, ell=1, a surviving C(t) would produce multiplicity 88, contradicting the theorem for k=11. Lesser points: several per-example Macaulay checks (Propositions 4.1, 4.11, the Section 5 tables) are given as \"direct computation\" without intermediates, and the exact cutoff for the Lemma 2.2 b_n checks is undocumented. These are minor; the shipping of the search is the substantive gap.\n\nFor whom: anyone tracking the cyclotomic Hilbert-series thread (Borzi-D'Alì, Braun-Liu, Higashitani-Ueyama). It deserves a serious referee, yes, sent to review with the requirement that a revision ship the search code or machine-checked certificates, or at minimum the complete excluded candidate lists. My own verdict: accept conditionally on exactly that.","headline":"The 8k classification is real and the Theorem 1.2 reduction is reusable, but the only-if half hangs on a finite search the paper describes rather than ships; treat it as conditional until the computation is reproducible.","tokens_in":19995,"tokens_out":5941,"duration_ms":39925},"pith_extraction":null,"created_at":"2026-11-13T06:23:54.181503+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":null,"supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}