{"id":"c81b3a5a-c1ad-4e8d-9466-0a541a204c9b","arxiv_id":"2608.07790","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For critical K=1 Boolean networks, attractor lengths are governed by the order of the permutation induced by feedback loops, yielding typical maximum lengths exp[(1/2)ln^2 N], extremal lengths exp[sqrt(N ln N)], and mean length exp[N^(1/3)].","lead":"A theoretical paper claims that the long-term behavior of critical Boolean networks with one connection per gene is controlled by a single permutation of the network's feedback loops. Using classical results on random permutations, it derives new scaling laws for attractor lengths, including a mean length that grows like exp[N^(1/3)].","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 2 is false (n=2, ell=4 yields negative count), so Theorem 1's lower bound \\bar A >= ord(sigma)/4 is unproven; the exp[N^{1/3}] mean-length result lacks support.","rationale":"The reader identifies the same load-bearing gap: the lower bound \\bar A >= ord(sigma)/4 is essential to the central mean-length claim, and its proof relies on Lemma 2, which is derived from an incorrect formula for negative-loop cycle counts. I independently verified the concrete failure: for n=2, Eq. (15) yields a negative count for ell=4, while direct enumeration of the negative 2-loop gives exactly one 4-cycle; and Lemma 2 fails at k=2 (LHS -2 vs RHS 4). The Möbius-inversion coefficient in the proof of Lemma 2 is also miscomputed for powers of 2 greater than 1. These are not mere typos: the identity is structurally false, so Step 1 of Theorem 1 does not go through. Since the mean-length result has no independent support in the paper (the Goh–Schmutz asymptotics only control order(sigma), not attractor lengths without the lower bound), the central claim is unproven. The typical and extremal bounds for maximum attractor length rely on different, valid arguments and may well be correct, but the headline contribution is not. I recommend REJECT: as a preprint, the proof of the main theorem contains a clearly false lemma, and the claimed formal verification is not verifiable. A corrected proof might restore the result, and the paper could then be reconsidered, but in its present form the central claim is not supported.","tokens_in":9174,"tokens_out":5711,"duration_ms":49696,"concrete_test":"Evaluate Lemma 2 and Eq. (15) for n=2, k=2: compute \\bar c(4)*4 from Eq. (15) and compare with the direct count of 4-cycles in a single negative 2-loop (enumerating all 4 states confirms exactly one 4-cycle). If the formula gives -2 instead of 4, the lemma and the proof of Theorem 1 are refuted. To test the theorem itself, brute-force enumerate all critical K=1 networks for N up to 8 (or 10), compute \\bar A/ord(sigma) for every realization, and check whether the minimum is at least 1/4; a counterexample would decisively refute the claim, while absence would still leave the proof needing repair.","verdict_should_be":"REJECT","load_bearing_attack":"The mean-attractor-length scaling exp[N^{1/3+o(1)}] rests on the lower bound \\bar A >= ord(sigma)/4 (Theorem 1). Without this bound, only the upper bound \\bar A <= 2 ord(sigma) is available, and the expectation could in principle be much smaller. Theorem 1's proof first reduces to positive loops (Step 1), and that reduction depends on Lemma 2, which claims \\bar c(2k)*2k = sum_{d|k, k/d a power of 2} d*c(d). This lemma is false. For a negative loop of length n=2, Eq. (15) gives \\bar c(4) = (1/4)*(mu(2)*2^{4/4}) = -1/2, but direct enumeration of the negative 2-loop (e.g., 00->10->11->01->00) shows there is exactly one 4-cycle, so \\bar c(4)=1. The error propagates: taking k=2 in Lemma 2 gives LHS = -2 while RHS = c(1) + 2 c(2) = 2 + 2*1 = 4. In the proof of Lemma 2, the asserted Möbius coefficient is wrong: for k/d' = 2^r, the coefficient is 1 if r=0, -1 if r=1, and 0 if r>=2, not '1 for any power of 2.' Because Lemma 2 is false, the inequality in Eq. (19) is not established, so Step 1 fails and Theorem 1 is unproven. The paper also claims formal verification via the Aristotle API, but no artifact or certificate is provided, so this does not rescue the proof. The typical and extremal maximum-attractor-length results (Erdős–Turán and Landau) appear sound, but the headline mean-length claim is unsupported as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies critical K=1 Boolean networks and claims that their attractor lengths are governed by the order of a permutation induced by the network's feedback loops. Using classical results of Erdős–Turán, Landau, and Goh–Schmutz, it asserts that almost all networks have maximum attractor length at most exp[(1/2)ln^2 N], that some networks attain exp[Theta(sqrt(N ln N))], and that the mean attractor length scales as exp[N^{1/3+o(1)}]. The main mathematical content is in the appendix, where Theorem 1 is intended to prove the lower bound \\bar A >= ord(sigma)/4 that connects the mean attractor length to the permutation order.","tokens_in":9528,"tokens_out":7801,"duration_ms":68218,"significance":"If the results were correct, the paper would provide a clean bridge between Boolean network dynamics and permutation theory, resolving a long-standing scaling question for K=1 Kauffman networks. The typical and extremal maximum-length statements are plausibly sound: they follow from well-known facts about random permutations, conditional on the standard result that attractor lengths divide 2 ord(sigma). The conceptual reformulation in terms of the induced permutation is attractive and could be valuable. However, the headline mean-length scaling exp[N^{1/3}] is not established: its proof depends on a lower-bound theorem whose key lemma is false, and a numerical claim in the appendicial proof is also contradicted by direct computation. The paper provides no machine-checked artifact for the claimed formal verification, and the verification claim is inconsistent with the false lemma.","major_comments":[{"comment":"Eq. (15) gives a negative number of attractors for a negative 2-loop at ell=4. For n=2, ell=4, the formula yields \\bar c(4) = (1/4) sum_{d|2, 2/d odd} mu(d) 2^{4/(2d)} = (1/4)(-2) = -1/2, whereas direct enumeration of the negative 2-loop (00->10->11->01->00) shows exactly one 4-cycle. Consequently Lemma 2 is false; for k=2 the left-hand side \\bar c(4)*4 equals -2 while the right-hand side is c(1)+2c(2)=4 under the standard reading of Eq. (17). This invalidates the use of Lemma 2 in Eq. (19).","section":"Appendix, Eq. (15) and Lemma 2"},{"comment":"The Möbius-inversion coefficient in the proof of Lemma 2 is miscomputed. The proof claims that the coefficient of c(d')d' is 1 whenever k/d' is a power of 2 and 0 otherwise. For r = k/d' = 2^s, the coefficient is sum_{e|r, r/e odd} mu(e), which equals 1 for s=0, -1 for s=1, and 0 for s>=2. This is not a minor typo: it produces exactly the counterexample to the lemma described in the previous comment.","section":"Lemma 2 proof"},{"comment":"Because Lemma 2 is false, Step 1 of Theorem 1 fails, so the lower bound \\bar A >= ord(sigma)/4 in Eq. (3) is unproven. This bound is load-bearing: the mean-attractor-length derivation in Eqs. (11)-(12) requires both the upper bound \\bar A <= 2 ord(sigma) and the lower bound \\bar A >= ord(sigma)/4. Without the lower bound, the expected value of \\bar A could in principle be much smaller than exp[N^{1/3+o(1)}]. The headline mean-length result is therefore unsupported as written.","section":"Appendix, Theorem 1 and Eqs. (3), (11)-(12)"},{"comment":"The claimed numerical value of the infinite product R_min is incorrect. The factors are p/(2p-1), whose product over all primes converges to 0; already the partial product over 2,3,5,7 equals 14/143 approximately 0.098, below 1/4. Thus the assertion that \"a numerical computation gives R_min ~ 0.32 > 1/4\" is false, and the Step 3 lower-bound argument fails for an additional reason even after the positive-loop reduction.","section":"Appendix, Step 3 and Eqs. (28)-(29)"}],"minor_comments":[{"comment":"The displayed inequality has an extra \"-1\" on the right-hand side: for j=0 it reads gcd(0,ab) <= gcd(0,a)+gcd(0,b)+ab-a-b-1, i.e. ab <= ab-1, which is false. The subsequent line uses the version without the \"-1\", so this appears to be a typographical slip that should be corrected.","section":"Eq. (23)"},{"comment":"The manuscript states that Lemma 2 and Lemma 3 were formally verified using the Aristotle API [26], but no artifact, certificate, or proof script is provided. Since Lemma 2 is false, this verification claim cannot be credited in its current form.","section":"Appendix, formal verification claim"},{"comment":"There are several typographical and grammatical errors, including \"simple alebraic manifpulation\" in the proof of Step 2 and \"Kjr\" in reference [4]; these should be corrected in a revision.","section":"Throughout"},{"comment":"The condition \"2n/ell odd\" in Eq. (15) is typeset ambiguously; please clarify the scope of the \"0 otherwise\" clause and explicitly define the support of \\bar c(ell).","section":"Eq. (15)"}],"recommendation":"reject","confidential_remarks":"For the editor: the typical and extremal maximum-attractor-length statements may be salvageable and could potentially form a shorter paper. However, the manuscript's abstract-level claim about the mean attractor length depends on a demonstrably false lemma and an erroneous numerical product computation, so the central contribution is not reliable as submitted. I therefore recommend rejection rather than major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: the paper's central idea—encoding the long-term dynamics of critical K=1 Boolean networks in the permutation induced by the feedback loops—is clean and correct as far as it goes. The typical and extremal maximum-attractor-length results are straightforward corollaries of Erdős–Turán and Landau, and they are assembled correctly. The new claimed result, the exp[N^{1/3}] mean attractor length, is not supported as written.\n\nWhat is actually new: the explicit lower bound A_bar >= ord(sigma)/4 (Theorem 1) that bridges permutation order and the mean attractor length. The typical and extremal bounds follow immediately from known theorems, as the authors acknowledge. The averaging over the number of relevant nodes using the Rayleigh distribution and Laplace's method is also correct in outline.\n\nThe soft spots are load-bearing. Lemma 2 is false. For a negative 2-loop, Eq. (15) gives a negative count for the 4-cycle, while direct enumeration gives exactly one. The Möbius inversion step in the lemma's proof is also wrong: the coefficient of c(d') is 1 for r=0, -1 for r=1, and 0 for r>=2 when k/d' is a power of 2, not uniformly 1. Since Step 1 of Theorem 1 relies on Lemma 2, the reduction to positive loops fails, and the lower bound is unproven. There is a second error in Step 2: inequality (23) fails at j=0, where gcd(0,ab)=ab but the right-hand side is ab-1. The claim that the lemmas were formally verified via the Aristotle API is not accompanied by any verifiable artifact, so it does not rescue the proof. The typical and extremal results do not depend on Theorem 1 and appear sound.\n\nMy overall read: the paper is not incoherent; the permutation reformulation is genuinely useful and the typical/extremal part is correct. But the headline mean-length scaling is the only really new result, and it rests on a false lemma and a false inequality. This is a major revision: the authors need to fix or replace Lemma 2, check Step 2, and provide a verifiable formalization if they claim machine-checked proofs. If the lower bound cannot be established, the mean-length claim should be stated as a conjecture.\n\nWho this is for: researchers working on random Boolean networks, Kauffman models, and the combinatorics of permutations. The typical/extremal part is worth teaching; the mean-length part will need scrutiny.\n\nRecommendation: send to a serious referee. The problem is important and the paper has a plausible core, but it is not acceptable as-is.","headline":"The permutation reformulation is clean and the typical/extremal max-length bounds are correct, but the mean-length scaling rests on a false lemma and is unsupported as written.","tokens_in":10136,"tokens_out":7573,"would_cite":false,"duration_ms":59344,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05A05","37B15","60C05","92C42"],"pacs":[],"model":"deepseek-v4-flash","headline":"The length of an attractor in a critical Boolean network is governed by the order of a single permutation induced by its feedback loops.","keywords":["Boolean networks","critical K=1","attractor lengths","permutation order","random permutations","Kauffman networks","feedback loops","mean attractor length"],"falsifier":"Evaluate the identity labelled Lemma 2 at k=2: formula (15) yields $\\bar c(4)=-1/2$, so $\\bar c(4)\\cdot 4 = -2$, whereas the right-hand side summed over divisors of 2 equals $c(1)+2c(2)=4$. The claimed equality fails, and the Möbius inversion in its proof assigns coefficient $+1$ to every power-of-two quotient when the true coefficient is $-1$ for quotient 2. Checking the theorem's reduction for a single negative 2-loop therefore decides whether the lower bound $\\mathrm{ord}(\\sigma)/4 \\le \\bar A$ is valid.","tokens_in":8936,"feed_emoji":"🔄","tokens_out":10463,"duration_ms":83345,"temperature":0.7,"pith_summary":"This paper claims that for critical Boolean networks of connectivity one, all asymptotic properties of long-term dynamics are determined by a single permutation built from the network's feedback loops. The paper shows that the mean attractor length of such a network is within a constant factor of the order of this permutation, so the distribution of attractor lengths becomes a question about the arithmetic of random permutations. Combining standard results on the order of random permutations, the paper concludes that almost all networks have maximum attractor length at most $\\exp[\\tfrac{1}{2}\\ln^2 N]$, some networks reach $\\exp[\\Theta(\\sqrt{N\\ln N})]$, and the average over networks of the mean attractor length scales as $\\exp[N^{1/3+o(1)}]$. A reader should care because this connects Boolean network dynamics with combinatorics and number theory, and because it shows that rare networks with exceptionally long attractors dominate ensemble averages.","feed_headline":"Permutation order sets Boolean attractor lengths","feed_subtitle":"Almost all K=1 networks stay sub-exponential; rare ones reach exp(√(N ln N)); the mean is exp(N^1/3).","key_machinery":"The machinery is the induced permutation $\\sigma$ on the $m$ relevant nodes of the network: each relevant node maps to its unique successor, so the cycles of $\\sigma$ correspond exactly to feedback loops, and its order $\\mathrm{ord}(\\sigma)$ is the least common multiple of the loop lengths. The load-bearing inequalities $\\mathrm{ord}(\\sigma)/4 \\le \\bar A \\le 2\\,\\mathrm{ord}(\\sigma)$ turn every attractor-length question into a permutation-order question; Theorem 1 proves the lower bound by reducing to positive loops, splitting composite loop lengths into distinct prime powers, and applying a factorization together with an infinite product bound. The asymptotics then come from known permutation-order results applied to $\\sigma$ under the relevant-node count distribution $P(m)\\sim (m/N)e^{-m^2/2N}$.","core_discovery":"The central claim is that the permutation $\\sigma$ induced by the feedback loops of a critical $K=1$ Boolean network is a faithful invariant for long-term dynamics: its cycle lengths are exactly the loop lengths, every attractor length divides $2\\,\\mathrm{ord}(\\sigma)$, and the mean attractor length $\\bar A$ satisfies $\\mathrm{ord}(\\sigma)/4 \\le \\bar A \\le 2\\,\\mathrm{ord}(\\sigma)$. The paper then proves that this invariant yields three asymptotic statements: almost all networks have sub-exponential maximum attractor lengths, extremal networks attain lengths of order $\\exp[\\Theta(\\sqrt{N\\ln N})]$, and the expected mean attractor length over network realizations scales as $\\exp[N^{1/3+o(1)}]$, dominated by rare high-order permutations.","pith_inferences":["If the permutation encoding is as faithful as claimed, the full finite-size distribution of attractor lengths should track the distribution of cycle structures of a random permutation conditioned on the relevant-node count; this could be checked by exact enumeration for moderate $N$.","Because $\\bar A$ is within a constant factor of $\\mathrm{ord}(\\sigma)$, one can estimate a network's mean attractor length directly from the lcm of its feedback-loop lengths without simulating the Boolean dynamics, offering a fast structural diagnostic for Boolean models.","The dominance of rare networks in the mean suggests that ensemble-level statements about Boolean network dynamics depend sensitively on the random-network distribution; conditioning on different loop-sign or loop-length statistics could shift the mean scaling exponent."],"forward_implications":["Almost every critical $K=1$ network of $N$ nodes has all attractor lengths at most $\\exp[\\tfrac{1}{2}\\ln^2 N]$, so typical long-term behavior is sub-exponential in $N$.","There exist critical $K=1$ networks whose longest attractor grows as $\\exp[\\Theta(\\sqrt{N\\ln N})]$, so network-to-network variation in attractor length is enormous.","The expected mean attractor length over network realizations scales as $\\exp[N^{1/3+o(1)}]$, meaning a handful of rare networks with very high permutation order dominate the ensemble average; simulations that sample typical networks will underestimate it.","Up to constant factors the number of attractors is $2^m/\\mathrm{ord}(\\sigma)$, so the same permutation controls both attractor counting and attractor length.","In critical $K=2$ networks the dynamically active core reduces asymptotically to a $K=1$ structure, so the permutation-based approach is a candidate route to higher-connectivity attractor statistics."],"supporting_citations":[{"why":"Exact solution of the K=1 model; establishes that attractor lengths divide twice the lcm of loop lengths, the base of the permutation reformulation.","marker":"[4]"},{"why":"Prior bound on number and length of attractors in critical K=1 networks; the new results improve on this benchmark.","marker":"[5]"},{"why":"Supplies the distribution of relevant-node counts P(m) used to average the mean attractor length over networks.","marker":"[7]"},{"why":"Supplies the relevant-node count formula and the reduction of the critical K=2 active core to an effective K=1 structure.","marker":"[15]"},{"why":"Supplies the attractor-count formulas for positive and negative feedback loops used in the proof of the lower bound.","marker":"[17]"},{"why":"Supplies cycle-count and feedback-loop theory underlying the formulas for attractor counts of positive and negative circuits.","marker":"[18]"},{"why":"Provides the typical-order bound for random permutations that gives the almost-all sub-exponential attractor bound.","marker":"[19]"},{"why":"Provides the maximal-order asymptotics for permutations that give the extremal attractor length.","marker":"[23]"},{"why":"Provides the expected-order asymptotics for random permutations that feed the mean attractor length calculation.","marker":"[24]"}],"fun_headline_variants":["Permutation order sets all attractor lengths in K=1 networks","Rare networks have huge attractors, dominating the average","Boolean network attractors scale via permutation order","Feedback loops' permutation dictates long-term dynamics","Rare permutations push average attractor length to exp(N^(1/3))"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument hinges on the claim that the average time before a network repeats is always at least one quarter of the least common multiple of its feedback-loop lengths; if that comparison is wrong, the reported average attractor length does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Permutation order sets all attractor lengths in K=1 networks","Rare networks have huge attractors, dominating the average","Boolean network attractors scale via permutation order","Feedback loops' permutation dictates long-term dynamics","Rare permutations push average attractor length to exp(N^(1/3))"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001385,"raw_usage":{"total_tokens":5611,"prompt_tokens":953,"completion_tokens":4658,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":569,"completion_tokens_details":{"reasoning_tokens":4576}},"tokens_in":569,"tokens_out":4658,"duration_ms":31585,"temperature":1.0,"reasoning_tokens":4576,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T04:15:23.601970+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the identity labelled Lemma 2 at k=2: formula (15) yields $\\bar c(4)=-1/2$, so $\\bar c(4)\\cdot 4 = -2$, whereas the right-hand side summed over divisors of 2 equals $c(1)+2c(2)=4$. The claimed equality fails, and the Möbius inversion in its proof assigns coefficient $+1$ to every power-of-two quotient when the true coefficient is $-1$ for quotient 2. Checking the theorem's reduction for a single negative 2-loop therefore decides whether the lower bound $\\mathrm{ord}(\\sigma)/4 \\le \\bar A$ is valid.","supporting_citations":[{"cited_title":"Exact solution of kauffman’s model with connectivity one","cited_arxiv_id":null,"evidence_quote":"Exact solution of the K=1 model; establishes that attractor lengths divide twice the lcm of loop lengths, the base of the permutation reformulation."},{"cited_title":"Number and length of attractors in a critical kauffman model with connectivity one.Physical review letters, 94(8):088701, 2005","cited_arxiv_id":null,"evidence_quote":"Prior bound on number and length of attractors in critical K=1 networks; the new results improve on this benchmark."},{"cited_title":"Number of attractors in the critical kauffman model is exponential","cited_arxiv_id":null,"evidence_quote":"Supplies the distribution of relevant-node counts P(m) used to average the mean attractor length over networks."},{"cited_title":"Random boolean networks.Reviews of nonlinear dynamics and complexity, pages 69–110, 2008","cited_arxiv_id":null,"evidence_quote":"Supplies the relevant-node count formula and the reduction of the critical K=2 active core to an effective K=1 structure."},{"cited_title":"On the number of attractors of posi- tive and negative boolean automata circuits","cited_arxiv_id":null,"evidence_quote":"Supplies the attractor-count formulas for positive and negative feedback loops used in the proof of the lower bound."},{"cited_title":"World Scientific, 2017","cited_arxiv_id":null,"evidence_quote":"Supplies cycle-count and feedback-loop theory underlying the formulas for attractor counts of positive and negative circuits."},{"cited_title":"On some problems of a statistical group-theory","cited_arxiv_id":null,"evidence_quote":"Provides the typical-order bound for random permutations that gives the almost-all sub-exponential attractor bound."},{"cited_title":"Über die maximalordnung der permutationen gegebenen grades.Archiv der Math","cited_arxiv_id":null,"evidence_quote":"Provides the maximal-order asymptotics for permutations that give the extremal attractor length."},{"cited_title":"The expected order of a random permutation.Bulletin of the London Mathematical Society, 23(1):34–42, 1991","cited_arxiv_id":null,"evidence_quote":"Provides the expected-order asymptotics for random permutations that feed the mean attractor length calculation."}],"review_version":1}