{"id":"52b9ed4d-775d-4e67-9041-353ba1ffe628","arxiv_id":"2611.16550","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Unconditional central limit theorems for Frobenius, Hecke, and Kloosterman angles are proved for a much larger class of test functions than previously known.","lead":"This paper proves central limit theorems for the fluctuations of Frobenius, Hecke, and Kloosterman angles averaged over large families, removing the need for a Riemann hypothesis assumption in the elliptic curve case. It provides a general framework that replaces independence with approximate independence and convergence to Weyl limits, yielding unconditional results and broader test functions.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"No significant objection identified; the paper is internally consistent and the weakest assumption (super-polynomial growth of the box) is stated and necessary for the CRT-based argument.","rationale":"The reader identified the growth condition as the weakest assumption; that is correct. The paper's logic is self-consistent, the arithmetic inputs are unconditional and correctly applied, and the moment computations are sound. No further issues were found that would alter the ACCEPT verdict.","tokens_in":55892,"tokens_out":7191,"duration_ms":55467,"concrete_test":"Attempt to prove Theorem 1 with the relaxed condition A,B ≥ x^c for some fixed c > 0 instead of log A/log x → ∞. If the error terms in (GC2) and (GC3) can still be shown to be o(π(x)^{r/2}) using a different handling of the boundary errors (e.g., by exploiting cancellation in the sum over primes), then the current growth condition is not essential.  A negative result would confirm that the super‑polynomial condition is necessary for the CRT‑based approach and therefore the paper's stated condition is appropriate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claims for Frobenius and Kloosterman angles depend on the condition that log A/log x → ∞ (and similarly for B,N), which makes the Chinese remainder theorem approximation error negligible.  This condition is required by the method: in Proposition 25(ii) the error bound contains a factor ∏ p_i which forces A,B to grow faster than any polynomial in x to keep the error o(π(x)^{r/2}).  The paper acknowledges this condition and does not claim minimality.  No internal inconsistency, hidden assumption, or circular reasoning was found.  The verification of (GC1)-(GC3) follows the paper's own explicit template, and the arithmetic inputs (Birch's formulas, Deligne bound, Katz's bounds, Eichler–Selberg trace formula) are unconditional and correctly cited.  The method-of-moments argument for convergence in distribution is standard and correctly applied.  Therefore no load-bearing concern that would undermine the central claim has been identified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper develops a general method-of-moments framework (Theorems 11, 13, 14) in which a central limit theorem for normalized sums Σ h(X_n) over a growing family of sequences X_n in [0,1] follows from three quantitative conditions: decay of the truncation tail of h (C1/GC1), approximate independence of the X_n at distinct indices (C2/GC2), and convergence of the family averages c_{m,n,k} to Weyl limits d_m at a suitable rate (C3/GC3). The framework is then applied to three arithmetic families. Theorem 1 proves an unconditional CLT for Frobenius angles of elliptic curves E(a,b) with |a| ≤ A, |b| ≤ B, for even test functions with polynomially decaying Fourier transform (κ > 2), under the growth condition log A/log x, log B/log x → ∞; this removes the modular Riemann hypothesis used by [BPS20] and covers dilated angles αθ_E(p). Theorem 2 extends Xi's Kloosterman-angle CLT to test functions of limited smoothness, to moments of K(p,a)/√p, and to dilated angles, using Katz's bound as the only arithmetic input at a single prime. Theorem 3 gives a graded family of CLTs for Hecke angles of newforms with weight and level varying simultaneously, with the admissible growth of kN tied to the decay class of the test function via the uniform Eichler–Selberg trace formula.","tokens_in":56269,"tokens_out":24426,"duration_ms":141424,"significance":"I found the central claims correct: the moment expansion in Theorem 11 is rigorous, the exceptional-set refinement in Theorem 13 is sound, and the verifications of (GC1)–(GC3) in Sections 3–5 follow the paper's own template with the stated arithmetic inputs from published sources (Birch's formulas plus the Ramanujan–Deligne bound, Katz's symmetric-power bound, and the uniform Eichler–Selberg trace formula of [MS24]). I checked the error bookkeeping in equations (28), (30)–(32), (44)–(45), (68)–(69), and Proposition 41, including the mixed terms in the Hecke bound and the m_i = 0 terms in (GC3), and found the estimates consistent. If the results hold — and on my reading they do — the paper is a substantial advance: it removes an unproved hypothesis from the Frobenius-angle CLT, widens the test-function class from Schwartz functions with exponential Fourier decay to polynomial decay, provides the first Hecke-angle CLT with weight and level varying simultaneously, and delivers a sharp-statistics CLT for Hecke angles along with a concrete explanation of why the other two families resist.","major_comments":[],"minor_comments":[{"comment":"The growth conditions for the Hecke family should be typeset unambiguously as ratios, log(kN/8^{ω(N)})/(M log x) → ∞ and log(kN/8^{ω(N)})/(√x log x) → ∞; in the present rendering they appear as products, and although the intended meaning is clear from the verification in Section 5.1 and the proof of Theorem 51, an unambiguous display would prevent misreading by a careful reader.","section":"Theorems 3 and 51","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Remark 12(e) credits the idea of working with powers of trigonometric polynomials to the unpublished note [SWZ18]; since the proofs here are self-contained and no result is imported from that note, a sentence stating explicitly that the attribution is contextual only would clarify that the paper does not rely on unpublished work.","section":"Remark 12(e) and [SWZ18]","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The parenthetical remark that the statement of [BP19, Lemma 8.2] omits a sign and the term 1/p^{(k+1)/2} is useful, but it should say plainly that the displayed identity in Proposition 20(iv) is the corrected form used in this paper, so that a reader comparing the two sources is not confused about which statement is being applied.","section":"Section 3, after Proposition 20","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The modification of the Beurling–Selberg polynomials to produce even Fourier coefficients is only sketched; since Lemma 45(e) is used in the key estimate of Remark 52(b), a sentence describing the symmetrization, or a precise pointer to the even-coefficient variant in [PS19, Section 3], would make the sharp-statistics section more self-contained.","section":"Lemma 45, Section 7.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"In the sentence introducing [MMP23], 'the Erdős–Kac theorem for ω(an)' should read 'for ω(a_n)'; the sequence (a_n)_{n≥1} is introduced immediately afterwards, so this is a typographical slip.","section":"Section 1.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The handling of the m_i = 0 terms in (GC3) is correct but terse; writing the estimate |c̃_{0,n,k} − d_0| = |G_{n,k}|/|F_k| and invoking (20) with u = 1, together with Lemma 18(ii), would make the claimed o(√A(x)) contribution immediate for the reader.","section":"Theorem 13, proof Step 2","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"accept"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"The main result is that the Frobenius angle CLT for elliptic curves is finally unconditional, removing the Modular Riemann Hypothesis that earlier work required. That alone makes this paper important. But it also gives a clean general framework (three conditions: approximate independence, convergence of averages to Weyl limits, and truncation) and then applies it to Kloosterman and Hecke angles, extending the class of test functions and, for Hecke angles, letting the weight and level vary together. The arithmetic inputs (Birch, Deligne, Katz, Eichler–Selberg) are unconditional and correctly cited. The moment method is used correctly, and the paper is honest about where the method works and where it does not.\n\nThe soft spots are real but not fatal. The growth condition for the box (log A / log x → ∞, same for B) is strong—the box must grow faster than any polynomial in x. That is necessary for the CRT argument to make the error negligible, and the paper states it plainly. For Hecke angles the growth condition on kN is even stronger in the sharp statistics section (requiring √x/log x divergence), which the authors acknowledge and explain. The sharp CLT only works for Hecke angles; for Frobenius and Kloosterman angles the method would need discrepancy bounds of square-root strength that are not available. The paper is upfront about this limitation.\n\nThere is an AI declaration (Claude helped with exposition and proofreading). That is unusual but transparent, and it does not bear on the mathematical correctness.\n\nWho should read this: anybody working on arithmetic statistics, Sato-Tate, or CLTs in number theory. The general framework is clearly laid out and could be reused for other families (the paper hints at this). The results themselves resolve a barrier (MRH) that had been open for the elliptic curve problem.\n\nVerdict: yes, send this to a serious referee. The paper is well-executed, the claims are supported, and the limitations are clearly stated. It deserves peer review.","headline":"Unconditional CLTs for Frobenius angles and more – this paper is the real thing.","tokens_in":56638,"tokens_out":1051,"duration_ms":9936},"pith_extraction":{"msc":["11G05","11F30","11L05","11K38","60F05"],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"The paper proves unconditional central limit theorems for fluctuations of Frobenius, Hecke, and Kloosterman angles around their limiting distributions, for a much wider class of test functions than previously possible.","keywords":["central limit theorem","Frobenius angles","Hecke angles","Kloosterman angles","Sato–Tate law","method of moments","Eichler–Selberg trace formula","unconditional"],"falsifier":"Take a fixed smooth test function h and compute the fourth moment of the normalized sum for Frobenius angles when A and B grow like $x^{{0.9}}$ instead of faster than any power; if the limiting distribution is still Gaussian, the proof's error estimate would need a different mechanism, since the paper's argument relies on the faster-than-polynomial growth to make the joint average error o(π(x)^{r/2}).","tokens_in":57421,"feed_emoji":"","tokens_out":3296,"duration_ms":25908,"temperature":0.7,"pith_summary":"The paper establishes that the normalized sums of Frobenius angles of elliptic curves, averaged over families with coefficients in rapidly growing boxes, converge in distribution to a standard Gaussian without assuming the modular Riemann hypothesis. This is the first unconditional result of its kind. The same general framework, which replaces independence by approximate independence and convergence to Weyl limits, also yields central limit theorems for Hecke angles and Kloosterman angles, extending them to larger classes of test functions and, for Hecke angles, allowing weight and level to vary simultaneously. The key advance is that the arithmetic information at each prime alone—rather than cancellation over primes—suffices when the test function decays sufficiently in frequency.","feed_headline":"Gaussian law for Frobenius angles without Riemann hypothesis","feed_subtitle":"Unified framework yields CLTs for elliptic curves, modular forms, and Kloosterman sums under weaker assumptions","key_machinery":"The machinery is a general theorem (Theorem 11) that replaces independence of the angles at distinct primes by two quantitative conditions: approximate independence, measured through cosines, and convergence of family averages to Weyl limits at a controlled rate. A refinement (Theorem 13) allows discarding thin exceptional sets. For elliptic curves, the arithmetic input is Birch's formula and the Ramanujan–Deligne bound; for Kloosterman sums, Katz's bounds for symmetric powers of the Kloosterman sheaf; for Hecke angles, the Eichler–Selberg trace formula uniform in weight and level.","core_discovery":"The central claim is that for families of elliptic curves with coefficients in boxes growing faster than any polynomial in x, the normalized sums of smooth test functions of Frobenius angles converge unconditionally to a Gaussian distribution. Earlier results required the modular Riemann hypothesis for L-functions of Hecke eigenforms. The authors bypass this requirement by using the decay of the Fourier coefficients of the test function to make a bound at each individual prime sufficient, eliminating the need for cancellation over primes that would otherwise require the Riemann hypothesis. The same general mechanism works for Kloosterman sums (using Katz's bounds) and for Hecke newforms (using the Eichler–Selberg trace formula), providing unified proofs for all three families.","pith_inferences":["The method may apply to other arithmetic families where analogous bounds at individual primes are available, such as families of cusp forms on GL(N) or Siegel modular forms, without requiring a Riemann hypothesis.","The growth conditions for Frobenius and Kloosterman angles require the family to grow faster than any power of x, which is less restrictive than it seems: typical choices like A = x^{log x} satisfy it easily.","The explicit comparison of the three families in Section 6 explains why sharp CLTs (for intervals) are possible for Hecke angles but not for Frobenius and Kloosterman angles: the geometric decay of one-prime differences for Hecke angles versus polynomial decay for the others."],"forward_implications":["The central limit theorem for Frobenius angles now holds for all even Schwartz functions, not just those with exponential Fourier decay assumed previously.","Kloosterman angles satisfy an unconditional CLT for smooth test functions with polynomial decay of Fourier coefficients, extending earlier results that required compact support.","Hecke angles now have a CLT with weight and level varying simultaneously, provided the product kN grows sufficiently fast relative to x.","The framework applies to dilated angles αθ for α≥2, whose limiting distribution becomes uniform, yielding CLTs for these as well."],"supporting_citations":[{"why":"previous CLT for Frobenius angles that assumed the modular Riemann hypothesis; this paper removes that assumption.","marker":"[BPS20]"},{"why":"provides the averaged moments of elliptic curve coefficients (Birch's formula) needed for the one-prime estimates.","marker":"[BP19]"},{"why":"previous CLT for Hecke angles with fixed level; extended here to varying level and larger test function classes.","marker":"[Nag06]"},{"why":"previous CLT for Kloosterman angles with compactly supported test functions; extended here to functions with polynomial decay.","marker":"[Xi17]"},{"why":"provides the needed bounds for symmetric powers of the Kloosterman sheaf used for the one-prime estimates.","marker":"[Kat88]"}],"fun_headline_variants":["Unconditional CLT for Frobenius angles, no RH needed","Gaussian fluctuations without the modular Riemann hypothesis","Unified theorem for Frobenius, Hecke, and Kloosterman angles","Elliptic curve Frobenius angles obey Gaussian law unconditionally","CLT for Hecke angles with weight and level varying"],"cache_read_input_tokens":0,"weakest_assumption_plain":"For Frobenius angles, the load-bearing premise is that the boxes A and B containing the elliptic curve coefficients must grow faster than any polynomial in the prime range x, so that the error from the Chinese remainder theorem in verifying approximate independence of angles at distinct primes becomes negligible.","fun_headline_variants_meta":{"raw":{"variants":["Unconditional CLT for Frobenius angles, no RH needed","Gaussian fluctuations without the modular Riemann hypothesis","Unified theorem for Frobenius, Hecke, and Kloosterman angles","Elliptic curve Frobenius angles obey Gaussian law unconditionally","CLT for Hecke angles with weight and level varying"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.000869,"raw_usage":{"cost":0.000868588,"is_byok":false,"cost_details":{"upstream_inference_cost":0.000868588,"upstream_inference_prompt_cost":0.000100828,"upstream_inference_completions_cost":0.00076776},"total_tokens":3667,"prompt_tokens":925,"completion_tokens":2742,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":256,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":2655}},"tokens_in":925,"tokens_out":2742,"cost_source":"openrouter_usage","duration_ms":20425,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":2655,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-11-13T14:21:23.804598+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":"Take a fixed smooth test function h and compute the fourth moment of the normalized sum for Frobenius angles when A and B grow like $x^{{0.9}}$ instead of faster than any power; if the limiting distribution is still Gaussian, the proof's error estimate would need a different mechanism, since the paper's argument relies on the faster-than-polynomial growth to make the joint average error o(π(x)^{r/2}).","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}