{"id":"d50f8305-4ee1-4cf1-b3ac-8fbb1bd6f5e5","arxiv_id":"2507.21401","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"On the Veronese curve in R^3, the set of simultaneously lambda-well approximable points has Hausdorff dimension (2-2lambda)/(1+lambda) for all 1/3 <= lambda <= 3/5.","lead":"This paper computes the exact size, in the sense of Hausdorff dimension, of the set of points on the standard cubic curve in three-dimensional space that are simultaneously approximable by rational vectors with a given quality. A generalist might read it because it settles a benchmark range predicted by the Beresnevich-Yang conjecture and introduces a new counting technique for cubic polynomials.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The upper bound depends on the imported counting estimates (9)–(10) from [1], which are not re-proved; a wrong exponent in (10) would invalidate the D3/D4 bounds and hence the dimension formula.","rationale":"The reader's weakest_assumption matches the point I would stress. The proof is an upper-bound machine; each branch is a product of the number of intervals J, the number of q per J, and the number of admissible polynomials. The first two factors come directly from (9)–(10), while the new contribution (Theorem 2, Proposition 1, Lemma 4) supplies the polynomial count and the zero-discriminant case split. I re-checked the exponent arithmetic in §§5–8 and found it consistent, with the maxima occurring at the stated boundary values of δ, η, κ, σ; no internal contradiction emerges if (9)–(10) hold. The genuine risk is that (10) silently omits a factor from the union over boxes sharing a common J, or that (9) misses a dependence on the number of possible hyperplanes. Because the endpoint λ=3/5 is saturated—the D3 bound is 1/2 and the S4 bound is exactly the target—any positive Q^c factor in these imports would break the theorem, while the existing ε absorbs only vanishing discrepancies. I also considered the intricate D12 emptiness argument in §8; there is a delicate point about a lower bound on |t|, but the determinant bounds shrink as |t| decreases, so that case appears repairable and is not the weakest link. The reader's CONDITIONAL verdict is the right calibration in the absence of the deferred proofs.","tokens_in":20118,"tokens_out":30454,"duration_ms":347956,"concrete_test":"Derive (10) independently from the geometry of ∆m: fix a dyadic box with τ4=Q^δ, an interval J of length Q^{−(1+λ)/2−η}, and the parameter η≥0, and count integer vectors q∈∆m whose associated polynomial satisfies (9) and whose balls lie in J. The count should be compared against Q^{(3−5λ)/2+2δ+2ε}. Work the extremal case λ=3/5, δ=0, η=0: if the number of q per J is not Q^{O(ε)} but grows like Q^c for some fixed c>0, then the Section 5 bound dim S3≤1/2 and the Section 7–8 bounds dim S_i≤(2−2λ)/(1+λ) both fail, so Theorem 1 at the endpoint collapses. Equivalently, inspect the proofs of equations (20)–(21) in [1] and confirm that the union over candidate hyperplanes and over sub-boxes inside J introduces no additional Q-power beyond the displayed ε.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is an upper bound, and every upper-bound branch (S3 in §5, S4–S6 in §6, S7–S9 in §7, S10–S11 in §8) is controlled by two inequalities imported from [1]: (9), bounding the height of the associated polynomial by Q^{λ−η−δ+ε} (or Q^{λ−δ+ε} when η<0), and (10), bounding the number of q sharing one interval J by Q^{(3−5λ)/2+2δ+2ε} (or Q^{(3−5λ)/2+2δ−η+3ε}). These estimates generate the factors Q^{2δ}, Q^{−η} and the constant (3−5λ)/2 that appear in every subsequent exponent calculation, including the application of Theorem 2 and the D8/D9/D10/D11 bounds. The paper states them with a reference to [1] and does not reproduce their proofs. If the exponent in (10) is off by a positive constant c, for example because the union over candidate hyperplanes a or over sub-boxes inside J introduces an additional Q^c factor, then the final inequalities #D_i ≤ Q^{2−2λ+O(ε)} would acquire that extra factor and would fail at λ=3/5, where the target (2−2λ)/(1+λ)=1/2 is met exactly. I checked the new sections for internal exponent consistency assuming (9)–(10) hold; the bookkeeping is coherent. The load-bearing point is therefore the unverified import, exactly as the reader concluded.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that for the Veronese curve V_3 = {(x,x^2,x^3)} in R^3, the Hausdorff dimension of the set S_3(I,λ) of simultaneously λ-well approximable points equals (2−2λ)/(1+λ) for every λ ∈ [1/3, 3/5]. This matches the range predicted by the Beresnevich–Yang conjecture for nondegenerate curves. The proof is an upper-bound argument that partitions the relevant rational vectors into a hierarchy of cases: nonzero-discriminant polynomials are treated via a new counting result for cubic polynomials (Theorem 2), and zero-discriminant cases are split into several geometric subcases controlled by the parameters δ, η, κ, σ, δ*. The paper is a continuation of the author's earlier work [1], and several central counting estimates are imported from that paper.","tokens_in":20472,"tokens_out":30368,"duration_ms":308267,"significance":"If the proof is correct, Theorem 1 provides the first verification of the Beresnevich–Yang dimension formula for a nondegenerate curve in R^n, n ≥ 3, up to the conjectured threshold 3/(2n−1), with n = 3. The paper also contributes a new counting result for cubic integer polynomials with small discriminant and large root spread (Theorem 2), which is a substantial technical tool. The exposition is structured and the case analysis is detailed; the exponent bookkeeping in Sections 5–8 is internally coherent once the imported estimates (9)–(10) are granted. However, the main theorem is conditional on those imported estimates, and the lower-bound half of the equality is not explicitly established in this paper.","major_comments":[{"comment":"The two counting estimates (9) and (10), imported from [1] without proof, are load-bearing for every upper-bound branch: they control S3 in §5, S4–S6 in §6, S7–S9 in §7, and S10–S11 in §8. In particular, the exponent (3−5λ)/2 in (10) appears in every subsequent estimate, and a multiplicative Q^c loss in that formula would destroy the final bound at λ = 3/5, where the target exponent (2−2λ)/(1+λ) = 1/2 is met exactly. The paper does not state the hypotheses under which (9) and (10) are valid, nor the uniformity of the implicit constants in ε, δ, and η. Please reproduce the proofs of these estimates in an appendix, or state them as formal lemmas with complete hypotheses and precise references to the corresponding results in [1].","section":"§3, Eqs. (9)–(10)"},{"comment":"The proof of Proposition 1 contains the unproved assertion that the number of pairs (a,b) with |a| ≤ A and |y1 + b/a| ∈ [|a|^{-t−ε}, |a|^{-t}] is bounded by ≪ A^{2−t}. This estimate is used directly to obtain (16), which in turn yields Theorem 2 and the D3 estimate. The proposition also relies on [1, Lemma 9] for the root-separation property |x_i − x_j| ≫ 1 of the reduced form R_a, and that separation lemma is not stated. These are load-bearing gaps in the new part of the proof; please supply the missing counting argument and a precise statement of the imported separation lemma.","section":"§5, Proposition 1"},{"comment":"The proof as written establishes upper bounds for the Hausdorff dimension of the sets S_j; it never explicitly proves the matching lower bound dim S_3(I,λ) ≥ (2−2λ)/(1+λ). If the lower bound for the full range 1/3 ≤ λ ≤ 3/5 is available from [1] or from Beresnevich's theorem, it should be stated explicitly and its hypotheses checked. If it is not known for λ > 1/2, then the equality in Theorem 1 is not justified. Please add a clear statement of the lower-bound theorem used and explain how it covers the range [1/3, 3/5].","section":"Theorem 1 and §3–§8"}],"minor_comments":[{"comment":"The displayed inequality \"H ≍ |P'''(x)| ≪ Q^{−(1+λ)/2 + 3η}\" appears to have a sign error: equation (12) with i = n = 3 gives the exponent +(1+λ)/2 + 3η, and it is that positive exponent which yields the stated bound η ≳ −(1+λ)/6. Please correct the displayed sign.","section":"§6, after Eq. (20)"},{"comment":"In the summation over discriminants, the intermediate expression \"min{H^{2/3+ε} d^{−1/6}, log H · R^{1+ε}}\" should have R, not R^{1+ε}, if it is quoting Proposition 1; the final bound in (19) is correct, but the intermediate line should be adjusted for consistency.","section":"§5, Theorem 2 proof"},{"comment":"The sentence \"If M is one dimensional ... the above inequality is achieved for 1/n ≤ λ ≤ 3/(2n−1)\" is ambiguous: it is not clear whether this means the lower bound is known to hold throughout that range or that equality is known. Since the conjecture is still open, please clarify that the lower bound is the known part and the present paper supplies the matching upper bound.","section":"§1, Introduction"},{"comment":"The abstract says the result \"confirm[s] the lower bound part of this conjecture,\" but the proof establishes an upper bound. If the lower bound is already known, it would be more accurate to say the paper confirms the full dimension formula for V3 in this range; please rephrase to avoid confusion about what is new.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the reliance on the imported estimates (9)–(10) from [1]; if those estimates are correct, the paper's case analysis appears coherent. The lower-bound issue in Theorem 1 must also be resolved, either by an explicit citation or by a proof. This is a technically substantial paper and the new counting theorem is interesting, but the manuscript should be self-contained in its central estimates before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline is simple: this paper reaches the conjectured threshold 3/5 for the Veronese curve in R^3, giving the first nondegenerate curve in n≥3 for which the lower-bound half of the Beresnevich–Yang conjecture is confirmed. That is a genuine within-subfield advance, not a cosmetic extension of [1]. The genuinely new ingredients are the root-separation parameter kappa and the counting theorem for cubic polynomials (Theorem 2); both appear new and the proof of Proposition 1 is included in full. The internal exponent bookkeeping in Sections 5, 7 and 8 is coherent. I spot-checked several of the asserted maximisations and they do work, though the paper compresses them heavily.\n\nThe soft spot is exactly where the reader put it: the two imported inequalities (9) and (10) from [1] control every upper-bound branch. They bound the height of the associated polynomial and the number of q sharing one interval J, and they are not re-proved here. If either exponent is off by a positive constant, the final dimension formula fails at the endpoint, where the target 1/2 is met exactly. That said, this is reliance on the author's own prior published result, not circularity or fitting, and the paper clearly flags the imports. The bigger practical issue is that a referee who wants to certify Theorem 1 cannot do so by reading this paper alone; they must also verify equations (20)–(21) of [1]. Some readers will also be annoyed by the asserted rather than derived exponent bounds in Sections 7 and 8, though they are fairly routine and I did not find a counterexample.\n\nIn proportion: the core argument is plausible and internally consistent, the new counting theorem is a real contribution, and the dependence on [1] is explicit and legitimate. This is not a flawed paper; it is a paper that outsources a load-bearing estimate. The right outcome is to send it to referees who are willing to check [1] as part of the review. For experts in metric Diophantine approximation this deserves close reading; for the broader number theory audience it is a solid progress report.\n\nRecommendation: yes, send to peer review, with a referee who knows [1] and can verify (9)–(10).","headline":"A real step to the conjectured endpoint 3/5 for the Veronese curve in R^3, with the proof's weight resting on counting estimates imported from the author's earlier paper—worth a serious referee.","tokens_in":20997,"tokens_out":1386,"would_cite":true,"duration_ms":18305,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11J13","11J54","11J82","11K55"],"pacs":[],"model":"deepseek-v4-flash","headline":"For every λ between 1/3 and 3/5, the Hausdorff dimension of simultaneously λ-well approximable points on the three-dimensional Veronese curve equals (2−2λ)/(1+λ).","keywords":["simultaneous Diophantine approximation","Veronese curve","Hausdorff dimension","nondegenerate curves","cubic polynomials","bounded discriminant","successive minima","metric Diophantine approximation"],"falsifier":"For $\\lambda=3/5$ and a large dyadic scale $Q=2^k$, enumerate the primitive integer vectors $q$ inside one of the boxes $\\Delta_m$ used in the proof and compare their number with the right-hand side of estimate (10); a single box that exceeds that bound would invalidate the exceptional-box count and with it the upper bound behind Theorem 1.","tokens_in":19934,"feed_emoji":"📐","tokens_out":15125,"duration_ms":159008,"temperature":0.7,"pith_summary":"The paper proves that for every $\\lambda$ in the interval $1/3 \\le \\lambda \\le 3/5$, the set of points on the three-dimensional Veronese curve $V_3 = \\{(x, x^2, x^3)\\}$ that are simultaneously $\\lambda$-well approximable by rational points has Hausdorff dimension $(2-2\\lambda)/(1+\\lambda)$. The endpoint $3/5$ is exactly the threshold that the leading conjecture for nondegenerate curves predicts for this dimension formula, so the result verifies the conjectured value on a concrete curve throughout the predicted interval. The author states that $V_3$ is the first nondegenerate curve in ambient dimension at least three for which the lower-bound half of that conjecture is confirmed in this range. Readers working on metric Diophantine approximation on manifolds care because $V_3$ is the natural test object: if the formula failed anywhere below $3/5$, the conjectured threshold would be wrong.","feed_headline":"Veronese curve hits predicted approximation dimension for λ up to 3/5","feed_subtitle":"It is the first nondegenerate curve in three dimensions to verify the conjectured lower-bound range.","key_machinery":"The central counting device is Theorem 2, which bounds the number of cubic polynomials with height at most $H$, discriminant $0 < |D(P)| \\le D$, and root separation at least $R^{-1}$ by $\\ll \\min\\{H^{2/3+\\epsilon}D^{5/6},\\log H \\cdot D R\\} + H^\\epsilon D$. The count is assembled from equivalence classes of cubics under fractional linear transformations, using the number of classes with a given discriminant as an input. The surrounding proof uses dyadic blocks, the boxes $\\Delta_m$ defined by closeness to tangent segments of the curve, the successive minima of these boxes, and parameters $\\delta,\\eta,\\kappa$ that measure how far a box is from the generic case. For zero discriminant the polynomial takes the form $(ax-b)^2(cx-d)$, and the remaining work is a partition by which linear factor's root lies in the associated interval $J$.","core_discovery":"The central claim is Theorem 1: for $V_3 = \\{(x, x^2, x^3) : x \\in I\\}$ with $I$ a closed interval not containing zero, and for every $1/3 \\le \\lambda \\le 3/5$, $\\dim S_3(I,\\lambda) = (2-2\\lambda)/(1+\\lambda)$, where $S_3(I,\\lambda)$ is the restriction to $V_3$ of the set of $x \\in \\mathbb{R}^3$ satisfying $\\|qx - p\\|_\\infty < q^{-\\lambda}$ for infinitely many integer pairs $(q,p)$. The proof supplies the upper bound needed for the formula; the matching lower bound for this range is already available from the theory of rational points near manifolds, so the equality follows. Grouping rational vectors into dyadic blocks in which the denominator $q$ lies between consecutive powers of two, the argument partitions each block by how close the rational points are to tangent segments of the curve. Each remaining case is then controlled by counting the associated cubic polynomials $P_a(x) = a_0 + a_1 x + a_2 x^2 + a_3 x^3$ whose height is bounded in terms of the block parameters. Polynomials with nonzero discriminant are counted by a new estimate for cubic polynomials with bounded height, bounded discriminant, and large root separation; polynomials with zero discriminant factor as $(ax-b)^2(cx-d)$ and are handled by a case analysis according to which root lies in the relevant approximation interval. The paper states that this makes $V_3$ the first nondegenerate curve in $\\mathbb{R}^n$, $n \\ge 3$, to confirm the lower-bound half of the conjecture in this range.","pith_inferences":["The polynomial-counting estimate used here is not tied to the Veronese parametrization; a similar count should apply to any nondegenerate curve in $\\mathbb{R}^3$ whose local approximating polynomials are cubic, which would widen the verified range beyond $V_3$.","The case analysis isolates the barrier at $\\lambda=3/5$: intervals that contain very simple rationals (denominator $\\ll H^{1/3}$) are the ones that just barely fit inside the dimension bound. A proof for $\\lambda>3/5$ would need a new way to handle exactly those intervals.","A numerical experiment at $\\lambda=3/5$ could probe whether the threshold is sharp: the proof predicts that the number of rational vectors in each exceptional box grows like $Q^{2-2\\lambda+o(1)}$, and a measurable excess on dyadic scales would suggest the conjectured range is not optimal."],"forward_implications":["For all $\\lambda$ in $[1/3,3/5]$, the Hausdorff dimension on $V_3$ is exactly $(2-2\\lambda)/(1+\\lambda)$, falling from $1$ at $\\lambda=1/3$ to $1/2$ at $\\lambda=3/5$.","The known range for $V_3$ is extended from $\\lambda \\le 1/2$ to the full interval predicted for the conjecture, so the conjectured threshold $3/(2n-1)$ is attained for $n=3$ on this curve.","The zero-discriminant analysis shows that the only delicate configurations are intervals containing rationals with denominator at most $H^{1/3}$; all other exceptional boxes contribute less than the dimension bound.","Because the lower bound was already known, the equality on $V_3$ turns the conjecture's predicted value into a theorem for this test curve."],"supporting_citations":[{"why":"Supplies the dyadic-block setup and the two counting estimates (9)-(10) that bound polynomial heights and shared intervals; they are reused in every later case of this paper.","marker":"[1]"},{"why":"Provides the general rational-points-near-manifolds lower bound that fixes the target dimension formula for nondegenerate curves in the range considered.","marker":"[2]"},{"why":"States the conjecture that predicts threshold $3/(2n-1)$; the endpoint $3/5$ for $n=3$ is the range verified here.","marker":"[3]"},{"why":"Supplies the classical covering lemma used throughout to turn cardinality estimates into Hausdorff dimension upper bounds.","marker":"[4]"},{"why":"Gives the count of equivalence classes of binary cubic forms by discriminant, which is the input for the polynomial-counting theorem.","marker":"[5]"}],"fun_headline_variants":["Veronese curve achieves predicted dimension for λ in [1/3, 3/5]","First nondegenerate curve in R^3 verifies approximation conjecture","Cubic curve matches exact Hausdorff dimension for λ range","Veronese curve confirms lower-bound conjecture in R^3","Exact Diophantine approximation dimension found for cubic curve in R^3"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof leans on two counting estimates imported from the author's earlier work: a height bound on the polynomial attached to each rational vector and a bound on how many rational vectors can share one small interval; the paper does not re-derive these estimates.","fun_headline_variants_meta":{"raw":{"variants":["Veronese curve achieves predicted dimension for λ in [1/3, 3/5]","First nondegenerate curve in R^3 verifies approximation conjecture","Cubic curve matches exact Hausdorff dimension for λ range","Veronese curve confirms lower-bound conjecture in R^3","Exact Diophantine approximation dimension found for cubic curve in R^3"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0011,"raw_usage":{"total_tokens":4619,"prompt_tokens":1002,"completion_tokens":3617,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":618,"completion_tokens_details":{"reasoning_tokens":3521}},"tokens_in":618,"tokens_out":3617,"duration_ms":29344,"temperature":1.0,"reasoning_tokens":3521,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T12:49:52.184326+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For $\\lambda=3/5$ and a large dyadic scale $Q=2^k$, enumerate the primitive integer vectors $q$ inside one of the boxes $\\Delta_m$ used in the proof and compare their number with the right-hand side of estimate (10); a single box that exceeds that bound would invalidate the exceptional-box count and with it the upper bound behind Theorem 1.","supporting_citations":[{"cited_title":"Badziahin","cited_arxiv_id":null,"evidence_quote":"Supplies the dyadic-block setup and the two counting estimates (9)-(10) that bound polynomial heights and shared intervals; they are reused in every later case of this paper."},{"cited_title":"Badziahin, Y","cited_arxiv_id":null,"evidence_quote":"Provides the general rational-points-near-manifolds lower bound that fixes the target dimension formula for nondegenerate curves in the range considered."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the conjecture that predicts threshold $3/(2n-1)$; the endpoint $3/5$ for $n=3$ is the range verified here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the classical covering lemma used throughout to turn cardinality estimates into Hausdorff dimension upper bounds."},{"cited_title":"Beresnevich","cited_arxiv_id":null,"evidence_quote":"Gives the count of equivalence classes of binary cubic forms by discriminant, which is the input for the polynomial-counting theorem."}],"review_version":1}