{"id":"5ad57acc-a055-48b7-a8d2-c5f332ad635e","arxiv_id":"2611.16358","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"For a genus-two curve over F_q, exact recovery of the Hasse-Weil numerator from noisy random Euler product readouts has minimax sample complexity Theta_{c,W}(q^4 log(1/delta)) under Sobolev-regular noise.","lead":"Two mathematicians prove that exactly recovering a genus-two curve's zeta numerator from repeated, noisy samples of a randomized Euler product has minimax sample complexity order q^4 log(1/delta), and that using the full observation law cannot beat this order. This pins down the information content of a new arithmetic inference channel and matches a simple phase-based estimator to a Hellinger lower bound.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The lower bound's reliance on the Sobolev regularity sqrt(f) in H^1(R) is acknowledged but the paper does not verify that the specific noise densities for which the bound is claimed actually satisfy this condition; the uniform-law counterexample shows the condition is not vacuous, but the paper…","rationale":"The reader's weakest_assumption correctly identifies the Sobolev regularity condition as load-bearing. However, the reader's verdict is ACCEPT, implying that the condition is not a red flag. I partially agree: the condition is acknowledged, but the paper does not verify that it holds for a broad class of densities beyond the two examples (Cauchy and Gaussian). The paper claims the result holds for 'every fixed, known additive noise density f satisfying sqrt(f) in H^1(R)', but provides no general criterion or test to determine whether a given density meets this condition. This is a concern because the lower bound collapses if the condition fails, and the paper does not demonstrate that the condition is satisfied by many natural densities. The concrete test would check a specific density (Laplace) to see if the condition holds. If it does, the concern is mitigated; if it does not, the claim is misleading. Therefore, the verdict should be CONDITIONAL: accept only if the authors provide a verification that the condition holds for a broader class of densities, or at least provide a general criterion for checking the condition. The paper's own acknowledgment of the uniform-law counterexample shows they are aware of the limitation, but they do not address it beyond that. The concrete test would settle whether the condition is restrictive or not.","tokens_in":19797,"tokens_out":925,"duration_ms":10238,"concrete_test":"Compute the H^1 Sobolev norm of sqrt(f) for the Laplace density f(x) = (1/2) exp(-|x|). Specifically, evaluate J_W = ∫ ( (d/dx) sqrt(f(x)) )^2 dx. If J_W is finite, then the condition holds and the lower bound applies. If J_W is infinite, then the lower bound does not apply to Laplace noise, and the paper's claim of optimal sample complexity for 'every fixed, known additive noise density f satisfying sqrt(f) in H^1(R)' would be false for this density, unless the upper bound also fails (which it does not, as the upper bound requires no density condition). This test would settle whether the condition is satisfied by a common density not explicitly checked in the paper.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The lower bound (Theorem 1.1) depends on Lemma 5.2, which requires sqrt(f) in H^1(R). The paper acknowledges that the uniform law on [-1,1] violates this (Section 6), but does not verify that any other common noise density (e.g., Laplace, logistic, Student's t with finite variance) satisfies the condition. The only examples given are Cauchy (which does satisfy H^1) and Gaussian (which also satisfies H^1). The paper claims the result holds for 'every fixed, known additive noise density f satisfying sqrt(f) in H^1(R)', but provides no general criterion or test to determine whether a given density meets this condition. This is a load-bearing concern because the lower bound collapses if the noise density does not satisfy the Sobolev regularity, and the paper does not demonstrate that the condition is satisfied by a broad class of densities beyond the two examples. The reader's weakest_assumption correctly identifies this as fragile, but the paper's own acknowledgment does not resolve the concern; it merely flags the limitation. The concrete test would be to check whether a specific density, such as the Laplace distribution (f(x) = (1/2) exp(-|x|)), satisfies sqrt(f) in H^1(R). If it does not, then the lower bound does not apply to that noise, and the claim of optimal sample complexity for 'every fixed, known additive noise density f satisfying sqrt(f) in H^1(R)' would be misleading because the condition is not easily verifiable and may exclude many natural densities. If it does, then the concern is mitigated but not eliminated, as the paper would still lack a general method to verify the condition.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper studies the minimax sample complexity for exact recovery of the Hasse–Weil zeta numerator of a genus-two curve over a finite field from repeated noisy observations of a randomized Euler product at a single scale t_q = c/√q. The main result (Theorem 1.1) gives matching upper and lower bounds of order Θ_{c,W}(q^4 log(1/δ)) for the number of independent observations needed to recover the numerator with error probability at most δ, under the assumptions that the noise density f satisfies √f ∈ H^1(R) and that the observation channel is as specified. The upper bound uses a characteristic-function phase estimator that requires no moment assumptions on the noise. The lower bound constructs a hard pair of curves (with ordinary, absolutely simple Jacobians) and bounds the squared Hellinger distance between their observation laws by K_W q^{-4}, using a combination of binomial comparisons, translation smoothing, and tail bounds. The paper also provides alternative estimators for finite-variance noise and a Gaussian special case.","tokens_in":20133,"tokens_out":6138,"duration_ms":39205,"significance":"If correct, the result is significant because it establishes the minimax optimal sample complexity for a nontrivial arithmetic inference problem, showing that the full observation channel does not improve over a simple phase estimator. The matching bounds are derived from structural properties of the Weil coefficients and the noise regularity, without imposing moment conditions. The paper is rigorous in its derivations and provides explicit constants. The lower bound is particularly notable for its use of actual curves (ordinary, absolutely simple Jacobians) and a careful Hellinger comparison that handles the infinite tail. The result is a contribution to the intersection of arithmetic statistics, information theory, and statistical inference.","major_comments":[{"comment":"The lower bound (Theorem 1.1) depends critically on the Sobolev regularity condition √f ∈ H^1(R) via Lemma 5.2. The paper acknowledges that the uniform law on [-1,1] violates this condition (Section 6), but it does not verify that any other common noise density (e.g., Laplace, logistic, Student's t with finite variance) satisfies it. The only examples given are Cauchy (which does satisfy H^1) and Gaussian (which also satisfies H^1). The paper claims the result holds for 'every fixed, known additive noise density f satisfying √f ∈ H^1(R)', but provides no general criterion or test to determine whether a given density meets this condition. This is a load-bearing concern because the lower bound collapses if the noise density does not satisfy the Sobolev regularity, and the paper does not demonstrate that the condition is satisfied by a broad class of densities beyond the two examples. The reader's weakest_assumption correctly identifies this as fragile, but the paper's own acknowledgment does not resolve the concern; it merely flags the limitation. The concrete test would be to check whether a specific density, such as the Laplace distribution (f(x) = (1/2) exp(-|x|)), satisfies √f ∈ H^1(R). If it does not, then the lower bound does not apply to that noise, and the claim of optimal sample complexity for 'every fixed, known additive noise density f satisfying √f ∈ H^1(R)' would be misleading because the condition is not easily verifiable and may exclude many natural densities. If it does, the paper should provide a proof or reference. I recommend that the authors either (a) provide a sufficient condition or a simple test for √f ∈ H^1(R) (e.g., f positive, differentiable, with square-integrable derivative of √f), or (b) explicitly verify that a few more common densities (e.g., Laplace, logistic) satisfy the condition, or (c) discuss the scope of the condition more precisely, noting that the lower bound is proved only for densities satisfying this Sobolev regularity, and that the upper bound holds without it. This would clarify the applicability of the result.","section":"§5.1, Lemma 5.2 and §6","authors_acknowledge":"This is a counterexample to removing Sobolev regularity fromthat lemma, not to the minimax rate itself.","authors_acknowledge_p":0.65},{"comment":"The lower bound relies on the existence of two curves C_+, C_- with zeta numerators P_±(T) = 1 + T ± T^2 + qT^3 + q^2 T^4 for every prime power q ≥ 3, with ordinary, absolutely simple Jacobians. The proof cites [4, Appendix, Step 2] and [4, Theorem 1.2(2) and Table 1.2] and [5, Theorem 6]. While these references are published and presumably contain the necessary results, the paper does not reproduce the argument that for every q ≥ 3, the isogeny class with these Weil polynomials contains a genus-two Jacobian that is ordinary and absolutely simple. The reader's weakest_assumption notes that the existence is cited rather than proved. However, the paper does provide an explicit example over F_3 (Section 4.3) as a check. The concern is whether the cited references indeed guarantee existence for all q ≥ 3, including q that are not prime (i.e., prime powers). The paper states 'for every prime power q ≥ 3' in Lemma 4.1. The references [4] and [5] are about abelian surfaces over finite fields; [4, Theorem 1.2] gives conditions for existence of genus-two Jacobians in isogeny classes. The paper should verify that the specific polynomials P_± satisfy the conditions of [4, Theorem 1.2] for all prime powers q ≥ 3, not just primes. The argument in Lemma 4.1 uses the factorization and the condition that D = 8q+1-4b ≡ 5 (mod 8) is not a square, which holds for all q ≥ 3? For q a prime power, 8q+1-4b with b=±1 gives 8q+5 or 8q-3. Are these always non-squares? For q=4 (a prime power), 8*4+5=37, which is not a square; 8*4-3=29, not a square. For q=8, 8*8+5=69, not a square; 8*8-3=61, not a square. It seems plausible, but the paper should either provide a brief verification or cite a result that covers all prime powers. This is a minor load-bearing point because the lower bound is only claimed for q ≥ q_0(W) (which may be larger than 3), but the existence for all q ≥ 3 is used in Proposition 5.4 and the proof of Theorem 1.1. I recommend that the authors add a short argument or reference to show that the polynomials P_± are indeed realized by ordinary absolutely simple Jacobians for all prime powers q ≥ 3 (or at least for all q ≥ q_0(W)).","section":"§4.1, Lemma 4.1 and the existence of the hard pair for all q ≥ 3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Proposition 5.4 decomposes the Hellinger distance into four steps and bounds each using the triangle inequality and Cauchy-Schwarz. The bound for the third step (degree-two block) uses Lemma 5.3 (binomial comparison) and gives squared cost at most 27 q^{-4}. However, the derivation of the constant 27 is not fully explained. The paper states: 'Since Hellinger distance is symmetric, (5.4) gives squared cost at most 27 q^{-4}.' The calculation: from Lemma 5.3, H^2(R_{N-2}, R_{N-2+2}) ≤ 2/((N-2+1)(N-2+2)) = 2/((N-2+1)(N-2+2)). With N-2 = (q^2 - q - 2)/2, we have N-2+1 = (q^2 - q)/2 and N-2+2 = (q^2 - q + 2)/2. Then H^2 ≤ 2 / ( ((q^2 - q)/2) * ((q^2 - q + 2)/2) ) = 8 / ((q^2 - q)(q^2 - q + 2)). For q ≥ 3, this is ≤ 8 / ((q^2 - q)^2) = 8 / (q^2(q-1)^2) ≤ 8 / (q^2 * 4) = 2/q^2? Wait, that gives O(q^{-2}), not O(q^{-4}). But the paper claims squared cost at most 27 q^{-4}. There is a missing factor: the Hellinger distance is between the laws of h_2 R_{N-2} and h_2 R_{N-2+2}, where h_2 = arctanh(t_q^2) ≈ t_q^2 = c^2/q. The scaling by h_2 changes the Hellinger distance? The paper says 'multiplication by a nonzero scalar does not change an f-divergence.' Is that true for Hellinger distance? For a random variable X and a constant a ≠ 0, the Hellinger distance between the laws of aX and aY is not necessarily the same as between X and Y; it scales? Actually, if we consider the densities, scaling by a changes the support and the density by a factor 1/|a|. The Hellinger distance is invariant under invertible linear transformations? Let's check: For random variables X and Y with densities p_X, p_Y, the density of aX is (1/|a|) p_X(x/a). Then the Hellinger distance between aX and aY is ∫ (√(p_{aX}) - √(p_{aY}))^2 = ∫ ( (1/√{|a|}) √{p_X(x/a)} - (1/√{|a|}) √{p_Y(x/a)} )^2 dx = (1/|a|) ∫ (√{p_X(x/a)} - √{p_Y(x/a)})^2 dx = ∫ (√{p_X(u)} - √{p_Y(u)})^2 du, after change of variable u = x/a. So indeed Hellinger distance is scale-invariant. So the scaling by h_2 does not affect the Hellinger distance between the sign-sum laws. Therefore the bound from Lemma 5.3 applies directly to the laws of R_{N-2} and R_{N-2+2}, and then after convolution with the common noise and other blocks, the Hellinger distance is at most the original Hellinger distance (by contraction).","section":"§5.2, Proposition 5.4 and the Hellinger bound","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The derivation of the constant 27 q^{-4} for the degree-two block comparison is not explained. Please expand the calculation to show how Lemma 5.3 and the values of N_2^- yield this bound. This would improve reproducibility.","section":"§5.2, Proposition 5.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof cites [4, Appendix, Step 2] and [4, Theorem 1.2(2) and Table 1.2] and [5, Theorem 6] for the existence of ordinary, absolutely simple Jacobians for all prime powers q ≥ 3. It would be helpful to briefly indicate why these references apply to all prime powers, not just primes, or to note that the argument works for any q ≥ 3 (including prime powers) because the conditions involve only the Weil polynomial and the discriminant condition.","section":"§4.1, Lemma 4.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The estimator uses the frequency ω satisfying 0 < ω ≤ 1/2 and |φ_W(ω)| ≥ 1/2. The existence of such ω is guaranteed by continuity of φ_W at 0. However, the paper does not discuss how to choose ω in practice given only the known noise density. This is a minor omission; a brief remark would suffice.","section":"§3.2, Proposition 3.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The condition √f ∈ H^1(R) implies that f is absolutely continuous? The paper should note that this condition is equivalent to f being absolutely continuous with derivative f' such that ∫ (f')^2 / f dx < ∞ (if f>0). This is a standard fact; a reference would be helpful.","section":"§5.1, Lemma 5.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper states that for the uniform law on [-1,1], H^2(Law(W+h), Law(W)) = |h| for |h| ≤ 2. This is correct, but it would be useful to note that this violates the quadratic bound (5.2) because the Hellinger distance is linear in h, not quadratic. This is a counterexample to the lemma, not to the minimax rate, as the paper correctly notes.","section":"§6","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The definition of d_c = (1/4) min{c^4, 2c^2(1-12c^2)} > 0. For c close to 1/√12, the term 2c^2(1-12c^2) becomes very small, making d_c small. The paper might note that the upper bound constant depends on d_c^{-2}, which blows up as c approaches 1/√12. This is expected and not a flaw.","section":"§1, Introduction","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The explicit example over F_3 is a nice verification. However, the paper should clarify that this example is not needed for the proof but serves as an illustration. Also, the field F_9 is described as F_3[z]/(z^2+1), but z^2 = 2 in F_3? Actually, in F_3, 2 = -1, so z^2 = -1, which is consistent. The example is correct.","section":"§4.3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"This proposition provides an independent check for Gaussian noise. It is not used in the main theorem but is a nice addition. The proof sketch is clear.","section":"§5.3, Proposition 5.7","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper is well-written and the logic is clear. The notation is consistent. The references are appropriate.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"This paper is a clean, self-contained piece of information-theoretic arithmetic geometry. It proves matching upper and lower bounds on the number of noisy scalar observations needed to recover the zeta numerator of a genus-two curve over a finite field. The result is new, the proof is honest, and the constants are explicit. The upper bound uses a clever phase estimator that works without any moment assumptions on the noise, and the lower bound constructs a genuine hard pair of curves with ordinary absolutely simple Jacobians and compares their full observation laws via Hellinger distance. The paper is careful about what it does and does not claim: it flags the Sobolev condition on the noise density as load-bearing for the lower bound, gives a counterexample (uniform law) where it fails, and notes that atomic or noiseless readouts are not covered. The stress-test concern about verifying the H^1 condition for common densities is real but not a flaw in the paper: the authors state the condition as a hypothesis, prove the bound under it, and give two examples (Cauchy, Gaussian) that satisfy it. They do not claim a general criterion, and the reader can check any specific density. The paper's own Section 6 is unusually frank about limitations. The only soft spot worth mentioning is that the lower bound's dependence on the Howe–Nart–Ritzenthaler and Howe–Zhu existence results for the hard pair over all q >= 3 is cited rather than proved here, but that is standard and the citation is to published, peer-reviewed work. The paper deserves a serious referee. I would cite it if I worked on zeta recovery or information-theoretic number theory. It is not a reading-group paper for a general audience, but a specialist would learn from it. Recommend peer review with no special urgency.","headline":"Matching minimax bounds for recovering genus-two zeta numerators from noisy scalar observations; clean, honest, and technically sound.","tokens_in":20824,"tokens_out":445,"duration_ms":5498},"pith_extraction":null,"created_at":"2026-11-13T06:29:35.791381+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":null,"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}