{"id":"10cfc10f-fd7b-49ba-a832-613f99c05ed9","arxiv_id":"2507.08652","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For 100% of hyperelliptic curves z^2 = f(x,y) ordered by height, every odd-degree algebraic point of degree at most 2g-1 has Weil height at least (1 + 1/(2g+2) - epsilon) log Ht(f).","lead":"For almost all hyperelliptic curves coming from integral binary forms of even degree, the authors prove that every algebraic point of odd degree at most 2g-1 has Weil height at least roughly (1 + 1/(2g+2)) times the log of the curve's height. The proof builds a new reduction theory for pairs of symmetric matrices and shows that the associated reduction covariant equidistributes, making small-height points rare.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the flagged Lemma 5.6 omission is filled by a direct matrix computation, and the density-1 height bound is not threatened by that step.","rationale":"The reader's own weakest-assumption pick is not actually weak. I verified the omitted m < n/2 case of Lemma 5.6 directly: the same complex basis diagonalizes every block of the form (psi(lambda), psi(mu)), and the normalization max(|lambda|,|mu|)=1 forces the reduction covariant to be the identity on L(m,tau). I then re-checked the downstream chain: Lemma 5.6 to Proposition 5.10 to Theorem 5.9 to Corollary 5.12 to Theorem 6.4 to Theorem 6.5. The logarithmic constants match: (n+1)log X is exactly what #F(X) ~ 2^{n+1} X^{n+1} predicts for the threshold 1 + 1/(2g+2), no constants are fitted, and there is no circularity. The remaining risk is the sheer depth of Sections 2 and 4 (Grothendieck duality, reflexive sheaves, ribbons) and the geometry-of-numbers adaptation, none of which is formally verified; that supports a cautious CONDITIONAL verdict but does not identify a concrete load-bearing flaw. I therefore disagree with the reader's identification of Lemma 5.6 as the main concern and leave the verdict unchanged.","tokens_in":30555,"tokens_out":44483,"duration_ms":559928,"concrete_test":"Run the omitted case of Lemma 5.6 explicitly: take representatives of L(m,tau) with complex blocks, e.g. (lambda,mu)=(1,i/2) and (1+i,1), form A=psi(lambda), B=psi(mu), and verify with F=[[1,1],[i,-i]] that F^T A F=diag(2lambda,2conj(lambda)), F^T B F=diag(2mu,2conj(mu)), and F F^H=2I, so H=I in standard coordinates. If this computation fails for some (lambda,mu) with max(|lambda|,|mu|)=1, Lemma 5.6 needs repair; otherwise the omitted case is settled.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's load-bearing concern is Lemma 5.6: the assertion that R is constant on each Bhargava fundamental set L(m,tau), with the m < n/2 case omitted. This concern does not land. For a complex 2x2 block, A = psi(lambda), B = psi(mu) with lambda = a+bi, mu = c+di, use the fixed complex basis F with columns (1,i), (1,-i). A direct calculation gives F^T A F = diag(2lambda, 2conj(lambda)) and F^T B F = diag(2mu, 2conj(mu)). Hence the same basis simultaneously diagonalizes every block, independent of lambda and mu, and the reduction covariant has diagonal entries max(2|lambda|,2|mu|) = 2max(|lambda|,|mu|) = 2 in that basis. Since F F^H = 2I, the Hermitian form in standard coordinates is the identity. Real root blocks contribute max(|lambda_i|,|mu_i|)=1 on the corresponding standard basis vector. Therefore R(L(m,tau)) = [H0] for every (m,tau), including all m < n/2; the omitted case is genuinely similar. I found no other specific flaw: the divisor-to-integral-orbit construction, the norm computation, the height inequality, and the density argument are internally consistent, and the exponent agrees with the count of forms. The remaining caveat is that the sheaf-theoretic and orbit-counting sections are intricate and not machine-checked, which is a verification burden rather than a detected error.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a reduction theory for the representation of SL_n on pairs of symmetric n×n matrices, constructs a reduction covariant R, proves its equidistribution over integral orbits, and uses this to show that in a density-1 family of integral binary forms of degree 2g+2, every Q-rational effective divisor of odd degree at most 2g−1 on the associated hyperelliptic curve has logarithmic Weil height at least (1+1/(2g+2)−ε) log Ht(f). For g=1 this gives a rational-point height lower bound of (5/4−ε) log Ht(f) for 100% of quartic forms. The proof combines a construction of integral orbits from divisors via coherent duality (Theorems 4.1 and 4.5), a norm computation (Proposition 4.7), an equidistribution theorem for the reduction covariant (Theorem 5.9), and a discriminant-based density-1 family (Proposition 6.1).","tokens_in":30772,"tokens_out":19691,"duration_ms":198406,"significance":"If correct, the main theorems give the first density-1 height lower bounds for odd-degree points on the full family of hyperelliptic curves, sharp up to ε, with the natural exponent 1+1/(2g+2). The method introduces a reduction covariant for a representation outside Vinberg theory and proves an equidistribution statement of independent interest, and the construction of integral orbits via Grothendieck duality is a technical innovation. The argument is internally consistent; the one flagged omission in Lemma 5.6 (the case m<n/2) is filled by a direct matrix calculation, so I do not regard it as a threat to the main claim. No free parameters are fitted to data, and the density-1 family is a standard sieve input.","major_comments":[],"minor_comments":[{"comment":"The proof of Lemma 5.6 handles only the case m = n/2 and states that m < n/2 is similar without details. Because this constancy is used to normalize R(L(m,τ)) = H0 and consequently in Proposition 5.10 and Theorem 5.9, the omitted case should be written out; it follows by simultaneously diagonalizing each complex 2×2 block ψ(λ), ψ(μ) with the fixed basis (1,i)^t and (1,-i)^t, which yields the standard Hermitian form after the normalization max(|λ|,|μ|)=1.","section":"§5.1, Lemma 5.6"},{"comment":"The statement 'D does not intersect the irreducible Weierstrass locus S_f' is used to apply the orbit construction of §4, but the reason is not given: because f is irreducible over Q, any Q-rational effective divisor supported on S_f has degree at least 2g+2, so a degree 2g−1 divisor cannot meet S_f. Please state this explicitly.","section":"§6.1, proof of Theorem 6.4"},{"comment":"In the final sentence of the proof, 'n log(w,w)_{HA,B} − det H_{A,B}' should read 'n log(w,w)_{HA,B} − log det H_{A,B}'.","section":"§6.2, proof of Theorem 6.4"},{"comment":"The statement of Theorem 6.5 (and similarly the abstract and Theorem 6.4) should specify that the divisors are effective Q-rational divisors, since the height h(D) is defined only for such divisors.","section":"§6.3, Theorem 6.5"},{"comment":"In the definition of V(0,1+) and V(0,1−), the text writes 'Let V(0,τ) denote the set...' but τ is not a free parameter there; the two cases should be named explicitly.","section":"§5.1, V(0,τ)"},{"comment":"The formula '#F(X) = 2n+1X^{n+1} + O(X^n)' appears with a formatting error; it should be 2^{n+1}X^{n+1}.","section":"§5.2, Lemma 5.8"}],"recommendation":"minor_revision","confidential_remarks":"This is a strong paper suitable for publication. I recommend minor revision mainly to add the omitted proof in Lemma 5.6 and clarify the Q-rational divisor hypotheses. I do not see any circularity or data-fitting concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a real result. Theorem 6.5 gives a density-1 lower bound of (1+1/(2g+2)-epsilon) log Ht(f) for odd-degree divisors of degree at most 2g-1 on the even-degree hyperelliptic family, and the g=1 case gives the corresponding statement for rational points. The bound is new for this family: [19] treats only the odd (monic) case, and the Bhargava–Gross–Wang and Bhargava papers cited here don't contain a height lower bound of this shape.\n\nWhat's good. The reduction covariant for SL_n on pairs of symmetric matrices is a new construction, and the particular choice of H is not arbitrary: it makes the height computation clean (Lemma 3.11 and Proposition 4.7). The integral orbit construction in Section 4, using Wood's ideal class viewpoint and Grothendieck duality, is a substantial upgrade over the earlier constructions in [8] and [19]. I checked the elementary parts: Proposition 6.2 uses the bound log|c| <= 0.5 log(n(n+1)) + log X from [25] and reproduces exactly the (n+1) log X term in Theorem 6.4; the exponent then follows from #F(X) = 2^{n+1} X^{n+1} + O(X^n). No constants are fitted; the conclusion is not circular.\n\nSoft spots. The one I'd flag is Lemma 5.6. It asserts that the reduction covariant is constant on each Bhargava fundamental set L(m,tau), and the proof only gives the m=n/2 case, saying the m<n/2 case is similar. Since this constancy normalizes R(L(m,tau)) = H0 and is used for the equidistribution Proposition 5.10, the omitted case matters. From the explicit forms (5.1)-(5.2), I agree with the stress-test note that the missing computation goes through: for the complex 2x2 blocks, a fixed basis diagonalizes both A and B independent of the entries, and the max condition gives the identity Hermitian form. So it's an omission of detail, not a flaw in the argument. The paper should expand this before publication. The other caveat is generic: Sections 2 and 4 are heavy sheaf theory and Section 5 is a long averaging argument; I found no specific error, just a high verification burden.\n\nWho it's for and what to do. Arithmetic geometers and number theorists working on heights, hyperelliptic curves, and arithmetic invariant theory will get real value from this. It deserves a serious referee. I'd send it out, with a request that the referee check Lemma 5.6 and the sheaf-theoretic orbit construction in detail.","headline":"A density-1 height lower bound for odd-degree points on even-degree hyperelliptic curves, with a long but coherent proof; one omitted case in Lemma 5.6 should be expanded before publication.","tokens_in":31467,"tokens_out":3130,"would_cite":true,"duration_ms":34061,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G30","11G50","14H10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For 100% of hyperelliptic curves, every odd-degree point of degree at most 2g−1 has height at least (1+1/(2g+2)−ε) log Ht(f), and the same bound holds for degree-(2g−1) divisors.","keywords":["hyperelliptic curves","Weil height","reduction theory","binary forms","arithmetic invariant theory","equidistribution","divisors","odd-degree points"],"falsifier":"Take n=4 and m=1, pick two elements of the same fundamental set L(1,τ) using the block forms (5.1)–(5.2), and compute their reduction covariants; if the classes in SL_4(Z)\\X differ, Lemma 5.6 is false and the equidistribution theorem fails.","tokens_in":30190,"feed_emoji":"🧮","tokens_out":6691,"duration_ms":66132,"temperature":0.7,"pith_summary":"The paper proves that, in a density-1 family of hyperelliptic curves, odd-degree points cannot be too small: their heights must be at least a positive fraction of the logarithm of the curve's defining binary form. Concretely, for 100% of integral binary forms of degree 2g+2 with nonzero discriminant, ordered by height, every effective divisor of degree 2g−1 has logarithmic height at least (1+1/(2g+2)−ε) log Ht(f). For genus 1, this says that every rational point on 100% of curves z²=f(x,y) with f a quartic form has height at least (5/4−ε) log Ht(f). The proof develops a new reduction theory for the action of SL_n on pairs of symmetric n×n matrices and shows that the associated reduction covariant equidistributes over integral orbits.","feed_headline":"Height bound holds for 100% of hyperelliptic curves","feed_subtitle":"On almost every z²=f(x,y), divisors of degree 2g−1 have height at least (1+1/(2g+2)) log Ht(f).","key_machinery":"The load-bearing object is the reduction covariant R: SL_n(Z)\\V(Z)_{Δ≠0} → SL_n(Z)\\X, where X=SL_n(R)/SO_n(R) is the symmetric space of inner products on R^n up to scaling. For a pair (A,B) of symmetric matrices one forms the pencil Ax−By, simultaneously diagonalizes it over C, and builds a positive definite inner product H_{A,B} by taking the maximum of the absolute diagonal values; this construction is equivariant and passes to a map on integral orbits. Odd-degree divisors on X_f produce integral orbits for this representation through the correspondence between orbits and ideal classes in the ring attached to f, using coherent Grothendieck duality; the divisor height controls the norm of a primitive vector in Z^n with respect to H_{A,B}. The proof then shows that these reduction covariants equidistribute over all integral orbits of bounded height, while lattices admitting a very short vector form a set of small measure; combining these facts gives the density-1 theorem.","core_discovery":"The central claim is Theorem 6.5: for every ε>0, the set S_ε of binary forms f∈Z[x,y] of degree 2g+2 and nonzero discriminant such that every effective divisor D on X_f: z²=f(x,y) of odd degree ≤2g−1 satisfies h(D) ≥ (1+1/(2g+2)−ε) log Ht(f) has density 1 when forms are ordered by height. Equivalently, asymptotically all forms belong to S_ε, so small-height divisors are rare: the count of forms admitting a degree-(2g−1) divisor whose height violates the bound is o($X^{{2g+3}}$). The same statement transfers to algebraic points: for a Galois orbit P of odd degree m≤2g−1, the quantity m·h(π(P)) obeys the same lower bound.","pith_inferences":["The proof of Lemma 5.6 explicitly leaves the case m<n/2 unverified: it checks constancy of the reduction covariant on the fundamental sets only when all blocks are real and says the mixed complex case is similar. The equidistribution theorem and therefore the density-1 height bound depend on this omitted case, so a reader checking the proof should verify it before treating the theorem as complete.","The construction of integral orbits from divisors via coherent Grothendieck duality is likely reusable in other arithmetic invariant theory settings, including the odd hyperelliptic family treated in the authors' earlier work.","A direct numerical check of the reduction covariant on the fundamental sets L(m,τ) for small n and m<n/2 would provide a concrete test of the omitted case.","The height bound concerns the naive divisor height h(D); it does not by itself yield statements about canonical heights or about proportions of curves with points of larger odd degree."],"forward_implications":["For genus 1, the result gives that on 100% of integral binary quartics, every rational point P satisfies h((x:y)) ≥ (5/4−ε) log Ht(f).","For every genus g≥1, every effective divisor of degree 2g−1 on 100% of hyperelliptic curves satisfies h(D) ≥ (1+1/(2g+2)−ε) log Ht(f).","For every algebraic point of odd degree m≤2g−1 on 100% of such curves, m·h(π(P)) obeys the same lower bound.","The equidistribution theorem covers all integral orbits of nonzero discriminant, not only irreducible orbits, so the density-1 conclusion follows without a separate reducible-orbit argument.","The reduction covariant provides a uniform method for reducing pairs of symmetric matrices, which the paper expects to be useful for computational arithmetic invariant theory."],"supporting_citations":[{"why":"Supplies the geometry-of-numbers counting method and the fundamental sets for orbits on pairs of symmetric matrices that the equidistribution proof adapts.","marker":"[5]"},{"why":"Establishes the invariant-form–orbit correspondence for this representation and the hyperelliptic curve context, motivating the question of odd-degree points.","marker":"[8]"},{"why":"Gives the parametrization of ideal classes in rings associated to binary forms by SL_n-orbits on pairs of symmetric matrices, used to construct integral orbits from divisors.","marker":"[32]"},{"why":"Provides the description of the ring R_f attached to a binary form, which underlies the sheaf-theoretic construction of orbits.","marker":"[31]"},{"why":"The authors' earlier work introducing the equidistributed-reduction-covariant strategy for odd hyperelliptic curves; the present paper adapts and extends that strategy.","marker":"[19]"},{"why":"Supplies the discriminant-lower-bound density-one family and auxiliary estimates used in Section 6.","marker":"[10]"},{"why":"Provides the root bound used in the height computation of Proposition 6.2.","marker":"[25]"}],"fun_headline_variants":["Odd degree points on almost all hyperelliptic curves have large height","Density-1 family: no small heights for odd degree points","Almost all hyperelliptic curves rule out small odd-degree point heights","For 100% of forms, odd degree divisors exceed height bound"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on the claim that the normalization procedure gives the same output for every representative in certain standard slices of the matrix space, but the paper checks this only for the slices where all blocks are real and says the mixed complex case is similar without giving details.","fun_headline_variants_meta":{"raw":{"variants":["Odd degree points on almost all hyperelliptic curves have large height","Density-1 family: no small heights for odd degree points","Almost all hyperelliptic curves rule out small odd-degree point heights","For 100% of forms, odd degree divisors exceed height bound"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000852,"raw_usage":{"total_tokens":3653,"prompt_tokens":846,"completion_tokens":2807,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":462,"completion_tokens_details":{"reasoning_tokens":2732}},"tokens_in":462,"tokens_out":2807,"duration_ms":22263,"temperature":1.0,"reasoning_tokens":2732,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:15:34.288028+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take n=4 and m=1, pick two elements of the same fundamental set L(1,τ) using the block forms (5.1)–(5.2), and compute their reduction covariants; if the classes in SL_4(Z)\\X differ, Lemma 5.6 is false and the equidistribution theorem fails.","supporting_citations":[{"cited_title":"Bhargava, B","cited_arxiv_id":null,"evidence_quote":"Establishes the invariant-form–orbit correspondence for this representation and the hyperelliptic curve context, motivating the question of odd-degree points."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the parametrization of ideal classes in rings associated to binary forms by SL_n-orbits on pairs of symmetric matrices, used to construct integral orbits from divisors."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the description of the ring R_f attached to a binary form, which underlies the sheaf-theoretic construction of orbits."},{"cited_title":"Soundararajan","cited_arxiv_id":null,"evidence_quote":"Provides the root bound used in the height computation of Proposition 6.2."}],"review_version":1}