{"id":"c12983d9-b829-43bf-a9db-ae9737d8aed4","arxiv_id":"2608.20507","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves log_2 h(n) = n/2 + O(sqrt(n) (log(n+2))^3), so the sharp exponential base for abelian covers is sqrt(2).","lead":"This paper proves that the worst-case number of abelian subgroups needed to cover a group grows like sqrt(2) raised to the size of the largest pairwise noncommuting subset. It settles the sharp exponential form of Erdős Problem #117 and shows the extremal examples must be essentially 2-groups.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The p=3 case of Lemma 4.4 is the load-bearing point: the claimed 13-vector set is supported only by a hand-listed Gram table, and a single zero pairing would make the odd-prime upper-bound coefficient exceed 1/2.","rationale":"The paper is dense but well-structured. I followed the isoclinism reduction, the central-factor descent, the interaction clique lemmas, and the nilpotent/extension steps, and found no internal contradiction in the main line; the error bookkeeping in Proposition 6.2 and Theorem 6.1 is consistent once the p=3 credit is granted. The single most exposed assumption is the finite Gram check in Lemma 4.4: it is the only point where the claimed constant 1/2 for odd primes is established, and the displayed table is easy to copy wrongly. The reader flagged the same item, and I agree. Since the check is explicit, finite, and easily machine-verifiable, and since the rest of the argument is independent of any further numerical input, I do not think the reader's ACCEPT should be changed; the appropriate review action is to run the verification (or request it) as part of the process. If the check passes, the proof goes through; if it fails, the odd-prime upper bound has a real gap. Hence the verdict remains UNCHANGED.","tokens_in":14358,"tokens_out":32399,"duration_ms":282789,"concrete_test":"Independently recompute the 78 quantities phi(v_i,v_j)=x_i1*x_j2 - x_i2*x_j1 + x_i3*x_j4 - x_i4*x_j3 + x_i5*x_j6 - x_i6*x_j5 mod 3 for the 13 vectors listed in Lemma 4.4, e.g. with a short script. If every upper-triangular entry is nonzero, kappa_3=2 is confirmed and the concern is settled; if any zero appears, the p=3 scalar-clique credit fails and the proof must be repaired.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim Theorem 2.2 depends on the uniform p-group bound Theorem 6.1, whose odd-prime coefficient alpha_3=(log_2 3)/4 is strictly below 1/2. That coefficient comes entirely from Lemma 4.4's assertion that a rank-6 alternating form over F_3 has a pairwise nonorthogonal set of 13 points (credit 12). The paper supports this by displaying 78 upper-triangular Gram entries and asserting none vanish, but no machine check, code, or formal proof is supplied. If one of the 78 pairings is zero, the best credit from the hyperbolic-plane construction would be 3m=(3/2)rho, giving coefficient (log_2 3)/3 approx 0.528 > 1/2; then Theorem 6.1 would no longer give linear slack over the desired n/2 bound, and the proof of Theorem 2.2 as written would fail. This is the only numerical/structural input I found that is both unverified and load-bearing; the rest of the descent, interaction, and extension arguments are internally coherent.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the extremal function h(n), the supremum over groups G with at most n pairwise noncommuting elements of the least number of abelian subgroups needed to cover G. Its main theorem, Theorem 2.2, states that log_2 h(n) = n/2 + O(sqrt(n) (log(n+2))^3), so h(n)^{1/n} tends to sqrt(2). The proof combines an isoclinism reduction to finite groups, a BFC-based control of conjugacy classes and the derived subgroup, a central-factor descent for finite p-groups using alternating forms and isotropic subspace covers, interaction estimates via nested-anchor and exact-centralization constructions, a Sylow-product analysis for nilpotent groups, and a quasipolynomial reduction to the nilpotent subgroup C_G(G'). The lower bound comes from extraspecial 2-groups. The paper also derives the same exponential rate for the minimum index of an abelian subgroup and shows that asymptotic extremality is concentrated in 2-groups.","tokens_in":14500,"tokens_out":36119,"duration_ms":319922,"significance":"If correct, the result settles the sharp exponential scale of Erdős's covering problem, improving Pyber's uniform exponential bounds to the exact base sqrt(2). The proof is modular and the main structural arguments are coherent: the finite reduction, the symplectic toolkit, the branch-cover bound, the weak- and strong-interaction mechanisms, and the final error balance are independently checkable. I specifically checked the load-bearing p=3 finite verification in Lemma 4.4: the displayed 78 upper-triangular Gram entries are all nonzero and match the listed vectors, so the claimed value kappa_3 = 2 is valid. The lower and upper bounds are derived by independent mechanisms, so the matching constant is not fitted. The corollaries on the least index of an abelian subgroup and on extremality concentration in 2-groups are natural and well supported by the proof.","major_comments":[],"minor_comments":[{"comment":"As typeset in the submitted text, the statement a(E_m)=2m+1 is inconsistent with the proof, which correctly proves a(E_m)=2^m+1; likewise, '2m+1 Lagrangian members' and 'at most 2m-1 nonzero vectors' should read 2^m+1 and 2^m-1 respectively. The displayed arithmetic (2^{2m}-1)/(2^m-1)=2^m+1 shows the intended meaning, but the notation should be fixed in the final version.","section":"Lemma 4.6"},{"comment":"The plain-text version flattens many superscripts, for example 'order 2 1+2m' and 'log2h(n)>= n2', which makes the main theorem and Lemma 4.6 unnecessarily hard to read. The published version should restore consistent superscript typesetting.","section":"Throughout"},{"comment":"The finite verification for p=3 is load-bearing because a single zero pairing would break the odd-prime coefficient in Theorem 6.1. I checked the 78 displayed Gram entries and they are all nonzero, so the proof is correct; nevertheless, adding a sentence that the table was machine-checked, or including a short verification script, would make this step easier for readers to certify.","section":"Lemma 4.4"}],"recommendation":"minor_revision","confidential_remarks":"The paper appears mathematically sound. The only reason I did not recommend direct acceptance is the conspicuous superscript-flattening in the submitted text, particularly in Lemma 4.6, which makes a key quantity appear to be 2m+1 instead of 2^m+1; this is clearly a typesetting artifact but should be corrected. I also verified the p=3 finite check in Lemma 4.4, so the skeptic's main concern does not land."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Lecomte proves log_2 h(n) = n/2 + O(sqrt(n)(log(n+2))^3), so h(n)^{1/n} -> sqrt(2). That closes Erdős's problem at the exponential scale. The lower bound is the standard extraspecial 2-group construction; the upper bound is the new work. The proof architecture is sound: isoclinism reduces to finite groups, then a central factor series through P' turns the covering problem into a recursion on alternating forms, with two interaction-control mechanisms (nested anchors and exact centralization) that keep the clique credits honest. The Sylow and Fitting reductions are clean, and the 2-group case gives the sharp sqrt(n) error term.\n\nThe paper earns its keep. The lemmas are clearly organized and mostly independently checkable. I went through the finite reduction, the symplectic toolkit, the branch cover bound, and the interaction lemmas; they hang together. The citations are appropriate: Pyber, Neumann–Vaughan-Lee, Neumann, etc. The result is a real advance in extremal group theory.\n\nThe soft spot is exactly the one the stress-test flags: Lemma 4.4's p=3 case. The claim that F_3^6 contains 13 vectors pairwise nonorthogonal for the displayed alternating form is load-bearing. It gives kappa_3 = 2, hence the odd-prime coefficient (log_2 3)/4 < 1/2. If a single one of the 78 pairings vanished, the best clique credit would drop and the upper bound would break at linear order. The paper lists all 78 Gram entries but provides no machine check or code. I didn't find an error by hand, but I didn't verify all of them either. This is a finite, mechanical computation, so it's not a structural objection, but it deserves independent verification before the proof is accepted as final.\n\nMy overall verdict: the paper is a serious piece of work, and the central argument holds up. It should go to peer review. I'd ask the referee to verify the p=3 list (or require the author to attach a short script) and otherwise check the interaction analysis. If that one check passes, the result is solid and important.","headline":"Sharp base sqrt(2) for h(n) is likely right; the p=3 scalar-clique check is the one load-bearing hand computation.","tokens_in":15090,"tokens_out":2279,"would_cite":true,"duration_ms":20119,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":["20D60","20D15","05C15","05C69"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the worst-case number of abelian subgroups needed to cover a group with no pairwise noncommuting set larger than $n$ grows like $\\sqrt{2}^{\\,n}$, up to a sharply quantified error.","keywords":["noncommuting graph","abelian cover","finite p-groups","alternating bilinear forms","isoclinism","chromatic number","extremal combinatorics","nilpotent groups"],"falsifier":"A reader could compute the 78 upper-triangular symplectic pairings among the 13 vectors listed in Lemma 4.4; if any pairing is zero, the $p=3$ clique-credit constant collapses and the upper-bound proof breaks. Alternatively, an exhaustive search over 13-element subsets of the 728 nonzero vectors of $\\mathbb{F}_3^6$ for the displayed form would settle whether such a clique exists.","tokens_in":14089,"feed_emoji":"🧮","tokens_out":10382,"duration_ms":90921,"temperature":0.7,"pith_summary":"Every finite group has two natural measures of noncommutativity: the size $\\omega(G)$ of its largest pairwise noncommuting subset and the least number $a(G)$ of abelian subgroups needed to cover it. A long-standing quantitative problem asks how the worst-case covering number grows as a function of the allowed clique size. This paper proves that the extremal function satisfies $\\log_2 h(n)=n/2+O(\\sqrt{n}\\,(\\log(n+2))^3)$, and in particular $h(n)^{1/n}\\to\\sqrt{2}$: the worst case grows like $(1.414\\ldots)^n$. The lower bound comes from extraspecial $2$-groups, and the upper bound is obtained by reducing to finite $p$-groups, reading noncommutation as alternating bilinear forms, and then passing from arbitrary finite groups to a nilpotent subgroup at negligible cost. A sympathetic reader should care because the result settles the exponential scale of the problem, and it also shows that asymptotically extremal examples must be essentially $2$-groups.","feed_headline":"Worst-case abelian cover count grows like 1.414^n","feed_subtitle":"With clique number at most n, about 1.414^n abelian subgroups always suffice, and extraspecial 2-groups show this is optimal.","key_machinery":"The load-bearing object is a finite $p$-group $P$ together with a normal series $1=K_0<K_1<\\cdots<K_L=P'$ whose successive quotients have order $p$ and lie in the centre. At each level, commutation modulo $K_{L-j-1}$ induces a nondegenerate alternating bilinear form $\\phi_j$ on $V_j=A_j/R_j$, an $\\mathbb{F}_p$-vector space; its rank $\\rho_j$ is the unit of account. Two complementary constructions do the work: a spread cover by isotropic subspaces shows that the group can be covered by about $p^{\\rho_j/2}$ subgroups, while orthogonal-anchor composition builds pairwise nonorthogonal sets of size roughly $\\kappa_p\\rho_j$ that force a clique of that size. Balancing the two gives the per-rank coefficient $\\alpha_p=(\\log_2 p)/(2\\kappa_p)$, with $\\kappa_2=1$, $\\kappa_3=2$, $\\kappa_p=p/2$ for $p\\ge5$, so $\\alpha_p\\le1/2$ with equality only at $p=2$. Around this core, a nested-anchor estimate controls weak interaction between stages and an exact-centralization product argument controls strong interaction.","core_discovery":"The central claim is that the extremal exponential base is exactly $\\sqrt{2}$: for $h(n)=\\sup\\{a(G):\\omega(G)\\le n\\}$ one has $\\log_2 h(n)=n/2+O(\\sqrt{n}(\\log(n+2))^3)$. Extraspecial $2$-groups provide the matching lower bound, so the constant cannot be lowered. On the upper-bound side the proof decomposes a finite group along a central series of its derived subgroup; each step produces a nondegenerate alternating form over $\\mathbb{F}_p$, and the ratio between the logarithmic covering cost and the clique forced by one unit of symplectic rank is maximized at $p=2$. For every odd prime this ratio is strictly below $1/2$, which is why asymptotic extremality is confined to $2$-groups: any sequence of groups with $\\log_2 a(G_r)=\\omega(G_r)/2-o(\\omega(G_r))$ must eventually have exactly one nonabelian Sylow subgroup, a $2$-group.","pith_inferences":["Beyond the paper, the location of the extremal constant in $p=2$ suggests that every near-extremal group should contain a large extraspecial $2$-section; the paper proves concentration at the Sylow level but does not fully extract such a structural embedding.","Beyond the paper, the finite $\\mathbb{F}_3^6$ clique used to get $\\kappa_3=2$ may be one member of an infinite family of small cliques in alternating spaces over odd fields; finding such families could replace the slack odd-prime bound by the exact constant.","Beyond the paper, the same central-series/alternating-form scheme should apply to other covering parameters such as covers by centralizers, where an analogous symplectic-rank constant would govern the growth.","Beyond the paper, the quasipolynomial extension via the centralizer of the derived subgroup is likely not optimal; a sharper bound on $[G:C_G(G')]$ would shrink the $O(\\sqrt{n}(\\log n)^3)$ error and may improve the qualitative error term."],"forward_implications":["The worst-case covering number $h(n)$ satisfies $\\log_2 h(n)=n/2+O(\\sqrt{n}(\\log(n+2))^3)$, so $h(n)^{1/n}\\to\\sqrt{2}$.","The least possible index $\\iota(G)$ of an abelian subgroup obeys the same rate: $\\log_2 i(n)=n/2+o(n)$, so $i(n)^{1/n}\\to\\sqrt{2}$.","Any sequence of groups reaching the extremal rate must asymptotically have exactly one nonabelian Sylow subgroup, and that subgroup must be a $2$-group.","For noncommuting graphs, the sharp asymptotic chromatic-versus-clique growth is $\\sup\\{\\chi(\\Gamma_G):\\omega(\\Gamma_G)\\le n\\}=2^{n/2+o(n)}$.","Within nilpotent groups, if at least two Sylow subgroups are nonabelian the rate drops to at most $n/3+o(n)$, so extremality is essentially confined to one nonabelian $2$-primary factor."],"supporting_citations":[{"why":"Supplies the conjugacy-class bound used throughout and the earlier exponential bounds for $h(n)$ that this paper sharpens.","marker":"[23]"},{"why":"Provides the theorem that bounded noncommuting sets force a finite center quotient, and the covering lemma used for the abelian-subgroup index corollary.","marker":"[20]"},{"why":"Provides the stem-group construction that makes the isoclinism reduction to finite groups possible.","marker":"[21]"},{"why":"Provides the bounded-conjugacy-class bound on the size of the derived subgroup, used to control the length of the central series and the cost of passing to the nilpotent subgroup.","marker":"[24]"}],"fun_headline_variants":["Abelian cover worst-case grows like sqrt(2)^n","Extremal abelian cover base is sqrt(2)","Covering groups: sharp rate sqrt(2)^n","Worst-case covering: 2-groups give sqrt(2)^n","Optimal cover base exactly sqrt(2) for groups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The decisive load-bearing premise is a finite check in the $p=3$ case: the 13 listed vectors in $\\mathbb{F}_3^6$ are pairwise nonorthogonal for the displayed alternating form, and if a single pairing vanished the claimed constant $\\kappa_3=2$ would fail, leaving a gap in the upper bound.","fun_headline_variants_meta":{"raw":{"variants":["Abelian cover worst-case grows like sqrt(2)^n","Extremal abelian cover base is sqrt(2)","Covering groups: sharp rate sqrt(2)^n","Worst-case covering: 2-groups give sqrt(2)^n","Optimal cover base exactly sqrt(2) for groups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00095,"raw_usage":{"total_tokens":4079,"prompt_tokens":999,"completion_tokens":3080,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":615,"completion_tokens_details":{"reasoning_tokens":2994}},"tokens_in":615,"tokens_out":3080,"duration_ms":21824,"temperature":1.0,"reasoning_tokens":2994,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-27T19:22:02.367444+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader could compute the 78 upper-triangular symplectic pairings among the 13 vectors listed in Lemma 4.4; if any pairing is zero, the $p=3$ clique-credit constant collapses and the upper-bound proof breaks. Alternatively, an exhaustive search over 13-element subsets of the 728 nonzero vectors of $\\mathbb{F}_3^6$ for the displayed form would settle whether such a clique exists.","supporting_citations":[{"cited_title":"Pyber, The number of pairwise non-commuting elements and the index of the centre in a finite group,J","cited_arxiv_id":null,"evidence_quote":"Supplies the conjugacy-class bound used throughout and the earlier exponential bounds for $h(n)$ that this paper sharpens."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the theorem that bounded noncommuting sets force a finite center quotient, and the covering lemma used for the abelian-subgroup index corollary."},{"cited_title":"Hall, The classification of prime-power groups,J","cited_arxiv_id":null,"evidence_quote":"Provides the stem-group construction that makes the isoclinism reduction to finite groups possible."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the bounded-conjugacy-class bound on the size of the derived subgroup, used to control the length of the central series and the cost of passing to the nilpotent subgroup."}],"review_version":1}