{"id":"16beb9cb-4d52-4d4f-a913-11461a9be60e","arxiv_id":"2607.25955","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"For polynomial phases with product Hilbert kernels in k variables, the exponential sums are uniformly bounded exactly when every exponent vector has at most one odd component (unless the whole phase cancels by symmetry).","lead":"This paper finds exactly when certain sums of oscillating fractions — the multi-parameter analogue of the Hilbert transform — stay bounded no matter how long they are or how the frequencies are chosen. The answer is a simple parity condition on the polynomial exponents, and it differs from the continuous integral case.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sufficiency proof depends on Lemma 9.1, an unproved anisotropic Konyagin-type congruence count; if its exponent is weak, the averaged Gauss-sum decay in E_gauss estimates collapses.","rationale":"I read the paper as a serious, technically dense attempt at a new multi-parameter circle method, and I found no internal contradiction or obvious fatal flaw. The necessity proof is self-contained and coherent; the parity/symmetry arguments in Section 7 check out. The lcm step in Proposition 3.1, though compressed, is valid once one quantifies lcm(L_linear, L_nonlinear): outside the (1/2,1)-major arcs gives lcm(all q_m)≥2^{j_k/20}, and the linear part is ≤2^{j_k/60}, so the nonlinear lcm is ≥2^{j_k/30}. Lemma 3.4, the reader's main worry, is recoverable from Carbery–Wright by enlarging the box by one unit and using |∇P|≤A/2^{j_k} together with r≤j_k; the missing proof is a short exercise. The genuinely load-bearing gap is Lemma 9.1, a nontrivial anisotropic congruence-counting theorem stated without proof or citation, on which Proposition 3.2 and hence the E_gauss estimates depend. If Lemma 9.1 is standard and can be supplied, the sufficiency proof likely works; if not, the averaged Gauss-sum decay may fail exactly in the unbalanced regime the paper is designed to handle. This does not change the reader's CONDITIONAL verdict: the concern is real and should be resolved by a proof or citation, but no conclusive failure has been demonstrated.","tokens_in":81771,"tokens_out":18442,"duration_ms":174235,"concrete_test":"Prove Lemma 9.1 in the minimal anisotropic case needed by the proof: k=2, j_1≫j_2, q a prime power comparable to 2^{j_2}, and P(t_1,t_2) a degree-2 primitive polynomial such as t_1^2−t_2^2 or t_1^2−t_2. Compute the exact maximal solution count in [1,2^{j_1}]×[1,2^{j_2}] and verify whether it satisfies ≤C2^{j_1+j_2}(2^{-c j_2}+q^{-c}) with a c that can be chosen uniformly in j_1/j_2. If the count contains an extra factor polynomial in j_2, re-run Lemma 5.5 and Lemma 6.4 with that factor; if the E_gauss sums no longer decay exponentially in j_{ℓ+1}, Proposition 3.2 needs repair before the sufficiency argument is complete.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The averaged Gauss-sum estimate Proposition 3.2, used as (3.18) in Lemma 5.5 and Lemma 6.4 to bound the E_gauss terms by q^{-δ_Λ}2^{-δ_Λ j_{ℓ+1}}, is proved only modulo Lemma 9.1. Lemma 9.1 is a multi-parameter Konyagin-type theorem asserting that, for a primitive polynomial of degree at least 2, the number of solutions of P(t)≡0 mod q in an anisotropic box ∏[1,2^{j_ν}] is ≤ C2^{j_1+…+j_k}(2^{-c j_k}+q^{-c}). The paper states it without proof and without citing a source. This is not a cosmetic omission: the proof of Proposition 3.2 uses Lemma 9.1 to control the 'sub' contribution, where the denominator of the layered Gauss sum is large; the divisor factor d(q) then requires the q^{-c} term to be strong enough to survive the divisor-sum and give a uniform geometric decay in j_{ℓ+1}. If Lemma 9.1 has an additional logarithmic factor, or if c must degrade with the aspect ratio j_1/j_k, then the E_gauss bounds in Lemmas 5.5 and 6.4 acquire logarithmic losses and the induction proving Theorem 4.1 is no longer summable. Because Lemma 9.1 is asserted rather than established, the sufficiency half of Main Theorem 1 is currently not fully checkable at this load-bearing point. By contrast, the reader's primary worry, Lemma 3.4, is less severe: it follows from Carbery–Wright with a derivative enlargement using j_k≥r, so the missing proof there is routine. Lemma 9.1 is the more substantive gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies uniform boundedness of multi-parameter exponential sums with a product Hilbert kernel and polynomial phase, and the ℓ^p boundedness of the associated discrete multiple Hilbert transform. Main Theorem 1 gives a complete dichotomy: the sums are uniformly bounded iff Λ contains no odd subset (in which case they vanish) or every m∈Λ has at most one odd component. Main Theorem 2 asserts ℓ^p boundedness for every 1<p<∞ under the same sufficient condition. The proof develops a multi-parameter circle method built on layered arcs and a 'major-arc rigidity' principle, together with Weyl, Gauss, and sublevel-set estimates. The necessity direction is proved by a residue-class decomposition and Poisson summation. The sufficiency proof is organized as an induction over k and occupies Sections 4–6; the ℓ^p result is obtained in Section 8 via the Ionescu–Wainger multiplier theory.","tokens_in":82267,"tokens_out":9280,"duration_ms":89908,"significance":"If correct, the main theorem is a substantial and natural completion of the one-parameter theorem of Arkhipov–Oskolkov and clarifies the contrast with the continuous multiple Hilbert transform. The paper gives an explicit and checkable characterization, and the proposed major-arc rigidity mechanism is a genuinely new idea that could have further use. The necessity proof is detailed and appears to be self-contained. However, the sufficiency half currently rests on two large unproved auxiliary estimates—Lemma 3.4 and Lemma 9.1—both of which are used at decisive points to obtain geometric decay. In particular, the averaged Gauss-sum estimate Proposition 3.2 is only proved modulo Lemma 9.1, and the sublevel-set decay used for the E_sub terms is asserted without proof. Until these estimates are supplied with complete proofs or precise references, the central claim is not fully verifiable. The paper is therefore significant but conditional.","major_comments":[{"comment":"Lemma 9.1 is the only support for the averaged Gauss-sum estimate Proposition 3.2. The text says the proof is by 'standard induction' and is omitted, and no reference is given. The one-variable Konyagin theorem [28] does not formally imply the anisotropic k-parameter bound (9.6). This estimate is used through (3.18) in Lemmas 5.5 and 6.4 to bound E_gauss by q^{-δΛ}2^{-δΛ j_{ℓ+1}}. If (9.6) carries logarithmic factors or a decay constant c that degrades with the aspect ratio j_1/j_k, the divisor sum d(q) in (3.14) cannot be controlled and the E_gauss bounds collapse. A complete proof, or a precise citation to a result containing (9.6), is indispensable.","section":"§9.2, Lemma 9.1; used in §5.5 and §6.3"},{"comment":"The discrete sublevel-set estimate is introduced 'without proof'. The text only mentions the continuous analogue from Carbery–Wright [13], but the passage from continuous measure to a sharp lattice-point count in dyadic boxes requires additional argument and is not automatic at the stated exponent 2^{-(1/d)r}. This estimate is used in Lemmas 5.6 and 6.5 to obtain the 2^{-cj2} / 2^{-cj_{ℓ+1}} decay of E_sub. Any logarithmic loss or weaker exponent would break the summability in the induction proving Theorem 4.1. Please provide a complete proof of Lemma 3.4 or a reference containing exactly this statement.","section":"§3.4, Lemma 3.4; used in §5.4 and §6.3"},{"comment":"The proof of Proposition 3.1 is compressed in Case 2. The text asserts that outside Ψ^{Ω,major}_{j,(1/2,1)} one has lcm(q_m : m∈Ω) ≥ 2^{j_k/20}, and that together with the bound on the linear denominators this recovers condition (3.9). The implied step is presumably lcm(|m|≥2) ≥ lcm(all)/lcm(|m|=1) ≥ 2^{j_k/20 - k j_k/(60k)} = 2^{j_k/30}, but this is not written out. Since Proposition 3.1 feeds into Lemma 4.1 and the balanced Weyl bound, the constants and the lcm argument should be spelled out so the exponent is verifiable.","section":"§3.2, Proposition 3.1"}],"minor_comments":[{"comment":"Typo: 'it suffies' should be 'it suffices'. Throughout, the distinction between HΛ_j and the continuous HΛ_j is easy to miss; consider using different fonts.","section":"§4.2"},{"comment":"In the proof following (5.35), 'indendent of t2' should be 'independent of t2'. Also, the summation index condition R_{m,j_2}(η)>0 under the sum is awkward; clarify the range of η.","section":"§5.4"},{"comment":"The notation h=[λ^ϵ] and N=[h^{D/s}] conflicts with the earlier convention [n]={1,...,n}. The authors note this conflict only in parentheses; it would be cleaner to use ⌊λ^ϵ⌋ and avoid a second meaning.","section":"§8.3"},{"comment":"The sentence 'Therefore, the uniform boundedness of HΛ_N(ξ) implies that of HΛ_N(ξ)' appears to contain a typo; the two displayed symbols are identical. The intended comparison with the continuous transform should be restated.","section":"§1.2, Remark 1.1"},{"comment":"The remark that Lemma 9.1 'is therefore omitted' is not acceptable in a research paper if the result is not standard; this is already the content of Major Comment 1, but the wording should also be changed in a revision.","section":"§9.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript addresses a natural and difficult question and the main structural ideas look promising. My recommendation is driven entirely by the two unproved estimates, especially Lemma 9.1, which are needed for the sufficiency half. If the authors can supply complete proofs of Lemma 3.4 and Lemma 9.1, or point to a source that contains them, the paper would likely be suitable for publication. The necessity proof and the major-arc rigidity construction are the strongest parts of the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main theorem is a real new result: a necessary and sufficient condition for uniform boundedness of multi-parameter exponential sums with product Hilbert kernels, and the discrete condition is genuinely different from the continuous one. The layered-arc / major-arc rigidity idea is the right way to attack the unbalanced regime, and the necessity proof is explicit and mostly self-contained. The parity argument in Section 7 is clean. Main Theorem 2 gives the corresponding ell^p boundedness and connects the multiplier bound to an actual operator, which is a meaningful payoff. The survey of ergodic context is useful, and the citations to [27] and [21] supply method and background rather than carrying the main theorem.\n\nThe soft spots are real but not evenly distributed. Lemma 3.4 (discrete sublevel sets) is stated without proof, but it looks routine from Carbery-Wright with a derivative enlargement, so I am not deeply worried. Proposition 3.1's lcm step is compressed and should be expanded. The multi-parameter induction in Section 6 is long; I could not check every exponent, but the structure is coherent and the induction hypothesis is used honestly.\n\nThe load-bearing problem is Lemma 9.1, the anisotropic Konyagin-type count. It is stated without proof and without citation, and Proposition 3.2, the averaged Gauss sum estimate, depends on it. Lemma 9.1 appears in Lemmas 5.5 and 6.4 through (3.18), so if it needs logarithmic factors, or if the c decays with the aspect ratio, the E_gauss estimates develop logarithmic losses and the induction in Theorem 4.1 stops being summable. The stress-test note is right: this is the main checkable gap. The manuscript itself flags that the proof is omitted, which is honest, but for a theorem of this size an omitted proof of a new multi-parameter counting lemma is too much to leave to the reader.\n\nI would not desk reject this. I would send it to a specialist referee with explicit instructions to look at Section 9.2, Lemma 9.1, and the transition from Proposition 3.2 to Lemmas 5.5 and 6.4. If Lemma 9.1 is supplied with the stated decay, the sufficiency proof looks credible. The paper deserves a serious referee even if the eventual verdict is conditional.","headline":"Genuinely new and likely correct, but the sufficiency proof rests on an unproved anisotropic Konyagin-type lemma (9.1) that must be supplied before the result is checkable.","tokens_in":82737,"tokens_out":2272,"would_cite":true,"duration_ms":28939,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B20","11L07","42B25"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the multi-parameter exponential sum with product Hilbert kernel is uniformly bounded exactly when a parity condition holds: every monomial exponent has at most one odd coordinate, unless the sum vanishes by symmetry.","keywords":["multi-parameter exponential sums","product Hilbert kernel","discrete Hilbert transform","multi-parameter circle method","layered arcs","major arc rigidity","parity condition","ℓ^p boundedness"],"falsifier":"A direct counting check of Lemma 3.4 would settle the proof: take the polynomial t1 t2 on the box [1,2^j]^2 and compare the claimed bound C 2^{2j} 2^{-r/2} with the actual number of pairs satisfying |t1 t2|≤2^{2j-r}; if a logarithmic factor appears, the lemma is false as stated and the E_sub estimates lose their summable decay. At the level of the theorem itself, a numerical search for Λ={(1,1)} in k=2 that kept sup_N,ξ |H^Λ_N(ξ)| bounded would refute the necessity half, since the paper proves divergence for this forbidden monomial.","tokens_in":81642,"feed_emoji":"🧮","tokens_out":6941,"duration_ms":70985,"temperature":0.7,"pith_summary":"The paper characterizes, for every finite set Λ of monomials in k variables, when the truncated multi-parameter exponential sum with the product kernel 1/(t1⋯tk) remains uniformly bounded in the coefficients and cutoff parameters. The answer is a parity rule: if no subset of Λ has all coordinate sums odd, the sum is identically zero by symmetry; otherwise uniform boundedness holds precisely when every monomial exponent in Λ has at most one odd coordinate. The paper proves both necessity and sufficiency, and under the same condition shows the associated discrete multiple Hilbert transform is bounded on ℓ^p(Z^{|Λ|}) for every 1<p<∞. The engine is a higher-dimensional multi-parameter circle method built from layered major arcs and a major-arc rigidity mechanism that propagates rational approximation information from one variable layer to the next.","feed_headline":"Hilbert-type sums bounded iff monomials have ≤1 odd coordinate","feed_subtitle":"A layered circle method makes the parity condition sharp and yields ℓ^p bounds for the associated transforms.","key_machinery":"The load-bearing device is a k-parameter circle method with layered arcs: the phase Σ_{m∈Λ} ξ_m t^m is read as a polynomial in the first ℓ variables whose coefficients are polynomials in the remaining variables. For each frozen outer variable, one defines major arcs for the sliced frequency ξ(t2); when these sliced major arcs occur for many values of the variable, a rigidity lemma forces the original frequency to lie in a much wider coarse major arc. This major-arc rigidity converts a heavy concentration of layered major arcs into rational-approximation information at the next layer, allowing the variables to be peeled off successively. The coarse arcs are then merged into the original major","core_discovery":"For fixed Λ⊂Z^k_+, consider H^Λ_N(ξ)=Σ_{t∈R(N)∩Z^k} e^{2πiΣ_{m∈Λ}ξ_m t^m}/(t1⋯tk). Main Theorem 1 states a strict dichotomy: if Λ contains no odd subset, the sum is identically 0 by symmetry; if Λ contains an odd subset, the supremum over N,ξ is finite if and only if every m∈Λ has at most one odd coordinate. The necessity proof exhibits, for any forbidden Λ, frequencies and truncations along which |H^Λ_N(ξ)| grows like log N. Main Theorem 2 upgrades this uniform multiplier bound to ℓ^p(Z^{|Λ|})→ℓ^p(Z^{|Λ|}) boundedness of the limiting discrete multiple Hilbert transform for every 1<p<∞. The proof splits the summation scales into balanced and unbalanced sectors, approximates major arcs by Gau","pith_inferences":["The layered-arc and major-arc rigidity scheme should transfer to multi-frequency Weyl sums that obstruct multi-parameter ergodic theorems; the model phase t1(ξ1+ξ2 t2^2) is the natural first test case.","A plausible next step is the general polynomial-mapping version of the problem, where Newton polyhedra and coefficient dependence will replace the pure monomial parity rule.","The ℓ^p result suggests that multi-parameter pointwise ergodic averages for commuting transformations with polynomial steps should converge for all 1<p<∞ whenever a similar parity condition holds on the dominant monomials.","The log N divergence for forbidden monomials gives a quantitative obstruction: any generalization admitting even one monomial with two odd coordinates must introduce an additional cancellation mechanism."],"forward_implications":["For k=2 the discrete and continuous boundedness conditions coincide, but for k≥3 they diverge; the classical discrete-to-continuous comparison m_disc = m_cont + O(1) fails for polynomial phases.","Uniform boundedness of the multiplier gives, by Plancherel, ℓ^2 boundedness of the truncated transforms; Main Theorem 2 extends this to every 1<p<∞.","The necessity construction yields quantitative divergence: for any forbidden Λ there are coefficient choices and truncations with |H^Λ_N(ξ)| ≳ log N, so no constant depending only on Λ can control the sums.","The condition is checkable directly from Λ and is uniform in both the real coefficients and the independent truncation parameters N1,...,Nk."],"fun_headline_variants":["One odd coordinate max: exact test for Hilbert sums","Multi-parameter sums: bounded iff ≤1 odd exponent per monomial","Higher-dim circle method: Hilbert sums bounded iff ≤1 odd exponent","One odd coordinate rule for bounded Hilbert sums"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The sufficiency proof leans on an unproved lattice sublevel-set counting estimate (Lemma 3.4): for a polynomial of degree d on a dyadic box, the number of lattice points where |Ση_m t^m|≤ε is claimed to be at most C 2^{Σj} (ε/A)^{1/d}; should this bound require logarithmic losses or a weaker exponent, the E_sub decay estimates and with them the sufficiency half of Main Theorem 1 collapse.","fun_headline_variants_meta":{"raw":{"variants":["One odd coordinate max: exact test for Hilbert sums","Multi-parameter sums: bounded iff ≤1 odd exponent per monomial","Higher-dim circle method: Hilbert sums bounded iff ≤1 odd exponent","One odd coordinate rule for bounded Hilbert sums"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00124,"raw_usage":{"total_tokens":4933,"prompt_tokens":755,"completion_tokens":4178,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":499,"completion_tokens_details":{"reasoning_tokens":4110}},"tokens_in":499,"tokens_out":4178,"duration_ms":28507,"temperature":1.0,"reasoning_tokens":4110,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T01:25:41.732652+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct counting check of Lemma 3.4 would settle the proof: take the polynomial t1 t2 on the box [1,2^j]^2 and compare the claimed bound C 2^{2j} 2^{-r/2} with the actual number of pairs satisfying |t1 t2|≤2^{2j-r}; if a logarithmic factor appears, the lemma is false as stated and the E_sub estimates lose their summable decay. At the level of the theorem itself, a numerical search for Λ={(1,1)} in k=2 that kept sup_N,ξ |H^Λ_N(ξ)| bounded would refute the necessity half, since the paper proves divergence for this forbidden monomial.","supporting_citations":[],"review_version":2}