{"id":"145ce4be-19b3-46f3-8d53-630c71288155","arxiv_id":"2608.09009","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For nondegenerate curves in R^3, rational points of height Q within δ of the curve number at most Cδ^2Q^2 plus Q^(4/3+ε), and the folklore conjecture for codimension at least two is false.","lead":"This paper counts rational points near smooth curves in three-dimensional space and proves an essentially optimal bound. It also shows a widely believed conjecture fails for manifolds of higher codimension, with the moment curve as an explicit counterexample.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 8.3's partition-of-unity identity (8.5) is algebraically false; the missing overlap term is not estimated, so Theorem 4.12 lacks a proof as written.","rationale":"The central claims of the paper—the near-sharp upper bound for rational points near nondegenerate space curves and the disproof of Huang's conjecture for codimension at least two—are supported by a long but largely coherent argument. The lower bound in Theorem 1.1 is elementary and sound. The upper bound is structured around the quantitative stationary phase Theorem 4.12, whose proof contains the flawed partition-of-unity identity in Lemma 8.3. The reader identified exactly this gap, and my independent reading confirms it: the displayed identity (8.5) is algebraically false, and the unaccounted overlap term is not estimated elsewhere in Section 8. Because Lemma 6.2 relies on Theorem 4.12, and Proposition 5.3 relies on Lemma 6.2, this gap currently blocks the proof of Theorem 1.2 as written. I do not see a reason to escalate to rejection: the scale condition (4.54) appears designed precisely to ensure that the two critical points are separated by more than the bump width, and with the correct partition of unity the extra term should be controllable. The proposed concrete test—checking the support separation and, if necessary, estimating the overlap integral—would resolve the issue. The verdict should therefore remain CONDITIONAL, as the reader recommended, pending this localized repair.","tokens_in":47420,"tokens_out":4830,"duration_ms":43854,"concrete_test":"Re-derive Lemma 8.3 with the corrected partition 1 = A + (1−A)(1−B) + B − AB, where A = η((t−θ⁻₁)/α) and B = η((t−θ⁺₁)/α). Using Lemma 4.10 and condition (4.54), check whether α ≤ c 2^{−ℓ} for a fixed small constant c; if yes, AB ≡ 0 and the lemma follows by the existing estimates. If no, bound the overlap integral ∫ e(λφ_ξ(t)) a(t) η((t−τν₀)/τ) AB dt by nonstationary phase, using the derivative lower bound |φ'_ξ(t)| ≳ |t−θ₂|² − O(2^{−2ℓ}) away from θ₂. This settles whether the proof of Theorem 4.12 can be repaired at the indicated point.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Lemma 8.3 in Section 8 claims the identity 1 = η((t−θ⁻₁)/α) + (1−η((t−θ⁺₁)/α))(1−η((t−θ⁻₁)/α)) + η((t−θ⁺₁)/α), with α = λ_ℓ^{−1/2+ε}. Expanding the middle product gives RHS = 1 + η((t−θ⁻₁)/α)η((t−θ⁺₁)/α), so the identity is false unless the two bumps have disjoint supports. This identity is used to split I(λ,ξ) into I₊ + I₋ plus a remainder r_ℓ; the missing cross term is not included in r_ℓ nor estimated. Lemma 8.3 feeds directly into Lemma 8.4 and hence Theorem 4.12, which is invoked in Lemma 6.2 to pass to the dual counting problem in Proposition 5.3. Thus the main upper bound currently relies on an unproved overlap estimate. The gap is plausibly fixable: Lemma 4.10 gives |θ⁺₁ − θ⁻₁| ≍ 2^{−ℓ}, and under (4.54) one can show α = λ_ℓ^{−1/2+ε} = o(2^{−ℓ}) (in fact at most λ^{−cε}2^{−ℓ} for large λ), so the supports of the two η factors are disjoint. But this separation argument is not written, and the displayed algebra as stated is wrong.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the counting function N_C(δ,Q) for rational points of height at most Q lying within δ/q of a compact C∞ curve C⊂R^3 with nonvanishing curvature and torsion. Its main upper bound, Theorem 1.2, asserts N_C(δ,Q) ≤ Cδ^2Q^2 + O_ε(Q^{4/3+ε}), which improves Huang's earlier Q^{8/5} bound and matches the lower bound up to endpoints in the range δ > Q^{-1/3+ε}. The paper also proves Theorem 1.1, an elementary lower bound N_{M_{m,n}}(δ,Q) ≳ δ^{1/2} Q^{n-m+1/2} for cylinders over the moment curve, which contradicts Huang's Conjecture 3.1 for codimension m≥2. The proof strategy combines a truncated Fourier expansion, a decomposition of frequencies according to distance to the binormal cone Γ, a quantitative stationary phase expansion (Theorem 4.12), and a reduction to a dual counting problem for the tangent developable surface S*; the dual count is ultimately bounded using the external Vaughan–Velani planar theorem.","tokens_in":47614,"tokens_out":18802,"duration_ms":173202,"significance":"If the proof is completed, the results are significant. Theorem 1.2 gives the essentially sharp exponent for rational points near nondegenerate space curves in R^3, and Theorem 1.1 provides a clean, elementary counterexample to a plausible general conjecture, with the 'major arc' obstruction clearly identified. The paper also introduces a reusable structural framework: dyadic decomposition near the slow-decay cone, a quantitative stationary phase expansion with uniform remainders, and a Legendre-dual counting argument adapted to a nonsmooth dual variety. The lower-bound argument in Section 2 is self-contained and checkable. The upper-bound argument is not circular: the dual counting problem is reduced back to the original curve count, but that count is then bounded using the external Vaughan–Velani planar theorem rather than the paper's own results. These strengths make the paper worth pursuing; however, one load-bearing proof step in the stationary phase section is currently incorrect as written.","major_comments":[{"comment":"The displayed partition identity is algebraically false. Expanding (1−η((t−θ+1)/α))(1−η((t−θ−1)/α)) gives 1 − η+ − η− + η+η−, so the right-hand side of (8.5) equals 1 + η((t−θ−1)/α)η((t−θ+1)/α), not 1. The term rℓ(t,ξ) defined immediately afterward omits this cross term, and no estimate for it is supplied. This is load-bearing because Lemma 8.3 is used in the proof of Theorem 4.12, and Theorem 4.12 is invoked in Lemma 6.2 to pass to the dual counting problem in Proposition 5.3; thus the main upper bound currently rests on an unproved overlap estimate. The gap appears fixable: Lemma 4.10 gives |θ+1−θ−1| ≍ 2^{−ℓ}, and condition (4.54) implies α = λ_ℓ^{−1/2+ε} = o(2^{−ℓ}) for large λ, so the supports of the two η factors are disjoint and the cross term vanishes. However, this separation argument is not written, and the displayed algebra as stated is wrong.","section":"Section 8, Lemma 8.3, Eq. (8.5)"}],"minor_comments":[{"comment":"Theorem 4.12 states the remainder amplitude as W^σ_ℓ((λ2^{−ℓ})^ε, ξ), while Lemma 8.4 concludes with A^σ_ℓ(λ_ℓ^{2ε}, ξ). These two normalizations should be reconciled; since ε is arbitrary, the discrepancy is harmless, but as written the notation is inconsistent.","section":"Theorem 4.12 and Lemma 8.4"},{"comment":"The notation N_{C_m} in Lemma 2.1 should be N̄_{C_m}, the pair count, because the proof counts pairs (a,q) before the Möbius-inversion step in Lemma 2.2; as written, the statement and proof use different counting functions.","section":"Section 2, Lemmas 2.1 and 2.2"},{"comment":"Expressions such as √δQ are typeset ambiguously; \\(\\sqrt{\\delta Q}\\) should be used throughout so that it is not read as \\((\\sqrt{\\delta})Q\\).","section":"Section 2, display (2.1)"},{"comment":"The sentence 'the implicit constant depends on φ, a and N' should specify whether the estimate is uniform in ξ and ℓ; the argument appears to give uniformity, but the statement should say so explicitly.","section":"Section 8, proof of Lemma 8.2"}],"recommendation":"major_revision","confidential_remarks":"The technical gap in Lemma 8.3 is local but load-bearing; it affects the proof of Theorem 4.12 and hence the main upper bound. I believe it is fixable with a short separation argument, so I recommend major revision rather than rejection. No concerns about novelty or scope arose in my reading."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a strong paper. It settles the order of magnitude for rational points near nondegenerate space curves in R^3: the upper bound in Theorem 1.2 improves Huang's Q^{8/5} (up to logs) to Q^{4/3+epsilon}, and the lower bound in Theorem 1.1 for the moment curve cylinders disproves Huang's conjecture for codimension at least two. The lower bound is elementary and sound, and the overall architecture of the upper-bound proof is coherent: reduce to a fixed-scale perturbation of the moment curve, decompose frequencies by distance to the binormal cone, pass to a Legendre dual counting problem, and invoke the Vaughan–Velani planar theorem for the final count. The paper is honest about what is new and credits the relevant literature, including Huang, Beresnevich–Yang, Hickman–Srivastava, and Schindler–Yamagishi.\n\nThe soft spot is real. Lemma 8.3, which feeds directly into Theorem 4.12, the stationary-phase expansion used to pass to the dual counting problem, claims the identity (8.5):\n\n1 = eta((t-theta^-)/alpha) + (1-eta((t-theta^+)/alpha))(1-eta((t-theta^-)/alpha)) + eta((t-theta^+)/alpha).\n\nExpanding the middle term gives an extra eta((t-theta^-)/alpha)eta((t-theta^+)/alpha), so the identity is false unless the two bumps have disjoint supports. The missing overlap term is not included in the remainder r_l and is not estimated. Since Lemma 8.3 is used in Lemma 8.4 and hence in Theorem 4.12, which is invoked in the proof of Proposition 5.3, the main upper bound currently rests on an unproved estimate. The gap is localized and plausibly fixable: Lemma 4.10 gives |theta^+ - theta^-| ~ 2^{-ell}, and under the scale condition (4.54) one has alpha = lambda_ell^{-1/2+epsilon} = o(2^{-ell}), so the supports are in fact disjoint; the separation argument is just not written. Still, as the paper stands, Theorem 4.12 lacks a complete proof.\n\nThis is a conditional accept situation, not a reject. The surrounding argument gives strong evidence that the intended approach works, and the flaw is confined to one step. A few smaller presentation issues (several omitted 'straightforward' modifications, a slightly overbroad abstract claim about folklore conjectures being incorrect for codimension at least two when the counterexample is exhibited for specific cylinders) do not change the picture. The paper deserves a serious referee, and I would send it out with the expectation that Section 8 needs fixing. The reader's verdict of CONDITIONAL with moderate confidence matches my own reading.","headline":"Important, likely correct result with a real but probably fixable gap in Lemma 8.3; worth refereeing carefully.","tokens_in":48288,"tokens_out":1980,"would_cite":true,"duration_ms":20193,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11J83","11J13","11J25","42B20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves essentially sharp upper bounds for rational points near non-degenerate space curves in R^3 and shows that a folklore counting conjecture fails for codimension at least two.","keywords":["rational points near curves","Diophantine approximation","space curves","nonvanishing curvature and torsion","stationary phase","slow decay cone","tangent developable surface","major arc obstruction"],"falsifier":"Verify Lemma 8.3’s decomposition for the normalized moment curve γ(t)=(t,t²/2,t³/6) with λ=$2^{{3ℓ}}$ and u(ξ)≍$2^{{−2ℓ}}$; if the stated equality I(λ,ξ)=I^+(λ,ξ,ℓ)+I^−(λ,ξ,ℓ)+O($λ^{{−N}}$) is not reproduced, or the claimed separation of the two critical points from the bump width fails, then Theorem 4.12—and hence the proof of the upper bound—is not currently supported. A separate computational test would be to count rational points for a specific nondegenerate curve at δ≈$Q^{{−1/3}}$ and check whether the count follows δ²Q² plus an error of size $Q^{{4/3+ε}}$.","tokens_in":1954,"feed_emoji":"📐","tokens_out":1922,"duration_ms":114586,"temperature":0.7,"pith_summary":"This paper counts rational points a/q of height q≤Q lying within distance δ/q of a smooth space curve in $R^{3}$ with nonvanishing curvature and torsion. It proves the essentially optimal upper bound N_C(δ,Q) ≤ C δ²Q² + O_ε($Q^{{4/3+ε}}$), together with an asymptotic for a smoothed count whose leading term is c δ²Q². It also exhibits a major-arc obstruction: for cylinders over the moment curve in codimension at least two, the number of near rational points is at least $δ^{{1/2}}$$Q^{{n−m+1/2}}$ in a range where previous conjectures predicted a smaller number. If correct, the results settle the sharp order of rational points near 3-nondegenerate space curves and disprove the folklore conjecture except in the hypersurface case.","feed_headline":"Space curves: rational point count sharpened to δ²Q² + Q^{4/3}","feed_subtitle":"The result is essentially optimal and shows a folklore conjecture fails for codimension at least two.","key_machinery":"The central object is the slow-decay (binormal) cone Γ = {θ : γ′(t)·θ = γ″(t)·θ = 0 for some t}, together with the function u(ξ) = φ′_ξ(θ_2(ξ)), where θ_2 locates the minimum of the phase derivative; u(ξ) measures signed distance to Γ. The argument decomposes the nonzero frequencies dyadically by |u(ξ)| ≍ $2^{{−2ℓ}}$; near Γ, the phase has two critical points separated by ≍ $2^{{−ℓ}}$, and Theorem 4.12 supplies a stationary-phase expansion with uniformly controlled remainders at every dyadic scale. This reduces each dyadic piece of the original count to a dual count of integer frequencies near the tangent developable S*, which is then bounded using a planar rational-point theorem and an involutive Legendre-transform reduction back to the original curve.","core_discovery":"On the paper's own terms: the number of rational points of height ≤Q in a δ/q-neighborhood of a compact C∞ curve in $R^{3}$ with nonvanishing curvature and torsion is O_ε(δ²Q² + $Q^{{4/3+ε}}$), and for δ ≥ $Q^{{-1/3+ε}}$ the count is comparable to δ²Q². The proof passes from the original counting problem to oscillatory sums, then to a dual counting problem near a tangent developable surface S* in frequency space. The key mechanism is a dyadic stationary-phase expansion near the slow-decay cone Γ, where the phase has two nearby critical points and the Fourier transform of the curve decays slowest. The same mechanism reveals a previously hidden major-arc obstruction: cylinders over moment curves have far more near rational points than the conjectured δ^m $Q^{{n−m+1}}$, so the codimension-m≥2 form of the conjecture fails.","pith_inferences":["A natural next step is to extend the dyadic slow-decay-cone analysis to nondegenerate curves in R^n; the structure of the proof suggests dimension-dependent exponents and analogous major-arc obstructions, although the paper proves only the case n=3.","The failure of the codimension≥2 conjecture suggests that the sharp exponent for rational points near a manifold is controlled by the slowest Fourier-decay directions of its surface measure rather than by codimension alone; this reading goes beyond what the paper states explicitly.","A testable outgrowth would be to count rational points near the tangent developable S* directly for a cubic model curve and compare the resulting exponent with the dual-count estimate (5.8); agreement would confirm that the Legendre-involution step is lossless."],"forward_implications":["For every compact C∞ curve in R^3 with nonvanishing curvature and torsion, N_C(δ,Q) ≤ C δ²Q² + O_ε(Q^{4/3+ε}); in the range δ ≥ Q^{-1/3+ε}, the two-sided bound c δ²Q² ≤ N_C(δ,Q) ≤ C δ²Q² holds.","The folklore conjecture (Conjecture 3.1) is false for codimension m≥2: for cylinders over the moment curve, N_{M_{m,n}}(δ,Q) ≳ δ^{1/2}Q^{n−m+1/2} whenever A Q^{ε−1} ≤ δ ≤ 1/2.","Because the conjectured uniform bound fails in this range, sharp counting near a special manifold cannot by itself imply the convergence case of Khintchine's theorem; generic/special decompositions are necessary rather than merely convenient.","The earlier estimate N_C(δ,Q) ≲ δ²Q² + Q^{8/5}(log Q)^{4/5} for such curves is improved to an error term of size Q^{4/3+ε}.","The smooth asymptotic N_{Ω,C}(δ,Q) = c_Ω δ²Q² + O_ε(δ^{4/3−ε}Q^{5/3+3ε} + δ^{1/2−ε}Q^{3/2+ε}) holds for δ in (Q^{ε−1/2}, 1/2), giving an explicit leading constant."],"supporting_citations":[{"why":"Supplies the planar rational-point theorem used as the main counting input in the dual problem and in Lemma 4.14.","marker":"[32]"},{"why":"Provides the sharp hypersurface model and the conjectural framework that the paper refutes for codimension at least two.","marker":"[19]"},{"why":"Gives the prior counting result for integral points near 3-nondegenerate space curves that Theorem 1.2 improves.","marker":"[18]"},{"why":"Handles the model case of frequencies whose directions stay away from the slow-decay cone; Proposition 5.1 follows this pattern.","marker":"[27]"},{"why":"Provides the finite integral-affine normalization and the geometric lemmas about θ_2, u, and critical-point separation used throughout the decomposition.","marker":"[4]"},{"why":"Supplies earlier lower bounds near manifolds and the moment-curve example that motivates and is contrasted with Theorem 1.1.","marker":"[5]"},{"why":"Gives the lower bound N_C(δ,Q) ≳ δ²Q² used to turn the upper bound into Corollary 1.3.","marker":"[26]"},{"why":"Provides the Möbius inversion and zeta identities used in the elementary lower-bound proof for cylinders over the moment curve.","marker":"[13]"},{"why":"Explains the divide-and-conquer Khintchine-type approach whose necessity is illuminated by the major-arc obstruction.","marker":"[7]"},{"why":"Provides the earlier dual-counting approach for integral points near n-nondegenerate curves that the paper refines by counting near the slow-decay cone.","marker":"[14]"}],"fun_headline_variants":["Optimal rational point count near space curves","Folklore conjecture fails for rational points near curves","Sharp bounds: δ²Q² + Q^{4/3} near space curves","Hidden major arc: more rational points near moment curve","Rational near points: counterexample to codimension-2 conjecture"],"cache_read_input_tokens":50304,"weakest_assumption_plain":"The load-bearing premise is that Theorem 4.12’s stationary-phase expansion, with its O($λ^{{−N}}$) remainder uniform in the dyadic scale ℓ, is a valid theorem; because the displayed identity in Lemma 8.3 used in its proof is false as written, the upper bound stands only if that lemma’s intended separation argument repairs the step.","fun_headline_variants_meta":{"raw":{"variants":["Optimal rational point count near space curves","Folklore conjecture fails for rational points near curves","Sharp bounds: δ²Q² + Q^{4/3} near space curves","Hidden major arc: more rational points near moment curve","Rational near points: counterexample to codimension-2 conjecture"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000241,"raw_usage":{"total_tokens":1486,"prompt_tokens":874,"completion_tokens":612,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":490,"completion_tokens_details":{"reasoning_tokens":530}},"tokens_in":490,"tokens_out":612,"duration_ms":6581,"temperature":1.0,"reasoning_tokens":530,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:20:08.877997+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Verify Lemma 8.3’s decomposition for the normalized moment curve γ(t)=(t,t²/2,t³/6) with λ=$2^{{3ℓ}}$ and u(ξ)≍$2^{{−2ℓ}}$; if the stated equality I(λ,ξ)=I^+(λ,ξ,ℓ)+I^−(λ,ξ,ℓ)+O($λ^{{−N}}$) is not reproduced, or the claimed separation of the two critical points from the bump width fails, then Theorem 4.12—and hence the proof of the upper bound—is not currently supported. A separate computational test would be to count rational points for a specific nondegenerate curve at δ≈$Q^{{−1/3}}$ and check whether the count follows δ²Q² plus an error of size $Q^{{4/3+ε}}$.","supporting_citations":[{"cited_title":"1, 103–124","cited_arxiv_id":null,"evidence_quote":"Supplies the planar rational-point theorem used as the main counting input in the dual problem and in Lemma 4.14."},{"cited_title":"J.169(2020), no","cited_arxiv_id":null,"evidence_quote":"Provides the sharp hypersurface model and the conjectural framework that the paper refutes for codimension at least two."},{"cited_title":"Ann.374(2019), no","cited_arxiv_id":null,"evidence_quote":"Gives the prior counting result for integral points near 3-nondegenerate space curves that Theorem 1.2 improves."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Handles the model case of frequencies whose directions stay away from the slow-decay cone; Proposition 5.1 follows this pattern."},{"cited_title":"Math.393(2021), Paper No","cited_arxiv_id":null,"evidence_quote":"Provides the finite integral-affine normalization and the geometric lemmas about θ_2, u, and critical-point separation used throughout the decomposition."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies earlier lower bounds near manifolds and the moment-curve example that motivates and is contrasted with Theorem 1.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Möbius inversion and zeta identities used in the elementary lower-bound proof for cylinders over the moment curve."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Explains the divide-and-conquer Khintchine-type approach whose necessity is illuminated by the major-arc obstruction."},{"cited_title":"14, rnaf200","cited_arxiv_id":null,"evidence_quote":"Provides the earlier dual-counting approach for integral points near n-nondegenerate curves that the paper refines by counting near the slow-decay cone."}],"review_version":1}