{"id":"38a6ab0a-96cb-4e14-96c2-2b4d51050c47","arxiv_id":"2509.03650","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The parity-refined classes of strata of k-differentials are tautological and explicitly computable, via a spin double ramification cycle formula and Segre classes of cones of spin sections.","lead":"Mathematicians have long sought formulas for loci of curves carrying a differential with prescribed zeros and poles; this paper adds a spin-parity label and proves those refined loci have computable formulas. Generalists should read it because the same loci control volumes and dynamics of translation surfaces, and the new formulas make spin-refined versions of those numbers accessible.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1 rests on an unpublished multiplicity count (Remark 3.16) that fixes the coefficients in Theorem 1.8.","rationale":"The reader's weakest_assumption isolates the point where the chain from Theorem 1.8 to Theorem 1.1 is least secure. The proof of Theorem 4.14 computes the coefficients c_{Γ,I} in (26) as the sum of lengths of Artin local rings of DRL over ζ_Γ(M_{Γ,I}); the passage from the known lengths at DRL_f (from [HS21]) to the lengths at DRL requires the number of points of DRL over a generic point of DRL_f. That number is asserted in Remark 3.16 with a pointer to the unpublished thesis [Pol25]. Since the thesis is not available, the equality H^± = DR^± is not self-contained, and Lemma 5.5's induction inherits the gap. I checked the local structure in §3.3–3.4: the count is plausible and consistent with the explicit saturation relations (e.g., a single outlying vertex with two edges of twists a,b yields gcd(a,b)), so the concern is a missing proof rather than a suspected error. I also scanned for other load-bearing assumptions: the polynomiality inputs in Lemma 1.9 are cited to public preprints [Spe24, Pix23]; the apparent sign mismatch in the proof of Theorem 1.10 (Λ(-2t)Λ(t) vs Λ(2t)Λ(-t)) traces to a typo, since Proposition 6.5 gives ~s^±(t) = 2^g Λ(2t)Λ(-t)/(1-tψ_1) after substituting t↔2t; and the remaining computations appear standard. Thus the single most load-bearing concern is the unpublished count in Remark 3.16. A direct re-derivation from the monoid description would settle whether the gap is easily fillable.","tokens_in":53010,"tokens_out":11720,"duration_ms":87759,"concrete_test":"Independently derive the general count in Remark 3.16 using only the explicit monoid description of Section 3.3.3: for a k-simple star graph (Γ,I), the normalization of k[N] is k[N^sat] with N^sat = N^E⟨∑_{e∈γ} I^γ_e e / gcd(γ)⟩ (Lemma 3.13). Show that over a generic point of DRL_f, where O_{DRL_f,p} ≅ k(p)[e]/(e^{I'(e)}) (Theorem 3.18), the pullback Spec(k(p) ⊗_{k[N]} k[N^sat]) decomposes into exactly ∏_e I(e)/∏_v lcm_{e→v}(I(e)) factors, by computing the number of maximal ideals of k[N^sat]/(e^{I'(e)}, e∈E). If this derivation succeeds, Theorem 1.8 no longer depends on the unpublished thesis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.1 is proven via Lemma 5.5, which invokes Theorem 1.8 (spin DR star-graph formula) and Theorem 1.10. The proof of Theorem 1.8 (Theorem 4.14) reduces the coefficients c_{Γ,I} in equation (26) to the lengths of Artin local rings at generic points of DRL^{1/2} over DRL_f. These lengths are computed in Remark 3.22 as ∏_{v∈Vout} lcm_{e→v}(I(e))/k^{|Vout|}, but this uses Remark 3.16, which asserts that the number of points of DRL over a generic point of DRL_f is ∏_{e∈E} I(e)/∏_{v∈Vout} lcm_{e→v}(I(e)), citing the unpublished thesis [Pol25]. The thesis is not publicly available. Without this count, the equality H^± = DR^± is not established, and since Lemma 5.5 relies on Theorem 1.8 (via Lemma 1.9), the proof of Theorem 1.1 is conditional on an unavailable input. The count is plausible—for one outlying vertex with two edges of twists a,b it gives gcd(a,b), consistent with the saturation relations in §3.3.3—but it is neither proved nor independently verified in the manuscript.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Chow classes of spin-parity strata of k-differentials on M_g,n. For odd k and odd a it defines the signed class [M_g(a,k)]_± = [M_g(a,k)_+] - [M_g(a,k)_-] and proves (Theorem 1.1) that this class is tautological and explicitly computable. The proof has two main pillars: Theorem 1.8, which identifies the spin double ramification cycle DR^±_g(a,k) with the star-graph expression H^±_g(a,k), and Theorem 1.10, which expresses signed Segre classes of cones of spin sections in terms of quadratic Hodge symbols. The authors also prove Proposition 1.4 (DR^± = Pixton's spin class P^±) and develop an induction in Lemma 5.5 that reduces the computation of spin strata classes to the base cases supplied by Theorem 1.10.","tokens_in":53392,"tokens_out":22260,"duration_ms":179984,"significance":"If the main results are correct, the paper resolves a natural and long-standing problem: spin-parity refined strata classes are tautological and computable in all genera. It confirms Conjecture 2.5 of CSS21, provides the first general algorithm for these classes that does not rely on the full tautological ring, and yields a concrete route toward computing spin Hurwitz numbers and Witten-type classes in FJRW theory. The technical work is substantial: it combines Cornalba's spin curves, the logarithmic double ramification cycle, Pixton's formula, and incidence-variety compactifications. Many of the inputs are cited from published sources, and the paper contains extensive original arguments, including the parity-vanishing Proposition 2.10 and the µ2-action in Section 4. However, one load-bearing sign inconsistency in Theorem 1.10's proof (detailed below) needs to be resolved before the main theorem can be considered reliable.","major_comments":[{"comment":"The statement of Proposition 6.5 and its proof are inconsistent. The proof ends with the equality ~s^±_g(t/2) = 2^g Λ(t)Λ(-t/2)/(1 - tψ_1/2), which after substituting t ↦ 2t gives ~s^±_g(t) = 2^g Λ(2t)Λ(-t)/(1 - tψ_1). Eq. (40), however, states ~s^±_g(t) = 2^g Λ(-2t)Λ(t)/(1 - tψ_1). These two expressions are not equal in general: for g = 1, for instance, they differ in the coefficient of λ_1 t unless λ_1^2 = 0 in the Chow ring, which is not the case on M_1,n. The subsequent proof of Theorem 1.10 multiplies Eq. (40) by (1-tψ_1) and compares the result with L_g(t) = 2^{g-1}Λ(2t)Λ(-t) - 2^{2g-1}; this comparison is valid for exactly one of the two sign conventions. The authors must determine the correct formula, correct either Eq. (40) or the derivation, and then re-verify the first identity of Theorem 1.10 and its use in the induction of Lemma 5.5, which is the final step in the proof of Theorem 1.1.","section":"§6.5, Proposition 6.5, Eq. (40); proof of Theorem 1.10"}],"minor_comments":[{"comment":"The point count over DRL_f is stated as provable in the unpublished thesis [Pol25]. I checked the dependency chain: the proof of Theorem 4.14 reduces the coefficients c_{Γ,I} directly to lengths at generic points of DRL_f via [HS21, Lemma 2.12], so Remark 3.16 appears not to be used in the main theorems. If that is correct, the remark should be removed or explicitly marked non-essential; if it is used anywhere, a proof must be supplied in the manuscript.","section":"§3.3.4, Remark 3.16"},{"comment":"The sentence 'd(0, 3) = 1, d(1, 0) = -3, and d(1, 0) = 1' contains a typo: the second occurrence of d(1,0) should presumably be d(1,1).","section":"§5.4, proof of Proposition 5.6"},{"comment":"The name Teleman is spelled 'Telemann' in the passage discussing the reconstruction theorem; the reference is to Teleman's theorem.","section":"Introduction, Conjecture 1.2 discussion"},{"comment":"The notation P^{c,r,1/2}_g(a,k) is somewhat confusing because the superscript 1/2 appears together with the integer powers in the Pixton formula; a brief explanation of the notation would improve readability.","section":"§2.3, Definition 2.17 and Proposition 2.21"},{"comment":"The definition of bi-colored graph uses 'non-trivial partition of the set of vertices V = V_0 ⊔ V_-1'; it would be helpful to state explicitly whether both sets are required to be non-empty, as suggested by 'non-trivial'.","section":"§5.2, Definition 5.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a strong and important contribution, but the sign inconsistency in Proposition 6.5 is load-bearing because Theorem 1.10 is one of the two main ingredients in the proof of Theorem 1.1. The issue appears local and likely fixable, but it must be resolved and checked carefully. I also note that several key supporting results are cited from unpublished preprints (Spe24, Pix23, Sau24) and one from an unavailable thesis (Pol25); the authors should clarify the status of these dependencies."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline result is real. The paper proves the spin DR formula H± = DR± that was Conjecture 2.5 in CSS21, and from it gets the unconditional tautologicality and computability of spin-parity strata classes. Wong's earlier algorithm required those classes to be tautological as an input; here the induction goes through without that assumption. The Segre-class computation for cones of spin sections (Theorem 1.10) is also new and looks sound. The proof is mostly a careful chain through published results—BHP+23, HS21, Sau19—with new pieces around the parity action on DRL and the spin DR cycle. I did not find an internal contradiction.\n\nThe soft spot is the one the stress-test note identifies, and it is real. Remark 3.16 states the number of points of DRL over a generic point of DRL_f as ∏ I(e)/∏ lcm_{e→v}(I(e)), and says the proof is in the unpublished thesis [Pol25]. That count fixes the coefficients c_{Γ,I} in Theorem 4.14, which feeds Theorem 1.8 and then Lemma 5.5. So the main theorem is conditional on an input the reader cannot verify from the manuscript. The count is plausible—the gcd check in §3.3.3 fits—but the text itself says the proof is elsewhere. A secondary version of the same problem appears in Lemma 1.9: the a∈(kN)^n case leans on polynomiality results cited to [Spe24] and [Pix23], which are less public than one would like.\n\nThis is not a fatal flaw in the sense of an internal contradiction, but it is load-bearing. A referee should not have to take the author's thesis on faith. The path to a publishable version is clear: include the proof of Remark 3.16, or make [Pol25] available and have the referee verify it.\n\nWho benefits: anyone working on spin-refined strata, DR cycles, Masur-Veech volumes, or Witten-class computations in FJRW theory. The paper is important enough to deserve a serious referee rather than a desk reject. I would send it out, with the missing proof as a condition.","headline":"Strong paper that proves the spin DR conjecture and computes spin strata classes, but the main theorem is conditional on an unpublished thesis; send it to referees and make the missing proof a condition.","tokens_in":53808,"tokens_out":3456,"would_cite":false,"duration_ms":34444,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14H10","14C17"],"pacs":[],"model":"deepseek-v4-flash","headline":"Spin-parity strata classes in the Chow ring of moduli spaces of curves are tautological and explicitly computable for odd k and odd a.","keywords":["spin structures","strata of differentials","double ramification cycles","tautological ring","moduli of curves","spin parity","Chow ring","Segre classes"],"falsifier":"Directly enumerate the points of the double ramification locus over a generic point of its non-saturated model for a two-vertex $k$-simple star graph and compare with $\\prod_e I(e)/\\prod_v \\mathrm{lcm}_{e\\to v}(I(e))$. Alternatively, run the paper's algorithm for $(g,n,a,k)=(3,1,(5),1)$, push the resulting class of $[\\mathcal{M}_3((5))]_\\pm$ forward to $\\mathcal{M}_3$, and compare with the independently known class of the even spin component in that genus; any mismatch would trace through the induction.","tokens_in":52772,"feed_emoji":"🌀","tokens_out":12020,"duration_ms":103242,"temperature":0.7,"pith_summary":"The paper proves that, on moduli spaces of stable curves, the difference between the even and odd spin-parity components of a stratum of k-differentials is a tautological class and can be explicitly written in the standard tautological generators. This settles the computability question for the spin refinement of classes of strata of differentials when the order k and the vector a are odd. The route is a spin version of the double ramification cycle: one expression identifies it with an explicit graph-sum tautological class, and another expresses it as a star-graph sum whose summands are spin classes of strata of lower genus and fewer markings. These identities, together with a computation of signed Segre classes of cones of spin sections, feed an induction that gives the formula in every genus.","feed_headline":"Spin-parity strata of differentials are now computable","feed_subtitle":"The even-minus-odd spin class is shown tautological, via spin double ramification cycles and quadratic Hodge symbols.","key_machinery":"The load-bearing mechanism is the spin double ramification cycle $DR^\\pm_g(a,k)$, formed by capping the spin DR class with the parity cycle on the moduli space of spin curves and pushing forward. The argument uses two computable expressions for it: a graph-weighting tautological class $P^\\pm_g(a,k)$ obtained as the constant term of a polynomial in $r$, and a star-graph expansion $H^\\pm_g(a,k)$ indexed by $k$-simple star graphs with odd twists. The second engine is the cone of squares of spin sections; Theorem 1.10 packages its signed Segre classes in the series $L_g(t)=2^{g-1}\\Lambda(2t)\\Lambda(-t)-2^{2g-1}$, and a Möbius inversion converts that series into the cone Segre classes. Together these reduce parity-filtered stratum classes to finite sums of $\\kappa$-, $\\psi$- and $\\lambda$-classes, making induction on genus and $|a|$ possible.","core_discovery":"The central claim is Theorem 1.1: for odd $k$ and an odd vector $a$, the spin class $[\\mathcal{M}_g(a,k)]_\\pm := [\\mathcal{M}_g(a,k)_+] - [\\mathcal{M}_g(a,k)_-]$ lies in the tautological ring $A^*(\\mathcal{M}_{g,n},\\mathbb{Q})$ and is explicitly computable. The proof identifies the spin double ramification cycle $DR^\\pm_g(a,k)$ with the explicit graph-sum spin class $P^\\pm_g(a,k)$ (Proposition 1.4), and, for $a$ not in $(k\\mathbb{N})^n$, with the star-graph expansion $H^\\pm_g(a,k)$ whose coefficients are products of twist values and whose summands are spin classes of lower-complexity strata (Theorem 1.8). Theorem 1.10 then computes the signed Segre series of cones of spin sections as $L_g(t)=2^{g-1}\\Lambda(2t)\\Lambda(-t)-2^{2g-1}$, a quadratic Hodge symbol built from tautological $\\lambda$-classes. Lemma 5.5 assembles these inputs into an induction on genus, number of markings and $|a|$, producing an explicit formula in the standard generators.","pith_inferences":["If the paper's conjectural relation to spin-refined $r$-spin classes holds, the same formulas would yield the first explicit values of those classes; the paper does not prove that relation.","The one unavailable input, the point count over the non-saturated double ramification locus, could plausibly be derived from the $\\mu_2$-action constructed in Section 4, which would remove the dependence on the unpublished thesis.","The cone-of-sections mechanism is likely to extend to other locally constant refinements of strata, for example $r$-spin or weighted spin settings, whenever an analogous parity invariant exists.","Adding the even and odd classes should recover the ordinary stratum class $[\\mathcal{M}_g(a,k)]$, giving a low-degree consistency check of the new formula."],"forward_implications":["Every parity-refined stratum class $[\\mathcal{M}_g(a,k)]_\\pm$ is an explicit linear combination of $\\kappa$-, $\\psi$- and $\\lambda$-classes in the tautological ring.","The spin double ramification cycle equals the star-graph expansion $H^\\pm_g(a,k)$ for $a\\notin(k\\mathbb{N})^n$, and the polynomiality reduction handles the remaining cases.","All classes obtained by pushing forward powers of the tautological class on the projectivized cone of squares of spin sections, with spin signs and residue conditions, are tautological and computable.","A previously conditional spin-stratum algorithm for 1-differentials now runs in all genera, since Theorem 1.1 supplies the tautologicity it assumed.","The identity $L_g(t)=2^{g-1}\\Lambda(2t)\\Lambda(-t)-2^{2g-1}$ expresses the signed Segre generating series as a computable quadratic Hodge symbol."],"supporting_citations":[{"why":"Supplies Pixton's formula on the Picard stack, used to prove the identity DR±_g(a,k)=P±_g(a,k).","marker":"[BHP+23]"},{"why":"Provides the star-graph expression and the Artin ring length computations of DRL_f that determine the coefficients in Theorem 1.8.","marker":"[HS21]"},{"why":"Formulated the conjectural spin star-graph and spin Pixton identities that Theorem 1.8 and Proposition 1.4 prove.","marker":"[CSS21]"},{"why":"Developed the degeneracy-locus and cone formalism for strata of differentials adapted here to spin parity.","marker":"[Sau19]"},{"why":"Gave the previous conditional algorithm for spin stratum classes and the intersection formulas used in Lemma 5.5.","marker":"[Won24]"},{"why":"Established tautologicity and computability of λ-classes, the base of the induction and of the quadratic Hodge symbol.","marker":"[Mum83]"},{"why":"Chiodo's formula computes the Chern characters of the cone of spin sections used in Theorem 1.10.","marker":"[Chi08]"},{"why":"Supplies the parity push-forward identity that produces the quadratic Hodge term in Theorem 1.10.","marker":"[GKL21]"},{"why":"Constructs the extended Abel-Jacobi map and birational model on which the double ramification cycle is defined.","marker":"[Hol21]"}],"fun_headline_variants":["Spin parity strata computable via DR cycles","Even-minus-odd spin class is tautological","Refined DR formula for spin sections","Spin classes in Chow ring are computable","Quadratic Hodge symbol yields spin classes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The computation rests on an unpublished count: over a generic point of the non-saturated double ramification locus belonging to a $k$-simple star graph there are exactly $\\prod_e I(e)/\\prod_v \\mathrm{lcm}_{e\\to v}(I(e))$ points, a fact cited to a thesis that is not included; if that count is wrong, the star-graph coefficients fail.","fun_headline_variants_meta":{"raw":{"variants":["Spin parity strata computable via DR cycles","Even-minus-odd spin class is tautological","Refined DR formula for spin sections","Spin classes in Chow ring are computable","Quadratic Hodge symbol yields spin classes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000153,"raw_usage":{"total_tokens":1151,"prompt_tokens":830,"completion_tokens":321,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":446,"completion_tokens_details":{"reasoning_tokens":255}},"tokens_in":446,"tokens_out":321,"duration_ms":3171,"temperature":1.0,"reasoning_tokens":255,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:30:39.355509+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly enumerate the points of the double ramification locus over a generic point of its non-saturated model for a two-vertex $k$-simple star graph and compare with $\\prod_e I(e)/\\prod_v \\mathrm{lcm}_{e\\to v}(I(e))$. Alternatively, run the paper's algorithm for $(g,n,a,k)=(3,1,(5),1)$, push the resulting class of $[\\mathcal{M}_3((5))]_\\pm$ forward to $\\mathcal{M}_3$, and compare with the independently known class of the even spin component in that genus; any mismatch would trace through the induction.","supporting_citations":[],"review_version":2}