{"id":"27038795-15a5-4af5-88a0-0cd3af9344df","arxiv_id":"2507.17986","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"The paper proposes heuristic prime-gap bounds of 180 and 8 by adding chaos and random matrix perturbations to Maynard's sieve, with no proof that the perturbations help.","lead":"A new heuristic framework combining chaotic maps and random matrix theory suggests that prime gaps are bounded by 180 unconditionally and by 8 under a partial Elliott-Halberstam assumption, but the arguments are explicitly non-rigorous. The methods extend Maynard's multidimensional sieve with a randomly expanded polytope and a Gaussian weight, yet the only numerical test performed shows no improvement in the sieve ratio.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed M'-gain from the chaotic polytope and RMT term is untested and contradicted by the paper's own F=1 experiment, so the improved gap bounds rest on an unsupported assumption.","rationale":"The paper is transparent that the bounds are heuristic, and it does provide a reproducible Monte Carlo volume expansion (Appendix B) showing |R'|/|R| > 1, plus prime-gap statistics consistent with small gaps. These support the plausibility of the framework but do not address the actual sieve-ratio gain. The decisive gap is Theorem 4.3: the claimed additive gains delta/2 and epsilon ln ln k are justified by a proof explicitly labeled hand-wavy, and the one numerical experiment that isolates the region change (Appendix A, F = 1) shows no improvement (-0.04%). Since the paper does not perform the promised re-optimization, the entirety of the improvements over Maynard's 246 rests on an unverified conjecture about the variational problem. Additionally, no theorem is stated or proved showing that the modified weight with support R' still satisfies the Maynard sieve estimates; the inference 'M' > m implies infinitely many m+1 primes' is assumed rather than derived. The internal arithmetic inconsistencies in Section 5 further undermine confidence. A concrete optimization test would settle whether the claimed M' > 3 for k ~ 40 actually holds; until then, REJECT is appropriate.","tokens_in":12065,"tokens_out":9292,"duration_ms":101345,"concrete_test":"Run the degree-5 symmetric-polynomial optimization of Section 5 for k = 40, delta = 0.3, epsilon = 0.1, computing sup I(F')/J(F') with F' = F + epsilon * product(Phi(t_j)) over R' (Definition 3.1, r = 3.9, 5 logistic iterates), by quasi-Monte Carlo integration. If the supremum is <= 3, the claimed M' > 3 and the derived gap bounds fail; if it exceeds 3, verify separately that the same F restricted to R reproduces M_base ~ 2.5, so the gain is attributable to the perturbation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the assertion (Theorem 4.3, Section 5) that the perturbed sieve ratio gains delta/2 + epsilon ln ln k over the unperturbed optimum. This is not derived; the proof is explicitly hand-wavy. The paper's own Monte Carlo experiment (Appendix A) shows that for F = 1 on R, moving to R' changes M by -0.04%, i.e., enlarging the region alone does not increase M. The paper hypothesizes that re-optimizing F would convert the extra volume into a gain, but no such optimization is carried out, so the predicted M' > 3 at k = 40 (Corollary 5.1) is untested. Moreover, no version of Maynard's sieve theorem is stated or proved for weights whose support is the chaotic region R'; the inference from M' > m to infinitely many m+1 primes is asserted as 'standard sieve arguments' without checking that the error-term analysis survives the enlarged support and the RMT factor. The internal arithmetic is also inconsistent (e.g., at k = 30 the text computes 0.85 + 0.15 + 0.12 = 1.12 but then reports M' ~ 1.97; at k = 40 the same perturbation is treated both as an additive increment and as a multiplicative exponent). Because the improvement is both unverified and partially contradicted by the only supplied experiment, the central claim does not hold up.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes an \"Enhanced Multidimensional Chaotic Heuristic Sieve\" (EMCHS) that modifies Maynard's multidimensional Selberg sieve by replacing the standard simplex R with a chaotically expanded region R' and adding a random-matrix-theoretic perturbation to the test function. The central claim is the perturbed sieve ratio satisfies M'(F') ~ (1/4) ln k + delta/2 + epsilon ln ln k (Theorem 4.3), which is then used to heuristically derive prime gap bounds of at most 180 unconditionally and at most 8 under a partial Elliott-Halberstam assumption. The paper is explicit that the gap bounds are heuristic and includes Monte Carlo volume experiments and prime-gap statistics as supporting evidence.","tokens_in":12397,"tokens_out":4248,"duration_ms":47541,"significance":"If the central formula and the subsequent gap derivation were actually established, the paper would offer a genuinely new heuristic connection between ergodic theory, random matrix theory, and sieve methods, with conditional bounds competitive with known results under stronger hypotheses. The manuscript has some strengths: it is unusually transparent about distinguishing proven components from heuristic ones, it provides executable Monte Carlo code, and it identifies a concrete mechanism (enlarged sieve support) that is interesting to consider. However, the central quantitative formula is not derived, the only direct numerical test of the mechanism in the paper (Appendix A) shows no improvement, and the numerical extrapolations in Sections 5 and 7 are internally inconsistent. As a result, the claimed gap bounds do not rest on any verified or logically derived foundation.","major_comments":[{"comment":"The perturbed sieve ratio formula M' ~ (1/4) ln k + delta/2 + epsilon ln ln k is the load-bearing claim of the paper, but its proof is explicitly described as a \"hand-wavy argument\" and does not derive the ratio I(F')/J(F') from the Selberg sieve sums. No quantitative connection is made between the volume enlargement of R' and the behavior of I and J, and the RMT adjustment is justified by a rough scaling argument rather than by computation. Since this formula is the basis for all subsequent bounds, the central claim is unsupported.","section":"Section 4, Theorem 4.3"},{"comment":"The paper's own Monte Carlo experiment (Appendix A) with F=1 shows that passing from R to R' changes M by -0.04%, i.e., the enlarged region alone does not increase the sieve ratio. The paper asserts that re-optimizing F would convert the additional volume into a gain, but no such optimization is performed anywhere in the manuscript. Consequently, the predicted M' > 3 at k=40 (Corollary 5.1) is untested, and the only direct numerical evidence for the proposed mechanism contradicts its effectiveness for the tested test function.","section":"Appendix A and Corollary 5.1"},{"comment":"The numerical extrapolation in Section 5 is internally inconsistent. For k=30, the text computes 0.85 + 0.15 + 0.12 = 1.12 and then reports M' ≈ 1.97, which corresponds to adding roughly 0.85 instead of 0.12 to the base value. For k=40, the perturbation is first treated as an additive increment (0.92 + 0.15 + 0.13 = 1.20, above base 1.0) and then as a multiplicative exponent e^(1.20-1.0) to obtain 3.0. These are different mathematical models, and neither is derived from Theorem 4.3; thus the claimed threshold M' > 3 is not well-defined.","section":"Section 5, Corollary 5.1"},{"comment":"The gap-bound derivation is ad hoc and numerically inconsistent. Section 7 first obtains H ≈ 56.5 for delta=0.3, epsilon=0.1, then for the unconditional case with delta=0 and epsilon=0.1 it computes H ≈ 163 and then arbitrarily states 180 as a conservative estimate; later the same section claims gap 8 and Conjecture 7.1 gives liminf(p_{n+1}-p_n) ≤ 11 under the same delta=0.3, epsilon=0.1. These values are mutually incompatible and the formula H ≈ k ln k / exp(2 delta - epsilon) is introduced as an ansatz without derivation or numerical support. The headline bounds of 180 and 8 are therefore not consequences of any consistent argument.","section":"Section 7, Conjecture 7.1"},{"comment":"The paper never states or proves a version of Maynard's sieve theorem for weights supported on the chaotic region R' and including the RMT factor xi(t). The implication from M'(F') > m to infinitely many m+1 primes is asserted via \"standard sieve arguments\" from Section 2, but that discussion applies only to the original polytope R and the unperturbed weight. The error-term analysis required for the enlarged, chaotically varying support and for the additional xi factor is not supplied, so even accepting Theorem 4.3, the final number-theoretic conclusion does not follow.","section":"Sections 2 and 4"}],"minor_comments":[{"comment":"The parameter tau is defined as theta/4 in Section 2 and as (1/2+delta)/4 in Section 3, but the relation between theta and delta is not made precise until Section 6, where theta = 1/2+delta is assumed; earlier sections should state this consistently.","section":"Sections 2 and 3"},{"comment":"The proof of Lemma 4.2 claims that the simplex {sum t_i <= 1+delta, 0 <= t_i <= tau} has volume at most (1+delta)^k/k! because tau >= 1+delta, but for the parameter values used in the paper, e.g. delta=0.3, tau=(1/2+delta)/4 = 0.2, so tau is far smaller than 1+delta. The coordinate constraints materially reduce the volume, and the stated bound is not justified.","section":"Lemma 4.2"},{"comment":"The abstract states that certain analytic components are rigorously proved, including bounding chaotic perturbations via ergodic theory, but Assumption 4.1 simply assumes the invariant measure for the logistic map at r=3.9 is close to the r=4 density and is bounded away from 0 and 1; no proof for r=3.9 is given.","section":"Assumption 4.1"},{"comment":"The prime-gap statistics up to 10^8 and up to 4 x 10^18 are consistent with standard heuristics and do not provide evidence for the specific liminf claims; the observation that many small gaps occur in a finite range is not quantitatively connected to the proposed formula M' ~ (1/4) ln k + delta/2 + epsilon ln ln k.","section":"Section 8"},{"comment":"The formula liminf(p_{n+1}-p_n) <= exp(2 delta - epsilon) * ln(e^2/delta) is introduced in the conjecture without a derivation matching the preceding text, and the notation ln(e^2/delta) is ambiguous; the heuristic justification does not reconcile the different gap values obtained in Section 7.","section":"Conjecture 7.1"}],"recommendation":"reject","confidential_remarks":"The rejection is based on the fact that the central formula (Theorem 4.3) is not derived, the only numerical test of the mechanism in Appendix A shows no improvement, and the extrapolations in Sections 5 and 7 are mutually inconsistent. These are load-bearing issues that cannot be repaired by local revisions within the manuscript's current scope; a convincing version would require an actual derivation or a successful numerical optimization of F on R' together with a consistent extension of the sieve theorem to the enlarged support."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague —\n\nYou should know two things about this paper up front. It is honestly labeled as heuristic—the author repeatedly says the new bounds are not proven. But the central mechanism that produces those bounds is not just unproven; the paper's own experiment shows it doing nothing, and the numerical path to the headline numbers contains arithmetic inconsistencies. The bounds 180 and 8 are effectively chosen by the free parameters.\n\nWhat is actually new: the idea of using a chaotic logistic-map perturbation and an RMT weight to enlarge the support of Maynard's sieve is novel, and the specific heuristic bounds (≤180 unconditionally, ≤8 under partial EHC) do not appear in prior literature. The paper is also transparent: it labels Theorem 4.3 as heuristic, admits the gap formula is an ansatz, and lists limitations. The volume-expansion Monte Carlo (|R'| ≈ 4.25×|R| in the toy case) is a real computation, and the code is included. The author clearly knows the difference between a heuristic and a proof, which is more than many speculative papers manage.\n\nThe soft spots are load-bearing. Theorem 4.3's proof is hand-wavy; the δ/2 and ε ln ln k gains are asserted with intuition, not derived. The only test of the core mechanism—Appendix A with F=1—shows M' unchanged (actually -0.04%). The paper's counter is that re-optimizing F would recover the gain, but no such optimization is carried out. So the claimed M'>3 at k=40 is an untested hypothesis, and the one data point we have points the other way. Section 5 also has an arithmetic inconsistency: for k=30, 0.85+0.15+0.12 is 1.12, but the text reports M'≈1.97; for k=40, the perturbation is both added and exponentiated (2.5×e^{1.20−1.0}). That makes it hard to take the numerical route to 180 and 8 seriously. The gap ansatz H≈k ln k/exp(2δ−ε) is invented, not derived, and the parameters δ=0.3, ε=0.1 are chosen to hit round numbers. The inference from M'>m to infinitely many primes is stated as 'standard sieve arguments,' but no version of Maynard's theorem is proved for the chaotic region R' with the RMT factor; the enlarged support could break the error-term analysis.\n\nThe numerical evidence on prime gaps up to 1e8 is irrelevant to a liminf claim—small gaps occurring often is well known and doesn't test the framework.\n\nSo I agree with the reader: reject, high confidence. This is not a paper for a serious referee. It might be interesting as a speculative note for people working on heuristics, but it does not meet the bar for an analytic number theory journal. If the author actually performed the optimization for the perturbed region and showed M' genuinely exceeds the unperturbed optimum, the story would change. As it stands, the central claim rests on an assumption that the paper itself fails to support.","headline":"An honest heuristic proposal whose load-bearing sieve-ratio gain is neither derived nor demonstrated—and its own Monte Carlo shows no gain—so the headline bounds of 180 and 8 rest on an unverified assumption.","tokens_in":12889,"tokens_out":7163,"would_cite":false,"duration_ms":67812,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11N05","11N35","11M50","37D45"],"pacs":[],"model":"deepseek-v4-flash","headline":"Heuristic bound on prime gaps: a sieve with chaotic polytope expansion and random-matrix weights claims gaps of at most 180 unconditionally and 8 under a partial Elliott–Halberstam hypothesis.","keywords":["prime gaps","Maynard sieve","Selberg sieve","Elliott–Halberstam conjecture","random matrix theory","chaotic dynamical system","GUE","heuristic number theory"],"falsifier":"A decisive test is to run the Appendix A Monte Carlo integrator for an optimized degree-5 or degree-6 symmetric polynomial on $R'$ at $k=40$, $\\delta=0.3$, $\\epsilon=0.1$. If the maximal sieve ratio remains near the unperturbed value $\\approx 2.5$ rather than exceeding 3, Theorem 4.3's additive gain is not realized and the claimed gaps 180 and 8 do not follow.","tokens_in":11826,"feed_emoji":"🔢","tokens_out":9159,"duration_ms":85340,"temperature":0.7,"pith_summary":"This paper proposes an extension of Maynard's multidimensional sieve in which the support region is chaotically expanded and the test function is given a random-matrix-theoretic bias. It claims that the sieve ratio then scales as $M'(F') \\sim \\frac{1}{4}\\ln k + \\frac{\\delta}{2} + \\epsilon\\ln\\ln k$, and that with $\\delta=0.3$ and $\\epsilon=0.1$ the ratio exceeds 3 by dimension $k\\approx 40$. That crossing would mean infinitely many admissible 40-tuples contain at least four primes, which the paper translates into heuristic prime-gap bounds of at most 180 unconditionally and at most 8 under a partial Elliott–Halberstam conjecture with $\\theta=0.8$. The author is explicit that these bounds are heuristic: the analytic volume and ergodic bounds are proven, but the additive gain in the sieve ratio and the translation to gap lengths rest on unproved modeling assumptions.","feed_headline":"Heuristic sieve predicts prime gaps of 180, or 8 with EHC","feed_subtitle":"Adding chaos and random-matrix weights to Maynard's sieve is claimed to push past the four-prime threshold.","key_machinery":"The engine is the pair of mechanisms defining $R'$ and $\\xi$. The perturbed polytope $R'=\\{t\\in[0,\\tau]^k:\\sum t_i\\le 1+\\delta\\chi(\\sum t_i)\\}$, with $\\tau=(1/2+\\delta)/4$ and $\\chi$ the five-fold logistic-map iterate at parameter $r=3.9$, expands the feasible region for the test function, and the random-matrix weight $\\xi(t)=\\prod_{j=1}^k\\Phi(t_j)$ biases the weight toward coordinates away from the singular edges of the $J$-integral. The quantitative claim carried by this machinery is Theorem 4.3's scaling law, which is what converts an assumed extra distribution range $\\delta$ and a GUE-like correlation strength $\\epsilon$ into an additive gain in $M$.","core_discovery":"The central claim is that two perturbations of Maynard's sieve raise the optimal ratio $M(F)=I(F)/J(F)$ by an additive amount $\\delta/2 + \\epsilon\\ln\\ln k$ beyond the baseline $\\frac{1}{4}\\ln k$. The first perturbation enlarges the simplex $R=\\{t\\in[0,\\tau]^k:\\sum t_i\\le 1\\}$ to $R'=\\{t\\in[0,\\tau]^k:\\sum t_i\\le 1+\\delta\\chi(\\sum t_i)\\}$, where $\\chi$ is a logistic-map iterate, and the second replaces $F$ by $F+\\epsilon\\xi$ with $\\xi(t)=\\prod_{j=1}^k\\Phi(t_j)$ built from the normal CDF as a stand-in for GUE spacing statistics. Theorem 4.3 encodes the resulting heuristic formula $M'(F')\\sim\\frac{1}{4}\\ln k+\\frac{\\delta}{2}+\\epsilon\\ln\\ln k$. The paper then uses the standard sieve criterion $M>m \\Rightarrow$ infinitely many $m+1$ primes in an admissible $k$-tuple to claim $M'>3$ at $k\\approx 40$, yielding 4-prime blocks and hence gaps of at most 180 unconditionally and at most 8 under a partial Elliott–Halberstam assumption with $\\theta=0.8$.","pith_inferences":["If the volume gain is the true source of the improvement, then any deterministic expansion of the simplex to $\\sum t_i\\le 1+\\delta/2$ would already push $M$ past 3 at $k\\approx 40$; this simpler mechanism is testable independently of the chaos and RMT machinery.","The $\\epsilon\\ln\\ln k$ term is the least grounded part of Theorem 4.3: replacing the normal-CDF product by a genuine $k$-point GUE correlation model would show whether the claimed slow gain survives a more faithful random-matrix input.","A practical extension would be to run the Appendix A Monte Carlo integrator for an optimized degree-5 polynomial on $R'$ at $k=30$ and $40$; that computation, which the paper does not perform, is the cheapest decisive check of the whole framework.","If the conditional gap of 8 is real, it suggests the same bound currently requiring strong distribution assumptions might be reproduced by an averaging procedure over chaotic orbits, a transfer the paper hints at but does not establish."],"forward_implications":["If the claimed values are correct, a 40-dimensional optimized sieve would exceed $M'>3$, so infinitely many admissible patterns of 40 integers would contain four primes simultaneously.","From that, the paper derives an unconditional heuristic bound $\\liminf(p_{n+1}-p_n)\\le 180$, beating the proven Polymath8b bound of 246.","Under a partial Elliott–Halberstam assumption with $\\theta=0.8$, the same framework would give at most 8, in line with the prime quadruplet pattern $\\{0,2,6,8\\}$.","The Monte Carlo volume test shows $|R'|/|R|$ can be large (about 4.25 in the toy example), so there is real extra integration volume available if a re-optimized $F$ can exploit it.","The paper's own numerical check with $F\\equiv 1$ shows the perturbation alone changes $M$ by only $-0.04\\%$, meaning the predicted gain depends entirely on re-optimizing $F$."],"supporting_citations":[{"why":"It introduces the weight method and the ratio $M(F)$ whose threshold $M>m$ implies $m+1$ primes in a tuple.","marker":"[2]"},{"why":"It supplies the multidimensional sieve framework, the polytope $R$, and the baseline asymptotic $M(F)\\sim\\frac{1}{4}\\ln k$ that Theorem 4.3 extends.","marker":"[3]"},{"why":"It provides the proven bound 246 that the paper's heuristic 180 is meant to surpass.","marker":"[4]"},{"why":"It supplies the invariant density of the logistic map at $r=4$, which Assumption 4.1 uses to justify the mean value $\\mathbb{E}[\\chi]=1/2$.","marker":"[7]"},{"why":"It establishes the GUE pair-correlation heuristic for zeta zeros that motivates the RMT factor $\\xi$.","marker":"[6]"},{"why":"It supplies the partial Elliott–Halberstam framework whose exponent $\\theta$ the paper converts into the parameter $\\delta$.","marker":"[1]"},{"why":"It provides the empirical maximal gap $1476$ up to $4\\times10^{18}$, used as a consistency check that gaps below 180 are common.","marker":"[8]"}],"fun_headline_variants":["Heuristic chaotic sieve hints prime gaps down to 8","Prime gaps 180 heuristically, or 8 with EHC","Chaos and random matrices suggest prime gap 8","Maynard's sieve plus chaos and RMT: gap 8","Heuristic sieve claims prime gaps of 8 under EHC"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that re-optimizing the test function $F$ inside the chaotically enlarged region $R'$ actually recovers the claimed additive gain in the sieve ratio, because the paper's only numerical check of the perturbation, with $F\\equiv 1$, shows essentially no change in $M$ ($-0.04\\%$).","fun_headline_variants_meta":{"raw":{"variants":["Heuristic chaotic sieve hints prime gaps down to 8","Prime gaps 180 heuristically, or 8 with EHC","Chaos and random matrices suggest prime gap 8","Maynard's sieve plus chaos and RMT: gap 8","Heuristic sieve claims prime gaps of 8 under EHC"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002076,"raw_usage":{"total_tokens":8103,"prompt_tokens":1003,"completion_tokens":7100,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":619,"completion_tokens_details":{"reasoning_tokens":7014}},"tokens_in":619,"tokens_out":7100,"duration_ms":54526,"temperature":1.0,"reasoning_tokens":7014,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T14:38:10.603505+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test is to run the Appendix A Monte Carlo integrator for an optimized degree-5 or degree-6 symmetric polynomial on $R'$ at $k=40$, $\\delta=0.3$, $\\epsilon=0.1$. If the maximal sieve ratio remains near the unperturbed value $\\approx 2.5$ rather than exceeding 3, Theorem 4.3's additive gain is not realized and the claimed gaps 180 and 8 do not follow.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It introduces the weight method and the ratio $M(F)$ whose threshold $M>m$ implies $m+1$ primes in a tuple."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the multidimensional sieve framework, the polytope $R$, and the baseline asymptotic $M(F)\\sim\\frac{1}{4}\\ln k$ that Theorem 4.3 extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It provides the proven bound 246 that the paper's heuristic 180 is meant to surpass."},{"cited_title":"Collet and J.-P","cited_arxiv_id":null,"evidence_quote":"It supplies the invariant density of the logistic map at $r=4$, which Assumption 4.1 uses to justify the mean value $\\mathbb{E}[\\chi]=1/2$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It establishes the GUE pair-correlation heuristic for zeta zeros that motivates the RMT factor $\\xi$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the partial Elliott–Halberstam framework whose exponent $\\theta$ the paper converts into the parameter $\\delta$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It provides the empirical maximal gap $1476$ up to $4\\times10^{18}$, used as a consistency check that gaps below 180 are common."}],"review_version":1}