{"id":"5fd76e72-68fc-41c5-9f82-d7c570f7e891","arxiv_id":"2603.14116","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Zaremba's conjecture holds for all large primes: an absolute M exists so every large prime p admits a coprime a whose continued-fraction partial quotients are all ≤ M, with quantitative lower bounds on the number of such a.","lead":"The paper proves Zaremba's conjecture for every sufficiently large prime (and many composite) denominators q: there is an absolute constant M such that some a coprime to q has all continued-fraction partial quotients of a/q at most M. It also supplies nearly optimal counts of such a and improved bounds on the sum of partial quotients, with direct consequences for discrepancy and combinatorial enumeration.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified beyond the already-flagged non-effective expansion constants.","rationale":"The reader correctly isolates the non-effective spectral-gap constant κ as the sole soft point and correctly classifies it as a limitation of the method rather than a flaw in the logic. The new critical-denominator independence (Section 4) and the middle-interval repulsion (Proposition 32) are self-contained once the expansion black box is granted; they contain no additional unstated assumptions that would collapse for the generating sets that actually arise. Because the paper already states that M is large and that the 1/9 barrier is far from optimal, no further downgrade of the ACCEPT verdict is warranted. The concrete test above would merely quantify how much of the size of M is forced by the present expansion technology versus the Diophantine core.","tokens_in":37261,"tokens_out":488,"duration_ms":4986,"concrete_test":"Independently re-derive the size of A ∩ A^{-1} for A = Z_M(t) ∩ Z_{M*}(tH) under the weaker hypothesis that only intervals of length q^{1/4-o(1)} expand (i.e., replace the 1/9-threshold in (129)/(133) by 1/4); if the same absolute M still emerges, the dependence on the short-interval expansion constant is not load-bearing for the existence claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Corollary 1 / Theorem 8) rests on the quantitative expansion estimates of Lemmas 13–14 (absolute κ>0 for the Bourgain–Gamburd machine and affine sieve on the particular generating sets coming from Ahlfors–David intervals of length q^{1/9-o(1)}). The paper itself records that this forces M large though computable and that 1/9 is far from the square-root barrier. No internal inconsistency, circularity, or hidden assumption appears in the new Diophantine independence machinery (Lemmas 24, 27, Proposition 32) or the repulsion argument that upgrades the earlier logarithmic bounds. The limitation is therefore methodological rather than a gap that would falsify the existence of an absolute (albeit huge) M.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The paper proves several strengthenings of classical results on bounded partial quotients of rationals a/q. For q belonging to a positive-density set Z (large primes, large square-free integers, high prime powers), Theorem 8 / Corollary 1 asserts the existence of absolute constants M ≥ 2 and ℳ ≪ M < ℳ such that there are at least q^{2w_M-1-o(1)} coprime a with all partial quotients of a/q bounded by M. Theorem 7 gives the asymptotically expected count q^{2w_M-1-o(1)} of a with M(a) ≤ M+2 once M ≥ C √log q. Theorem 6 improves Larcher’s bound on the sum of partial quotients to O(log q · √log log q) and supplies a matching lower bound on the number of such a. The argument combines the Cantor/Ahlfors–David structure of Z_M(t) (Lemmas 12, 20), expansion estimates in SL_2(Z/qZ) (Lemmas 13–16), a new Diophantine independence theory for critical denominators (Section 4, Lemmas 24, 27, Proposition 32), and a final repulsion step that eliminates denominators near √q.","tokens_in":37485,"tokens_out":771,"duration_ms":7076,"significance":"If correct, the paper settles Zaremba’s conjecture for every sufficiently large prime (and for a positive-density set of composite moduli) with an absolute though non-effective bound M. This is a substantial advance over Korobov’s O(log q) bound and the author’s earlier O(log q / log log q) result, and it improves Larcher’s estimate on the sum of partial quotients. The new independence machinery for critical denominators (Lemmas 24, 27) and the Ahlfors–David analysis appear to be of independent interest for Diophantine approximation and fractal geometry. The lower bounds on the number of good numerators match the expected order of magnitude in the regime M = Ω(√log q), which is a clean and sharp feature of the method.","major_comments":[],"minor_comments":[{"comment":"The absolute constant M produced by Theorem 8 is acknowledged to be large and non-effective because of the expansion constant κ and the 1/9-threshold in (99) and (129). A short remark quantifying the dependence of M on κ (or stating that no explicit numerical bound is claimed) would help the reader.","section":null},{"comment":"Notation for the two constants M and ℳ in Theorem 8 is slightly overloaded with the running parameter M used throughout Sections 2–5; a typographic distinction (e.g., script M versus roman M) would improve readability.","section":null},{"comment":"Lemma 19 (Möbius inversion for reduced fractions) is used crucially for composite q; a one-sentence reminder that the same argument is vacuous for prime q would clarify the logical structure for readers interested only in the prime case.","section":null},{"comment":"A few typographical slips appear (e.g., “Furthemore”, “acordingly”, missing spaces around some ≪ symbols). They do not affect the mathematics but should be cleaned in the final version.","section":null}],"recommendation":"accept","confidential_remarks":"The logical skeleton is complete and the new Diophantine independence arguments appear sound. The only non-effective black boxes are the expansion estimates already present in the author’s earlier work; they do not introduce circularity with respect to the present target. I see no load-bearing gap that would justify major revision or rejection. The paper is suitable for a top number-theory journal."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The punchline is Corollary 1 / Theorem 8: for every large prime (and every q in the positive-density set Z of square-frees and high prime powers) there is an absolute M such that at least q^{2w_M-1-o(1)} coprime a have all partial quotients of a/q bounded by M. That settles the classic Korobov case after sixty years.\n\nWhat is actually new is the Diophantine machinery in Section 4. The Ahlfors–David structure of Z_M(t) (Lemma 20) plus the independence of critical denominators of Type I (Lemmas 24, 27 and Proposition 32) let him kill the denominators near √q by a repulsion argument that upgrades his own earlier log/log-log bound to an absolute constant. The counting lower bounds in Theorems 6–7 for the sum of partial quotients and for M ≫ √log q are asymptotically tight and improve Larcher cleanly. Error terms are tracked, reduced fractions are removed by Möbius (Lemma 19), and the logical skeleton (modular inverses → Cantor decomposition → expansion → independence → repulsion) is complete.\n\nThe soft spot is exactly the one the paper flags: the absolute expansion constant κ and the 1/9-threshold coming from the Bourgain–Gamburd machine and affine sieve force M large and ineffective. That is a limitation of the method, not a hole in the new independence lemmas. The reliance on the author’s earlier expansion results is not circular; those papers rest on Helfgott and do not target absolute bounded quotients. No post-hoc fitting or internal contradiction appears.\n\nThis is for people working on continued fractions, discrepancy, thin orbits, or numerical integration. The math is solid enough that a serious editor should send it to referees. I would bring it to reading group and cite the counting statements.","headline":"Proves absolute-constant Zaremba for every large prime (and positive-density composites) via new critical-denominator independence; M huge but the argument holds.","tokens_in":38155,"tokens_out":489,"would_cite":true,"duration_ms":10388,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11J70","11K38","11A55","11B30"],"pacs":[],"model":"grok-4.5","headline":"Zaremba's conjecture holds for every large prime denominator: some a/q has all partial quotients bounded by an absolute constant.","keywords":["Zaremba conjecture","continued fractions","partial quotients","Korobov bound","Larcher sum","Ahlfors-David sets","discrepancy","modular expansion"],"falsifier":"Compute, for a sequence of large primes p, the minimal possible max partial quotient M(a) over a coprime to p; if this minimal value tends to infinity, the absolute-constant claim is false. Alternatively, check whether the product theorem used for intervals of length p^{1/9} holds with a positive spectral gap.","tokens_in":38124,"feed_emoji":"🔢","tokens_out":721,"duration_ms":5848,"temperature":0.7,"pith_summary":"Zaremba's conjecture asks whether every positive integer q admits a numerator a coprime to q whose continued-fraction expansion has all partial quotients bounded by a fixed absolute constant. The paper proves this for every sufficiently large prime q (and more generally for a positive-density set of denominators that includes all large primes, large square-free integers, and high prime powers). It also shows that the bound can be taken as small as O(sqrt(log q)) while still producing asymptotically the expected number of such numerators, and that the sum of the partial quotients can be kept as small as O(log q times sqrt(log log q)). These statements improve classical bounds of Korobov and Larcher and give concrete discrepancy estimates for lattice-point sequences used in numerical integration.","feed_headline":"Zaremba holds for every large prime: bounded partial quotients","feed_subtitle":"Absolute constant works for primes; weaker O(sqrt(log q)) bound already gives the expected count of numerators","key_machinery":"Critical denominators of Type I (denominators of convergents lying near sqrt(q)) living inside Ahlfors-David intervals of the Cantor set of rationals with bounded partial quotients. Their near-independence (linear relations with small coefficients are forbidden) is combined with Diophantine repulsion to eliminate those denominators that would force a large partial quotient.","core_discovery":"For every q belonging to a positive-density set Z that contains all large primes, there exist absolute constants M >= 2 and an intermediate bound such that at least q to the power 2w_M - 1 - o(1) residues a coprime to q have every partial quotient of a/q bounded by M. The same circle of ideas yields the weaker but still absolute bound O(sqrt(log q)) with the expected count of numerators, and an analogous lower bound for numerators whose partial-quotient sum is O(log q * sqrt(log log q)).","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Zaremba holds for all large primes: absolute M works","Every large prime has a with partial quotients O(sqrt(log q))","Primes yield expected count of a with bounded partial quotients","For large M, Omega(q^{1-O(1/M)}) coprime a with quotients <=M","Sum of partial quotients O(log q sqrt(log log q)) for primes"],"cache_read_input_tokens":32896,"weakest_assumption_plain":"The argument treats quantitative expansion bounds for the modular group SL_2(Z/qZ) as black boxes; if the spectral gap fails for the short Ahlfors-David intervals that arise, the independence of critical denominators collapses.","fun_headline_variants_meta":{"raw":{"variants":["Zaremba holds for all large primes: absolute M works","Every large prime has a with partial quotients O(sqrt(log q))","Primes yield expected count of a with bounded partial quotients","For large M, Omega(q^{1-O(1/M)}) coprime a with quotients <=M","Sum of partial quotients O(log q sqrt(log log q)) for primes"]},"model":"grok-4.5","effort":"low","cost_usd":0.005108,"raw_usage":{"total_tokens":1450,"prompt_tokens":802,"num_sources_used":0,"completion_tokens":109,"cost_in_usd_ticks":51080000,"prompt_tokens_details":{"text_tokens":802,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":539,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":802,"tokens_out":109,"duration_ms":6598,"temperature":1.0,"reasoning_tokens":539,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-14T21:29:36.372064+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Compute, for a sequence of large primes p, the minimal possible max partial quotient M(a) over a coprime to p; if this minimal value tends to infinity, the absolute-constant claim is false. Alternatively, check whether the product theorem used for intervals of length p^{1/9} holds with a positive spectral gap.","supporting_citations":[],"review_version":1}