{"id":"6f20a2c7-1072-4c2f-b5a3-f76c82de459e","arxiv_id":"1908.02355","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a generic genus 5 curve, the theta hyperplanes determine the set of enveloping quadrics, hence determine the curve; this yields a theta-hyperplane description of the Schottky locus in genus 5.","lead":"This paper proves that for a generic genus 5 curve, the theta hyperplanes determine the quadrics enveloping the curve, and those quadrics determine the curve itself. The proof mixes classical algebraic geometry with certified computer calculations, and gives a new description of the Schottky locus in genus 5.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1 hinges on the unverified SVD accuracy assumptions in §7.2; a misclassification of a zero versus nonzero singular value in the finite checks of Theorem 7.6 would break Proposition 4.2 and the numerical certification.","rationale":"The reader's weakest assumption matches my analysis: the certified numerical protocol is the single load-bearing concern. I found no independent mathematical flaw in the degeneration argument, the Steiner-system combinatorics of Proposition 4.2, or the representation decomposition in Corollary 7.8; those parts are coherent, and the finite computations are small enough that exact verification is feasible. The theorem therefore stands or falls on whether the floating-point SVD classification in Theorem 7.6 is correct, and the paper currently supplies only an empirical accuracy assertion in §7.2 rather than a formal certificate. The proposed exact rerun would settle the question because all matrices are tiny and the coordinate field is a biquadratic extension of Q. I also note that Remark 7.10's claim that a single printed witness 'proves the theorem' is an overstatement unless that witness is propagated by an explicit group action to all 510 classes; however, that remark is supplementary and the code is the actual proof vehicle. In agreement with the reader, the verdict remains CONDITIONAL: the mathematical architecture is likely sound, but the numerical certification needs a rigorous, machine-checkable validation before the theorem can be accepted unconditionally.","tokens_in":22986,"tokens_out":24731,"duration_ms":287242,"concrete_test":"Re-run the entire finite computation behind Theorems 7.6 and 7.7 with rigorous interval arithmetic or exact linear algebra over Q(i,√φ), using the same 18 orbit representatives from the supplied program. Concretely, compute the SVD/rank of every matrix appearing in the classifications with MPFR at 256-bit precision and directed rounding, or with exact rational arithmetic over the biquadratic extension, and verify that (1) every singular value classified as zero is truly below 10^-14 and every classified nonzero value is truly above 10^-2 with margin, and (2) the resulting set A and relation R are transitive with exactly 510 equivalence classes. If both conditions hold, the numerical certification in §7.2 is validated; if either fails, Theorem 7.6 and hence Theorem 1.1 are not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 2.4 legitimately reduces Theorem 1.1 to finitely many numerical checks on the Wiman curve, and the degeneration and representation-theoretic framework of Sections 2–5 is coherent. The load-bearing step is Theorem 7.6, which certifies that the relation R from Proposition 4.2 is transitive with exactly 510 equivalence classes and that these classes are exactly the intersections of O160 with Steiner systems. That certification is a classification of computed singular values of small matrices (at most 12×16 in Corollary 6.7, 8×16 in Corollary 6.9, 15×48 for Corollary 5.5) into the intervals [0,10^-14] and [10^-2,10^3], plus analogous tests for double and triple points. The only bridge from these floating-point numbers to true mathematical statements is the assertion in §7.2 that the SVD outputs satisfy entrywise residuals near 10^-14 and 3×10^-14, based on running the accompanying program. This is not a machine-checked certificate or a proven error bound; for matrices whose true singular values lie near the threshold, the computed classification could in principle flip. A single flipped classification changes the set A in Proposition 4.2, can destroy transitivity or the count of 510 classes, and removes the only rigorous connection between the numerical tests and the Steiner systems. Remark 6.10 explicitly concedes that the multiplicity tests are insufficient when four theta characteristics share a common point, so Proposition 4.2 is not a mere formality; it carries the full weight of the numerical certification. Therefore the central claim is conditional on the unverified accuracy assumption in §7.2.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper gives an effective reconstruction procedure for a generic genus 5 curve from its theta hyperplanes. The main theorem (Theorem 1.1) identifies the ideal of quadrics I2(C) of a generic genus 5 curve C with the intersection, over all nonzero 2-torsion points α of the Jacobian, of the span of the quadrics q{θ,θ+α} associated to pairs of theta hyperplanes, as θ ranges over the Steiner system Σ_{C,α}. The proof specializes to the Wiman curve W, computes the 160 relevant odd theta characteristics explicitly, proves combinatorial facts about their intersections with Steiner systems (Proposition 4.2), and then uses a certified numerical argument to verify the remaining finite checks. Corollary 1.2 gives the actual curve by Enriques–Babbage, and Corollary 1.3 gives a description of the Schottky locus in genus 5.","tokens_in":23226,"tokens_out":18214,"duration_ms":181190,"significance":"If fully established, the theorem is a natural and nontrivial extension of the author's earlier genus 4 reconstruction result, and it would give an effective, theta-hyperplane-based characterization of the genus 5 Schottky locus. The algebro-geometric reductions in Sections 2–5 are elegant: the specialization to the Wiman curve, the explicit coordinates for the 160 theta characteristics, and the representation-theoretic decomposition of Sym^2 H^0(K_W) are convincing and well motivated. Proposition 6.4 also provides explicit and usable error bounds. The paper is honest about its limitations, including Remark 6.10 on the insufficiency of the multiplicity tests. However, the claimed ``certified numerical proof'' is not yet a fully rigorous mathematical proof, because the key numerical certification in Section 7 rests on empirical assertions about floating-point SVD accuracy rather than on machine-checked certificates or interval/ball arithmetic. This gap is load-bearing for Theorem 1.1.","major_comments":[{"comment":"The numerical certification is not a proven bound. The proof of Theorem 7.6 depends on the assertion that the computed singular values of the matrices from Corollaries 5.5, 6.7, and 6.9 lie in the disjoint intervals [0, 10^{-14}] and [10^{-2}, 10^3] (and analogous assertions in later steps), and on the claim in §7.2 that all SVD outputs satisfy entrywise residuals near 10^{-14} and 3·10^{-14}. These claims are based on running the supplied program, not on a rigorous error analysis of the SVD, no condition-number estimates are provided, and no machine-checkable certificate of the singular value classification is supplied. A single misclassification of a true zero singular value as nonzero, or of a small nonzero value as zero, would change the set A in Proposition 4.2 and could destroy the transitivity or the count of 510 equivalence classes, thereby invalidating Theorem 7.6 and with it Theorem 1.1.","section":"§7.2, Theorem 7.6"},{"comment":"The paper concedes in Remark 6.10 that the multiplicity tests of Section 6 can pass incorrectly when the four theta characteristics share a common point, and that the only protection is Proposition 4.2. But the verification of the hypotheses of Proposition 4.2 in Theorem 7.6 uses the same numerically classified singular values that come from the tests. Thus the exclusion of the problematic quadruples is not an independent check; it is the same numerical output that is being certified. This circularity needs to be broken, for example by verifying the bad cases with exact arithmetic over finite fields (as sketched in Remark 7.10) or with rigorous interval enclosures that are checked independently of the floating-point SVD output.","section":"Remark 6.10, proof of Theorem 7.6"},{"comment":"The proof of Corollary 7.8 is compressed on a load-bearing point. Proposition 7.7 verifies the dimension and the representation-theoretic projections of the orthogonal complement for 6 of the 12 orbits of Steiner systems, and Corollary 7.8 then obtains the intersection statement for all Steiner systems by ``applying the symmetry group.'' Please spell out the action of the full automorphism group on the 12 orbits and on the irreducible decomposition, and state explicitly how the remaining 6 orbits (and hence all 510 Steiner systems) are covered. As written, the reader cannot check that the dimension-3 conclusion for the full intersection follows from the data in Proposition 7.7 alone.","section":"Corollary 7.8, Proposition 7.7"}],"minor_comments":[{"comment":"The definition of the relation R is ill-typed: A is a subset of 4-element subsets, but the condition ``θ1∪θ2∈A'' treats θ1 and θ2 as individual theta characteristics even though R is said to be a relation on (O160 choose 2). Please rewrite the definition, presumably to say that two pairs P1,P2 ∈ (O160 choose 2) are related when P1∪P2 lies in A.","section":"Proposition 4.2"},{"comment":"The proof reuses the letter A for two different sets: first ``Let A be the set of a's...'' and later ``A := (O160 choose 4) \\ A.'' Use distinct names for the certified set and for the set entering Proposition 4.2 to avoid confusion.","section":"Theorem 7.6"},{"comment":"The sentence ``By Corollary 7.8 this witness proves the theorem'' is an overclaim: the supplied spanning tree is a witness for one partial Steiner system only. To serve as a traditional proof certificate, the paper needs witnesses for representatives of all 18 G0-orbits, or an explicit explanation of how a single tree plus the group action covers all Steiner systems.","section":"Remark 7.10"},{"comment":"The statement that IEEE-754 ``guarantees at most half a bit error for each multiplication, addition, or subtraction'' is imprecise; the standard formulation is that each of these operations is correctly rounded to within 1/2 ulp. Please rephrase, and specify which operations and rounding modes are assumed.","section":"§7.1"}],"recommendation":"major_revision","confidential_remarks":"The algebro-geometric part of the paper is coherent and the claimed result is substantial. My main concern is the level of rigor of the numerical certification: the proof as written relies on empirical SVD accuracy assertions and on a circular exclusion of bad cases. This is fixable in principle by adding rigorous certificates, interval/ball arithmetic bounds, or exact finite-field verification for the finite checks, so I recommend major revision rather than rejection. I do not see concerns about novelty or about the citation to the author's earlier work [L2]; the specialization lemma is standard and is used appropriately."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nRead Lehavi's genus 5 reconstruction paper. Short version: the main theorem is new and likely correct, the algebro-geometric part is careful, and the only serious weakness is the numerical certification in §7, which is real but fixable.\n\nWhat's actually new: Theorem 1.1 gives a formula for I2(C) as the intersection of spans of quadrics associated to theta hyperplanes over Steiner systems. For a generic genus 5 curve that yields effective reconstruction (Cor 1.2) and a Schottky-locus description (Cor 1.3). The strategy follows the author's genus 4 paper [L2], but that's not a demerit; the specialization to the Wiman curve and the representation-theoretic decomposition in §5 are clean. Section 6's error bounds are explicit and genuinely useful, and the code is provided.\n\nWhere it's soft: the bridge from floating-point numbers to mathematical truth is asserted, not proven. Section 7.2 says the SVD outputs satisfy residuals near 10^-14 and 3*10^-14 'up to the accuracy' of IEEE-754 operations, based on running VerifiedZSVD. That's an empirical claim, not a machine-checked interval certificate. Since the classification of singular values into [0,10^-14] and [10^-2,10^3] is load-bearing for Theorem 7.6, a skeptical referee can fairly ask for interval arithmetic or an exact finite-field verification. Remark 6.10 makes this concrete: the triple-point test can pass incorrectly when four theta characteristics share a common point, so the burden falls on the numerical definition of A and Proposition 4.2. And Remark 7.10 overreaches a bit: a single printed witness for one partial Steiner system doesn't, by itself, prove the theorem unless the group action is shown to propagate it; that part needs a clearer argument.\n\nNone of this makes me think the theorem is false. The singular value gaps are enormous (10^-14 vs 10^-2), the algebro-geometric reductions are coherent, and the author flags the limitation in 6.10. I'd bet the result is right. But as written, the proof is conditional on an unverified numerical assumption.\n\nWho this is for: anyone working on reconstruction from theta characteristics, Coble/Steiner systems, or effective Schottky problems. It deserves a serious referee. I'd send it out, and ask the author to tighten §7.2 and Remark 7.10—ideally with interval arithmetic or a full witness/group-action proof. If those are fixed, this is an accept.\n\nBest","headline":"Solid new effective reconstruction theorem for generic genus 5 curves; the degeneration and Steiner-system analysis are coherent, and the numerical certification is the one genuinely soft spot.","tokens_in":23812,"tokens_out":3412,"would_cite":true,"duration_ms":36969,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14H40","14Q20"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a generic genus 5 curve, the theta hyperplanes determine the enveloping quadrics, and those quadrics cut out the curve.","keywords":["theta hyperplanes","genus 5 curves","Steiner systems","Schottky locus","effective reconstruction","certified numerical proof","enveloping quadrics","Wiman curve"],"falsifier":"Run the supplied program's verification steps with exact rational arithmetic (for example using the Chinese-remainder approach mentioned in Remark 7.10) on the Wiman curve, and check whether the 510 equivalence classes of pairs of the 160 theta characteristics match the claimed Steiner-system intersections; any mismatch would invalidate Theorem 7.6 and hence Theorem 1.1.","tokens_in":22699,"feed_emoji":"📐","tokens_out":9970,"duration_ms":96739,"temperature":0.7,"pith_summary":"The paper seeks to prove that a generic curve of genus 5 is determined, in an effective way, by its theta hyperplanes: the 496 hyperplanes attached to its odd theta characteristics. The route is to recover the ideal $I_2(C)$ of quadrics containing the canonical model of $C$, because the Enriques–Babbage theorem says those quadrics cut out the curve. The proof specializes the statement to a single highly symmetric curve, the Wiman curve, and then certifies a large numerical computation with explicit error bounds, making the reconstruction algorithmic rather than merely existential. A corollary gives a description of the Schottky locus in genus 5 in terms of intersections of spans of theta-hyperplane quadrics.","feed_headline":"Theta hyperplanes rebuild generic genus 5 curves","feed_subtitle":"A certified computation shows the enveloping quadrics follow from Steiner systems, and the quadrics cut out the curve.","key_machinery":"The load-bearing object is the Steiner system $\\Sigma_{C,\\alpha}$: for a nonzero 2-torsion point $\\alpha$ on the Jacobian, the set of 240 odd $\\theta$ characteristics $\\theta$ for which both $\\theta$ and $\\theta+\\alpha$ have odd $h^0$. The argument also rests on the quadrics $q_{\\theta,\\theta+\\alpha}$, the images of pairs of $\\theta$ hyperplanes in the symmetric square of the canonical system; Theorem 1.1 says the intersection of their spans over all $\\alpha$ is exactly $I_2(C)$. The proof machinery then specializes to the Wiman curve $W_{160}$, uses its automorphism group to decompose $\\operatorname{Sym}^2 H^0(K_W)$ into irreducible representations, and certifies numerically—with singular values separated into $[0,10^{-14}]$ and $[10^{-2},10]$ ranges and with propagation bounds from Section 7.2—that the relevant spans have dimension 13 and that their intersection is $I_2(W)$.","core_discovery":"The central claim, Theorem 1.1, is that for a generic genus 5 curve $C$, with $q_{\\theta,\\theta+\\alpha}$ denoting the quadric obtained from the pair of $\\theta$ hyperplanes $l_\\theta, l_{\\theta+\\alpha}$ (for an odd $\\theta$ characteristic $\\theta$, the unique hyperplane whose intersection with the canonical curve is everywhere double), the ideal of quadrics containing $C$ satisfies $$I_2(C)=\\bigcap_{\\$\\alpha$\\in J_C[2]\\setminus\\{0\\}} \\operatorname{span}(\\{q_{\\$\\theta$,\\$\\theta$+\\$\\alpha$}\\}_{\\$\\theta$\\in\\Sigma_{C,\\$\\alpha$}}),$$ where $\\Sigma_{C,\\alpha}$ is the Steiner system of the pair $(C,\\alpha)$, i.e. the 240 odd $\\theta$ characteristics $\\theta$ for which both $\\theta$ and $\\theta+\\alpha$ are odd. The author establishes this by proving the analogous statement for the Wiman curve and then extending to the generic curve by a semi-continuity specialization argument. Once $I_2(C)$ is known, Enriques–Babbage gives that the canonical curve is the intersection of these quadrics, so $C$ is effectively reconstructed; in addition, interpreting the right-hand side for any principally polarized abelian variety defines a $\\theta$-hyperplane condition for the Schottky locus in genus 5.","pith_inferences":["Beyond the paper, the certified-numerical template—separated singular values plus an exact witness—should transfer to other effective-reconstruction problems; genus 6 is a natural test case because the general curve is not a complete intersection there.","One could test the author's conjecture that the statement generalizes to higher genera numerically, by running the same protocol on the natural symmetric candidate curves for genus 6 and 7 and checking whether the corresponding intersection of Steiner-system quadrics has the predicted dimension.","Because the right-hand side of Theorem 1.1 is defined for any principally polarized abelian variety via the Gauss map, the result suggests a numerical Jacobian-recognition test: compute the intersection of theta-hyperplane quadrics and check whether it is a curve of arithmetic genus 5.","The exception noted in Remark 6.10—four theta characteristics sharing a common point can fool the triple-point test—indicates that a purely algebraic proof would need a sharper multiplicity analysis; searching for such exceptional configurations on other curves could reveal where the numerical certificate is genuinely essential."],"forward_implications":["For a generic genus 5 curve, the curve can be algorithmically reconstructed from its theta hyperplanes alone; no ordering or additional level structure on the characteristics is required.","The Schottky locus in genus 5 is characterized as the locus of principally polarized abelian varieties for which the intersection of the theta-hyperplane quadrics is a curve of arithmetic genus 5 with the same moduli point.","The proof supplies a finite, exact-checkable witness for the numerical part—a spanning tree of pairs of theta characteristics—so the computational claim can be verified without rerunning floating-point code.","Because the statement is verified on one curve and extended by semicontinuity, any future check on a different genus 5 curve would only need to reproduce the dimension counts on that single curve."],"supporting_citations":[{"why":"Supplies the specialization lemma (its Corollary 7) and the genus 4 reconstruction result that the present genus 5 argument directly extends.","marker":"[L2]"},{"why":"Established the current framing by showing curves can be characterized by their odd theta-characteristics without an ordering of the characteristics.","marker":"[CS2]"},{"why":"Provides the Enriques–Babbage theorem, used to pass from the recovered $I_2(C)$ back to the curve itself, and the Humbert-curve facts used in Section 3.","marker":"[ACGH]"},{"why":"Supplies the definitions and structure of theta characteristics and Steiner systems that the main formula is stated in.","marker":"[D]"},{"why":"Introduced the Wiman curve, the specific genus 5 curve to which the proof specializes.","marker":"[W]"},{"why":"Gives the $W_{160}$ notation and the model of the curve used for explicit coordinates.","marker":"[E]"},{"why":"Originally introduced Humbert curves, whose five elliptic double covers underlie the explicit set $O_{160}$ of theta characteristics.","marker":"[H]"}],"fun_headline_variants":["Theta hyperplanes fully determine generic genus 5 curves","Certified computation rebuilds genus 5 curves from theta hyperplanes","Effective genus 5 curve reconstruction from theta hyperplanes","Generic genus 5 curves rebuilt from theta hyperplanes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The theorem stands on the premise that the supplied program's singular value decompositions really have the claimed accuracy, at error levels near $10^{-14}$ and $3\\cdot 10^{-14}$, since that numerical output, rather than a machine-checked certificate, is what verifies the dimension computations.","fun_headline_variants_meta":{"raw":{"variants":["Theta hyperplanes fully determine generic genus 5 curves","Certified computation rebuilds genus 5 curves from theta hyperplanes","Effective genus 5 curve reconstruction from theta hyperplanes","Generic genus 5 curves rebuilt from theta hyperplanes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001886,"raw_usage":{"total_tokens":7359,"prompt_tokens":869,"completion_tokens":6490,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":485,"completion_tokens_details":{"reasoning_tokens":6424}},"tokens_in":485,"tokens_out":6490,"duration_ms":47384,"temperature":1.0,"reasoning_tokens":6424,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:47:29.751235+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the supplied program's verification steps with exact rational arithmetic (for example using the Chinese-remainder approach mentioned in Remark 7.10) on the Wiman curve, and check whether the 510 equivalence classes of pairs of the 160 theta characteristics match the claimed Steiner-system intersections; any mismatch would invalidate Theorem 7.6 and hence Theorem 1.1.","supporting_citations":[],"review_version":1}