{"id":"ab2a20ec-a02e-4f01-b3c1-ba54e2ed3d98","arxiv_id":"2411.17551","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Intersection numbers on moduli spaces of stable curves are shown to equal, up to sign, the chromatic polynomial of a graph evaluated at a negative integer.","lead":"The authors define a new family of intersection numbers on moduli spaces of stable curves, indexed by finite graphs, and prove they equal evaluations of the graph's chromatic polynomial. The result connects enumerative geometry with algebraic statistics, hyperplane arrangements, and scattering amplitudes, giving a clean bridge between graph theory and moduli spaces.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the main theorem is proved by two independent routes, and the intricate boundary analysis in Theorem 4.4(3) is auxiliary; I found no concrete flaw.","rationale":"The reader identified Theorem 4.4(3) and its case analysis as the weakest assumption, and I agree that this is the most intricate passage. However, I do not regard it as a load-bearing risk for the central claim: the first proof of Theorem 1.4 is independent of the hyperplane-arrangement machinery, and its deletion-contraction recursion is a direct computation with psi-classes and boundary strata. I re-checked the structure of that recursion: the base case is standard, the expansion into (C1)-(C6) follows from Facts 2.3 and 2.4, and the final addition uses Fact 2.3 in the form (C1)+(C6)=(C2)+(C5)=-(C3), which is consistent. For the second proof, the induction in Theorem 4.4(3) is intricate but its two critical ingredients, Lemma 4.8 and the graph-of-a-rational-function step, check out. No concrete error, missing case, or unproved assertion that affects the theorem was found. The paper's self-identification of 'some involved computations' in Section 4.2 is accurate but does not amount to a gap. A computational verification of the deletion-contraction identity for a small graph would provide additional empirical support and is the most useful single check.","tokens_in":30622,"tokens_out":29960,"duration_ms":282286,"concrete_test":"Use admcycles or KapranovDegrees in Sage to verify the deletion-contraction identity (10) for a nontrivial graph with at least four vertices, e.g. the 4-cycle C_4, in genus 0 with m=3: compute omega_{C_4}, omega_{C_4 \\ e}, and omega_{C_4/e} as integrals of psi-class pullbacks and check omega_{C_4} = omega_{C_4 \\ e} + omega_{C_4/e}, then compare each with the closed formula (-1)^|V| chi_G(-1).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most delicate part of the manuscript is the proof of Theorem 4.4(3), which asserts that for generic parameters u the intersection of the hypersurfaces L_{v,u} with the fiber pi_M^{-1}(y) is contained in the interior. This is established by an induction that uses Lemma 4.8 to build a sequence of hypersurfaces whose codimensions add one at a time. I checked the two points on which the induction rests: Lemma 4.8(3) supplies the independence of a fresh parameter u_{v_{s+1}, phi(v_{s+1})}, and Lemma 4.8(2) ensures the denominator x_{h_{alpha -> phi(v_{s+1})}} is a legitimate coordinate on the relevant projective space. The codimension count k > |Q_W| - m is exactly what is needed to exceed dim(M_{0,Q_W}) = |Q_W| - 3. I did not find a gap in this argument. Moreover, Theorem 1.4 is already established by the first proof, a self-contained deletion-contraction induction on psi-classes; the second proof is independent confirmation. The base case and the six contributions (C1)-(C6) in Section 3 are consistent, and the toy cases I checked satisfy the recursion. Thus the central claim appears sound.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a family of intersection numbers on moduli spaces of stable curves associated to a finite simple graph G. For a graph G=(V,E) and extra markings M with 2g-2+|M|>0, the class Ψ_{G,g,m} is the product over v of the pullback of ψ_v along the forgetful map remembering the closed neighborhood N[v] together with M, and ω_{G,g,m} is its integral against the pullback of a point class. The main theorem (Theorem 1.4) states that these numbers are given by evaluations of the chromatic polynomial: ω_{G,g,m}=(-1)^{|V|}χ_G(-(2g-2+m)) when (g,m)≠(1,0), with the exceptional formula ω_{G,1,0}=(-1)^{|V|-1}χ'_G(0). Two proofs are given: the first is a self-contained induction on edges using deletion-contraction and intersection-theoretic facts about ψ-classes and boundary strata; the second, in genus zero, identifies the critical points of a Varchenko-type multivalued function on a graphical hyperplane arrangement with the intersection number, after a delicate boundary analysis showing that all intersections occur in the interior of M_0,V⊔M. The paper also connects these numbers to maximum likelihood degrees and scattering amplitudes, and introduces directed-graph analogues χ^in_G and χ^out_G.","tokens_in":30876,"tokens_out":43937,"duration_ms":424925,"significance":"If correct, Theorem 1.4 gives a surprising and clean geometric realization of all negative integer evaluations of the chromatic polynomial, while the exceptional case captures its linear coefficient. The genus-zero special case unifies several previously studied objects: Kapranov degrees, maximum likelihood degrees of graphical models, and bounded regions of graphical hyperplane arrangements. A notable strength is the presence of two independent proofs: the first is direct intersection theory on moduli spaces and does not depend on the more intricate boundary analysis of the second, which substantially raises confidence in the central claim. The paper is also refreshingly explicit: the base cases are computed, small examples are checked, and the relevant computational tools (admcycles, KapranovDegrees) are named.","major_comments":[],"minor_comments":[{"comment":"In the proof of Lemma 4.8, subcase (b), the map φ is only needed on the sequence W, which is chosen as V(C)\\setminus\\{r\\}; the definition φ(r)=m_i is therefore extraneous. If W were instead intended to include r, the choice φ(r)=m_i would violate condition (2) in the case where α induces the trivial partition of M, so the role of the root should be clarified.","section":"Section 4.2, Lemma 4.8, subcase (b)"},{"comment":"The notation π^*_{1234}([pt]) is potentially confusing: the point class on M_0,4 is being identified with a ψ-class on that space. Adding a sentence explaining that ψ_i=[pt] as classes on M_0,4 would make the connection to the factors π^*_{\\{v\\}\\cup A}(ψ_v) in equation (20) easier to follow.","section":"Section 3.1, Fact 2.9"},{"comment":"In the sentence preceding equation (43), the displayed comparison of polynomials should involve χ_{G\\setminus\\{w\\}}(x) on the right-hand side, not χ_G(x); the subsequent displayed conclusion is correct.","section":"Section 5, proof of Proposition 1.13"},{"comment":"The phrase 'Let L_v be as in Theorem 1.10' should read 'as in Theorem 4.4'; the hyperplanes L_v are defined in Section 4.2, not in Theorem 1.10.","section":"Proof of Theorem 1.10"},{"comment":"There is a small typo in the sentence 'one may we characterize arbitrary evaluations of χ_G at nonpositive integer values'; the word 'we' should be deleted.","section":"Section 1.4"}],"recommendation":"minor_revision","confidential_remarks":"The second proof, especially the boundary case analysis in Section 4.2 culminating in Lemma 4.8 and Theorem 4.4(3), is intricate and deserves careful checking by an expert in combinatorial intersection theory. However, the first proof of Theorem 1.4 is self-contained and appears sound, so any residual risk in the second proof does not affect the paper's central claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper. First, the main formula is genuinely new and clean: the intersection number omega_{G,g,m} on Mbar_{g,V⊔M} equals (-1)^|V| chi_G(-(2g-2+m)), with the (1,0) case giving the linear coefficient. Second, it's proved twice, and the first proof is self-contained enough that a skeptical reader can verify it without trusting the second. What the paper does well: Theorem 1.4 gives a geometric realization of all negative chromatic evaluations, and it connects genus-zero intersection numbers to Kapranov degrees, counting bounded regions of a graphical arrangement, and maximum likelihood degrees. The first proof is a deletion-contraction induction on psi-classes, and the base case handles the edgeless graph correctly. The bookkeeping of the six contributions (C1)–(C6) is complicated but the cancellation works out. The second proof, through hyperplane arrangements and Varchenko's critical point theorem, is a nice independent route; the main burden is Theorem 4.4(3), which rules out boundary contributions in the intersection. I went through Lemma 4.8 and the induction—the codimension count k > |Q_W| - m is exactly what is needed, and the fresh parameter trick is legitimate. I did not find a gap. Soft spots: they are real but not fatal. The boundary analysis in Section 4.2 is the sort of thing where a subtle incidence could fail, and it deserves a close read by the referee. The paper is also dense in places; the directed graph section is clearly exploratory and the combinatorial interpretations there are open questions. None of that affects the main theorem, because the first proof stands alone. The citation pattern is clean—Stanley, Zaslavsky, Varchenko are standard, and the only self-citation, Fact 2.9, is a minor elementary computation. Who this is for: anyone working on psi-class intersection numbers, Kapranov degrees, or chromatic polynomials. It is a solid subfield contribution rather than a breakthrough, but it is correct and useful. I'd send it to a serious referee; the first proof alone justifies publication. If it were my call, accept.","headline":"Reinke–Silversmith prove a clean closed formula realizing negative chromatic evaluations as psi-class intersection numbers, with two independent proofs that hold up under scrutiny.","tokens_in":31373,"tokens_out":2514,"would_cite":true,"duration_ms":21954,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14H10","05C31"],"pacs":[],"model":"deepseek-v4-flash","headline":"For every finite simple graph $G$, the intersection numbers $\\omega_{G,g,m}$ on moduli spaces of stable curves equal $(-1)^{|V|}\\chi_G(-(2g-2+m))$, with a derivative formula in the exceptional genus-one case.","keywords":["chromatic polynomial","moduli of stable curves","intersection numbers","psi-classes","hyperplane arrangements","maximum likelihood degree","deletion-contraction","acyclic orientations"],"falsifier":"Take $G=K_3$ and $m=4$; the theorem predicts $\\omega_{K_3,0,4}=24$, since $\\chi_{K_3}(x)=x(x-1)(x-2)$ and $(-1)^3\\chi_{K_3}(-2)=24$. Compute this intersection number on $\\overline{M}_{0,7}$ directly from psi-class and boundary-divisor relations. A result different from 24 would disprove the main theorem; agreement is the expected check.","tokens_in":30428,"feed_emoji":"🎨","tokens_out":17216,"duration_ms":145827,"temperature":0.7,"pith_summary":"For every finite simple graph $G$, the paper defines a family of intersection numbers $\\omega_{G,g,m}$ on moduli spaces of stable curves by pulling back cotangent (psi) classes and integrating, and proves that the family is governed by the chromatic polynomial $\\chi_G$. The main formula is $\\omega_{G,g,m}=(-1)^{|V|}\\chi_G(-(2g-2+m))$ except in the genus-one, no-marked-points case, where it gives the derivative of $\\chi_G$ at zero. If correct, the theorem makes every negative evaluation of the chromatic polynomial into a geometric count, and makes the whole family computable from one classical graph invariant. Two independent proofs are supplied: a deletion-contraction induction in the intersection ring of moduli space, and a route through hyperplane arrangements, bounded regions, and critical points of a product of powers of linear functions.","feed_headline":"Negative chromatic values are stable-curve intersections","feed_subtitle":"The identity ties curve geometry to graph colorings, and connects to hyperplane arrangements and likelihood degrees.","key_machinery":"The central object is the class $\\Psi_{G,g,m}=\\prod_{v\\in V}\\pi^*_{N[v]\\cup M}\\psi_v$ on $\\overline{M}_{g,V\\sqcup M}$, where $N[v]$ is the closed neighborhood of $v$ and $\\psi_v$ is the cotangent class at the marked point $v$. This class encodes the graph's adjacency data in a codimension-$|V|$ cohomology class, and integrating it against $\\pi^*_M([pt])$ produces $\\omega_{G,g,m}$. The argument is carried by two mechanisms: the deletion-contraction identity for these classes, which reduces the intersection number to the edgeless case; and, in the second proof, the identification of logarithmic derivatives of $\\prod_H f_H^{u_H}$ with pullbacks of hyperplane classes under the psi-class linear system from the genus-zero moduli space to projective space. A technical lemma rules out boundary intersections for generic weight choices, so the critical-point locus represents the same intersection product.","core_discovery":"The central discovery is Theorem 1.4: for finite simple $G=(V,E)$ and stable-curve data $(g,m)$ with $2g-2+m>0$, the intersection number built from $\\Psi_{G,g,m}=\\prod_{v\\in V}\\pi^*_{N[v]\\cup M}\\psi_v$ and the pullback of a point class satisfies $\\omega_{G,g,m}=(-1)^{|V|}\\chi_G(-(2g-2+m))$ when $(g,m)\\neq(1,0)$, and in the remaining case $\\omega_{G,1,0}=(-1)^{|V|-1}\\chi'_G(0)$. The paper proves this twice. The first proof is an induction on edges: the identity $\\omega_{G,g,m}=\\omega_{G\\setminus e,g,m}+\\omega_{G/e,g,m}$ mirrors the chromatic polynomial's deletion-contraction recursion, with the equality of numbers coming from an equality of cohomology classes on a single moduli space. The second proof, in genus zero, identifies $\\omega_{G,0,m}$ with the number of bounded regions of a refined graphical arrangement and, through a classical theorem on products of powers of linear functions, with a count of nondegenerate critical points; this also realizes $\\omega_{G,0,m}$ as a maximum likelihood degree. A directed-graph variant of the construction produces two new polynomials that satisfy the same moduli-space formula.","pith_inferences":["One testable extension is to check whether the directed-graph polynomials satisfy a deletion-contraction recursion on directed edges; if they do, they would be genuine chromatic-like invariants with a moduli-space definition.","The maximum-likelihood interpretation points in the opposite direction as well: tautological-ring calculations on moduli space could yield new closed forms for maximum likelihood degrees of graphical statistical models.","The two proofs suggest a dictionary between boundary strata of moduli space and degenerations of colorings or acyclic orientations; making it explicit could produce bijective proofs of the formula and of the directed variants."],"forward_implications":["Every negative integer evaluation of the chromatic polynomial becomes an intersection number on a moduli space of stable curves, so the whole polynomial is encoded in these geometric counts.","In genus zero, the same numbers count bounded regions of refined graphical hyperplane arrangements and nondegenerate critical points of products of powers of linear functions; they are therefore maximum likelihood degrees.","The deletion-contraction identity gives a direct geometric proof of the chromatic polynomial's recursion, reducing all computations to the edgeless case.","For complete graphs, the genus-zero case recovers the classical count of $n!$ critical points of the scattering potential in the $m=3$ model, connecting the formula to scattering amplitudes.","For directed graphs, the construction produces two polynomials of degree $|V|$ that satisfy the same moduli formula and factor as products of linear factors on acyclic digraphs."],"supporting_citations":[{"why":"Interprets negative evaluations of the chromatic polynomial as acyclic orientations with compatible colorings; this is the counting identity behind Theorem 1.8.","marker":"[Sta73, Thm. 1.2]"},{"why":"Gives the number of regions of the graphical arrangement as a negative chromatic evaluation; supplies the arrangement-to-coloring bridge.","marker":"[Zas75, Thm. A]"},{"why":"Supplies the theorem that a product of powers of linear functions has one nondegenerate critical point in each bounded region; this turns region counts into critical-point counts.","marker":"[Var95, Thm. 1.2.1]"},{"why":"Provides the pullback formula for boundary strata on moduli space, used throughout the deletion-contraction manipulation in the first proof.","marker":"[Kee92, Fact 3]"},{"why":"Gives the comparison formula for pulling back psi-classes along forgetful maps, the main computational engine for the induction step.","marker":"[Koc01, Lem. 1.3.1]"},{"why":"Supplies the recursion and vanishing criterion for genus-zero intersection numbers used to discard boundary contributions.","marker":"[BELL23, Thm. B]"},{"why":"Defines the psi-class linear system on genus-zero moduli space used to express logarithmic derivatives as hyperplane pullbacks in the second proof.","marker":"[Kap93]"},{"why":"Provides the likelihood correspondence used to show generic critical points avoid the complete-graph hyperplane arrangement.","marker":"[HS14, Thm. 1.6]"}],"fun_headline_variants":["Graph colorings count stable-curve intersections","Negative chromatic values reveal curve geometry","Stable curves meet graph theory via chromatic polynomials","How coloring a graph counts curves in moduli space","Chromatic polynomial encodes curve intersection numbers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the claim that, for almost all weight choices, the hypersurfaces $L_{v,u}$ representing the intersection classes meet only in the interior of moduli space and never in its boundary; the authors verify this by a long case analysis, and a missed boundary intersection would break the equality between the critical-point count and the intersection number $\\omega_{G,0,m}$.","fun_headline_variants_meta":{"raw":{"variants":["Graph colorings count stable-curve intersections","Negative chromatic values reveal curve geometry","Stable curves meet graph theory via chromatic polynomials","How coloring a graph counts curves in moduli space","Chromatic polynomial encodes curve intersection numbers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000289,"raw_usage":{"total_tokens":1717,"prompt_tokens":994,"completion_tokens":723,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":610,"completion_tokens_details":{"reasoning_tokens":671}},"tokens_in":610,"tokens_out":723,"duration_ms":7408,"temperature":1.0,"reasoning_tokens":671,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:59:53.607760+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $G=K_3$ and $m=4$; the theorem predicts $\\omega_{K_3,0,4}=24$, since $\\chi_{K_3}(x)=x(x-1)(x-2)$ and $(-1)^3\\chi_{K_3}(-2)=24$. Compute this intersection number on $\\overline{M}_{0,7}$ directly from psi-class and boundary-divisor relations. A result different from 24 would disprove the main theorem; agreement is the expected check.","supporting_citations":[],"review_version":1}