{"id":"71900946-e0c4-4661-82c5-ebfdfc7b68a4","arxiv_id":"2505.03487","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A wall-crossing formula at the level of cycles on moduli of curves relates stable-map cycles to admissible-cover cycles, refining the ELSV formula and the Gromov-Witten/Hurwitz correspondence.","lead":"The authors prove a cycle-level formula that expresses Gromov-Witten cycles of curves as sums of Hurwitz cycles, which come from admissible covers, with correction classes attached. This unifies two major enumerative geometry frameworks and yields new proofs of the ELSV formula and the Gromov-Witten/Hurwitz correspondence.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 2.7 hinges on the cycle-level residue identity (3.3), which is only partially proved: it is delegated to numerical wall-crossing results and Remark 3.3 invokes the (h,m)=(0,1) case proved only inside Theorem 6.1, whose final step uses (3.5), derived from (3.3).","rationale":"The reader's weakest_assumption correctly identifies equation (3.3) as the load-bearing step. My review confirms this: the cycle-level upgrade of the wall-crossing is precisely where the cited references are used, and Remark 3.3 exposes a potential circularity that is not resolved in the text. The Appendix A computations are independent and are genuine evidence, but they verify only low-genus cases of degree 1 and 2 maps to rubber P^1 with two relative points and trivial insertions; they do not establish the general cycle-level localization formula or the (0,1) wall-crossing. The theorem may well be true, and the intended proof strategy is plausible, but as printed the complete proof is not supplied. This does not change the reader's verdict: CONDITIONAL remains appropriate, and the proposed concrete check would settle whether the concern is purely expositional or reflects a genuine gap.","tokens_in":23327,"tokens_out":5801,"duration_ms":53591,"concrete_test":"Re-derive (3.3) as a standalone lemma directly from the master-space fixed locus description (3.1) and the obstruction analysis of [Nes24, Prop. 6.19], without invoking Theorem 6.1 or (3.5). Concretely, compute the residue contributions of the F_\\Gamma terms as classes in H^*(M_{g,n}(X,\\mu)), tracking the divisor classes D_i, the ordering/automorphism factors, and the coefficients [I_{g_i,n_i,\\eta_i}(z-\\tilde{\\psi}_i)]_{b_\\ell}; then isolate the (h,m)=(0,1) wall-crossing as an independent lemma proved from the explicit master space of Section 6.1. If the cycle-level residue formula agrees with (3.3) and the (0,1) lemma can be proved without (3.5), the circularity is resolved; if any step requires the numerical pushforward or uses (3.5), Theorem 2.7 remains unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a cycle-level identity in H^*(M_{g,n}), and its proof is driven entirely by the master-space residue computation. The paper states (3.2) from [Nes24, Prop. 6.19], but the passage from (3.2) to (3.3) is not carried out: it is asserted to follow from the analysis in [Zho22, Thm 7.3.3], and Remark 3.3 adds that the (h,m)=(0,1) wall-crossing, presented inside the proof of Theorem 6.1, is needed. Theorem 6.1 finishes by applying (3.5), the wall-crossing relation of Section 3.6 whose proof uses (3.3). As written, this is a circular dependency: (3.3) requires Theorem 6.1, Theorem 6.1 requires (3.5), and (3.5) requires (3.3). In addition, the cited results in [Nes24] and [Zho22] are formulated for numerical wall-crossing and pushforwards to a point, while the theorem needs the same identity for pushed-forward cycle classes in H^*(M_{g,n}(X,\\mu)), including the substitution z = -\\tilde{\\psi}_i with negative powers set to zero. The low-genus checks in Appendix A are strong evidence for a special case (rubber P^1 with two relative points, trivial insertions), but they do not cover the general (h,m) set-up or the (0,1) wall. Thus the proof of Theorem 2.7 is incomplete as printed.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Fix a degree d, a moving target curve of genus h, and ramification profiles assigned to m target markings. The paper claims a cycle-valued wall-crossing formula (Theorem 2.7): the Gromov–Witten cycle of degree-d stable maps, pushed forward to the moduli space of possibly disconnected source curves, equals a sum over star-shaped graphs of gluings of a Hurwitz cycle (at the root vertex) with products of explicit I-functions (at contracted vertices), the I-functions being tautological classes in lambda- and psi-classes given in Proposition 2.6. The claimed applications are: a cycle-level refinement of the ELSV formula (Theorem 6.1 and (6.3)); a recursion for hyperelliptic cycles (Corollary 5.1); effective computation of Torelli pullbacks of [A1 x A_{g-1}] (Corollary 4.1); and a new proof of the Okounkov–Pandharipande Gromov–Witten/Hurwitz correspondence (Section 8). The proof proceeds by localization on the epsilon-unramification master spaces of [Nes24] with Zhou's entangled-tail analysis [Zho22], and is verified computationally in Appendix A (Schmitt) for rubber P1 with two relative points, degrees 1 and 2, in low genus.","tokens_in":23627,"tokens_out":38400,"duration_ms":326026,"significance":"If Theorem 2.7 is valid, it is a strong and useful result: a geometric, cycle-theoretic explanation of the relation between admissible covers and stable maps, with all correction terms given by explicit, parameter-free lambda/psi formulas, subsuming ELSV and the OP06 correspondence. The degree-one and degree-two applications (Torelli pullbacks; hyperelliptic loci) are new and would yield effective recursions. Positive features: the I-functions are fully explicit (Proposition 2.6); Appendix A provides reproducible computational checks, with code linked in the paper, against independently computed double-ramification and hyperelliptic cycles, in tautological rings of rank up to 838; the checks are not circular, since the comparison cycles are computed via [JPPZ17] and [SvZ20]. The risk is concentrated exactly where the claimed novelty lies: the proof of the cycle-level residue identity (3.3) is delegated to numerical wall-crossing results, and Remark 3.3 introduces a self-referential dependency on Theorem 6.1 (see major comments). The verification covers only the (h,m)=(0,2) case with trivial insertions and does not exercise the (0,1) wall.","major_comments":[{"comment":"The proof of Theorem 2.7 contains a circular dependency at its engine. Remark 3.3 states that passing from (3.2) to (3.3) requires the (h,m)=(0,1) wall-crossing 'presented in the proof of Theorem 6.1.' The proof of Theorem 6.1 derives (6.2) from the explicit empty-chamber master space, but then finishes by applying formula (3.5) ('To finish the proof, we apply the formula (3.5) to the class on the right of (6.2)'), and (3.5) is obtained in §3.6 from (3.4), which is a consequence of (3.3). Hence the argument runs in a cycle: (3.3) depends on Theorem 6.1, Theorem 6.1 depends on (3.5), and (3.5) depends on (3.3). The polar-part identity needed in Remark 3.3 is the content of Theorem 6.1's formula, not merely the intermediate statement (6.2), so a charitable reading does not break the cycle. The (h,m)=(0,1) wall-crossing must be proved directly, independently of (3.3)/(3.5), before (3.3) is derived; without that, Theorem 2.7 rests on an unproved lemma.","section":"§3.5 (Remark 3.3), §3.6 (3.5), §6.1"},{"comment":"The passage from (3.2) to (3.3) in §3.5 is the load-bearing residue computation, and it is not carried out in the paper: (3.2) is imported from [Nes24, Prop. 6.19], (3.3) is attributed to 'the analysis presented in the proof of [Zho22, Theorem 7.3.3]', and the divisor classes D_i are deliberately left unspecified. The cited results are numerical wall-crossing statements (pushforwards to a point), whereas the paper needs (3.3) as an identity of pushed-forward cycle classes in H*(M_{g,n}(X,mu)), including the substitution z = -psi-tilde_i with negative powers set to zero. The b-sum in (3.3), the identification of the D_i contributions with polar parts of I-functions (Remark 3.3), and the cancellation of the b<0 terms leading to (3.4) all require justification at the level of cycles, not numbers; this is exactly the 'extra layer of complexity' announced in §1.2. Please give a complete derivation of (3.2) to (3.3), or state a precise lemma under which the [Zho22]/[Nes24] analysis applies verbatim to pushed-forward cycle classes and verify its hypotheses.","section":"§3.5, equations (3.2)–(3.4)"}],"minor_comments":[{"comment":"The displayed list of degree-two I-functions should be proofread against (2.5) and Proposition 2.6: entries with a (0,1) connected component, such as the one over M^◦_{1,1} × M^◦_{0,1}, appear to omit the factor z^{-1} contributed by I^◦_{0,1,(1)}, and some genus subscripts do not match the component genera; since the verification in Appendix A depends on these formulas, please state the corrected table explicitly.","section":"§5.3"},{"comment":"The step from Theorem 6.1 to (6.3) appeals to a dimension constraint in order to discard all contributing I-functions except I_{0,(2)}; since this is the point where the ELSV formula is recovered, the dimension count, including the treatment of the stable no-marking I-functions of Lemma 7.3, should be written out rather than asserted in one sentence.","section":"§6.2, equation (6.3)"},{"comment":"The class gamma_Gamma of (2.3) is built from source-marking insertions gamma_j that are first defined on the universal curve over M^◦_{h,m}, while the Hurwitz cycle H_{mu,eta} evaluates them at the k moving target markings, i.e., on the universal curve over M^◦_{h,m+k}; the base change implicit in this identification is never stated and should be clarified.","section":"§2.3, §2.8"}],"recommendation":"major_revision","confidential_remarks":"The core of the proof relies on [Nes24] and [Nes22], both preprints by the first author, for the epsilon-unramified moduli spaces, the master-space construction and perfect obstruction theory, properness, and Proposition 6.19 entering (3.2); refereeing the present paper therefore depends on unpublished work. I recommend that the editor confirm the status of [Nes24] (arXiv:2405.18398) and consider asking the authors to state the imported results as precise quoted lemmas. The circular dependency flagged in major comment 1 is the main blocker. The paper fits the journal's scope well; the advertised applications, in particular the degree-one application to [COP24] and the hyperelliptic recursion, are genuinely novel if the main theorem is secured."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing you should know: this is a serious paper with a real proof gap. Theorem 2.7 is a cycle-valued wall-crossing formula that expresses Gromov–Witten cycles as a sum over star-shaped graphs of Hurwitz cycles on the root component times I-functions on contracted components. If it is right, it is a genuine cycle-level refinement of ELSV and of the Gromov–Witten/Hurwitz correspondence, and the applications in Sections 4–8 (Torelli pullbacks, hyperelliptic cycles, completed cycles) are worth taking seriously.\n\nWhat is new: the formula itself. It is not a relabeling of [Nes22, Nes24] or [OP06b]; those were numerical wall-crossings or integrated statements. Here the identity is in H*(M_{g,n}), with z substituted by -psi_tilde and negative powers killed. The organization is clear and the I-function bookkeeping is careful. Credit where due: Appendix A by Schmitt gives independent low-genus checks using admcycles, with code on GitLab, for degree 1 and 2 maps to rubber P^1 relative to two points. Those are real computations, not heuristic argument, and they match in tautological rings of reasonably high rank. That is meaningful evidence.\n\nNow the soft spot, and it is not minor. The engine of the proof is the residue identity (3.3). The paper asserts it follows from [Nes24, Prop. 6.19] and [Zho22, Thm 7.3.3], but those are numerical statements, and the cycle-level lift is not shown. Worse, Remark 3.3 says (3.3) needs the (h,m)=(0,1) wall-crossing 'presented in the proof of Theorem 6.1,' while Theorem 6.1's proof finishes by applying (3.5), which is derived from (3.3). As printed, (3.3) depends on Theorem 6.1, Theorem 6.1 depends on (3.5), and (3.5) depends on (3.3). That is a circle. It might be that the (0,1) wall-crossing can be proved independently—the master space in Section 6 is different and simpler—but the authors do not do that separation here. The appendix does not cover the (0,1) wall or general (h,m), so it does not close the loop. The fix is probably straightforward: isolate the (0,1) wall-crossing as its own lemma, prove it without (3.5), and then derive (3.3). But as it stands, Theorem 2.7 is not established. There are also minor numbering typos (e.g., Corollary 7.6 cites itself), but those are easily fixed.\n\nThe paper is not a lost cause. The cited machinery exists, the computations are concrete, and the applications are well-motivated. I would send it to a serious referee, with the explicit request that the proof gap be addressed or resolved. The authors should be given the chance to extract the missing lemma.","headline":"Cycle-level GW/Hurwitz formula that looks true and is computationally supported, but the printed proof has a genuine circular dependency that must be fixed before publication.","tokens_in":24260,"tokens_out":3900,"would_cite":true,"duration_ms":35881,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14N35","14H10","14N10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves a cycle-valued wall-crossing formula expressing every Gromov–Witten cycle as a sum of Hurwitz cycles with explicit I-function corrections, refining the ELSV formula and the Gromov–Witten/Hurwitz correspondence.","keywords":["admissible covers","stable maps","Hurwitz cycles","Gromov-Witten cycles","wall-crossing","ELSV formula","I-functions","moduli of curves"],"falsifier":"Verify both sides of Theorem 2.7 as explicit classes in the tautological ring for degree $d=2$ maps to the relative $\\mathbb{P}^1$ with profile $((2),(2))$ at source genus $g=4$, a case beyond the paper's table; any nonzero difference in a graph coefficient would settle that the cycle-valued equality is false.","tokens_in":23003,"feed_emoji":"📐","tokens_out":12566,"duration_ms":107460,"temperature":0.7,"pith_summary":"Two compactifications of the space of degree-$d$ maps to a moving curve — stable maps and admissible covers — both yield cycles on the moduli space $\\mathcal{M}_{g,n}$. The paper proves a cycle-valued wall-crossing formula: every Gromov–Witten cycle $GW_{g,\\mu}(\\gamma;\\alpha)$ is equal to a finite sum, over star-shaped graphs, of Hurwitz cycles $H_{\\mu,\\eta}(\\gamma_\\Gamma;\\alpha)$ glued to I-functions $I_{g_i,n_i,\\eta_i}(-\\tilde{\\psi}_i)$ that encode contracted components and ramifications. The I-functions admit explicit closed forms in $\\lambda$- and $\\psi$-classes obtained by equivariant localization on $\\mathbb{P}^1$. Because the equality holds for cycles rather than only for numbers, the paper recovers the ELSV formula and the Gromov–Witten/Hurwitz correspondence as refinements and extends them to arbitrary descendent insertions and moving targets. Low-genus computations for degree one and two maps to a relative $\\mathbb{P}^1$ confirm the formula on tautological cohomology.","feed_headline":"Stable-map cycles split into admissible-cover sums","feed_subtitle":"A cycle-valued wall-crossing formula refines ELSV and unifies Gromov–Witten with Hurwitz theory.","key_machinery":"The proof is driven by the master space for $\\epsilon$-unramification, a stability parameter that interpolates from stable maps ($\\epsilon\\ll1$) to admissible covers ($\\epsilon>1$); for each wall $\\epsilon_0=1/d_0$ there is a proper master space with a $\\mathbb{C}^*$-action, and its fixed loci are indexed by star-shaped graphs. Virtual localization on this master space, together with the entangled-tail residue identity, converts the wall-crossing into the graph sum. The I-function is the localized virtual class of the space of contracted components,\n\\[\nI_{g,n,\\eta}(z)=z\\prod_{j=1}^{\\ell(\\eta)}\\eta_j\\,\\pi_*\\left(\\frac{[V_{g,n,\\eta}^{\\mathbb{C}^*}]}{e_{\\mathbb{C}^*}($N^{{\\mathrm{vir}}$})}\\right),\n\\]\nexpanded in the range $|z|>1$; Proposition 2.6 evaluates it in terms of the Hodge bundle $\\Lambda^\\vee(z)$ and $\\psi$-classes. The star-shaped graph carries the Hurwitz cycle on its root vertex and I-functions on its non-root vertices, and the gluing map $\\mathrm{gl}_\\Gamma$ assembles these factors by identifying labelled markings; dividing by $|\\operatorname{Aut}(\\Gamma)|$ removes overcounting.","core_discovery":"The central statement, Theorem 2.7, is that for $(h,m)\\neq(0,1)$,\n\\[\nGW_{g,\\mu}(\\gamma;\\$\\alpha$)=\\sum_{\\Gamma}\\frac{1}{|\\operatorname{Aut}(\\Gamma)|}(\\mathrm{gl}_\\Gamma)_*\\left(H_{\\mu,\\eta}(\\gamma_\\Gamma;\\$\\alpha$)\\boxtimes\\prod_{i=1}^{k} I_{g_i,n_i,\\eta_i}(-\\tilde{\\psi}_i)\\right),\n\\]\nwhere the sum is over star-shaped graphs whose edge labels are partitions $\\eta_i\\vdash d$, the genus is fixed by $g=\\sum_i(g_i+\\ell(\\eta_i))+g_0-k$, and the degree relation reads $2g_0-2=d(2h-2)+\\sum_i(d-\\ell(\\mu_i))+\\sum_i(d-\\ell(\\eta_i))$; negative powers of $\\psi$-classes are set to zero. The left-hand side is the pushforward of the virtual fundamental class of degree-$d$ stable maps to the universal curve over $\\mathcal{M}_{h,m}$, with ramification profiles $\\mu$ at target markings and cohomology insertions $\\gamma$ and $\\alpha$. The right-hand side is a graph sum of Hurwitz cycles — admissible covers with profiles $\\mu$ and $\\eta$ — with an I-function correction attached at every contracted component. The special case $(h,m)=(0,1)$ is governed by a separate polar wall-crossing formula, Theorem 6.1, which recovers the ELSV formula by extracting the coefficient of the minimal power of $z$.","pith_inferences":["Beyond the paper, the same master-space residue calculus is likely reusable for other stability changes, producing cycle-level expansions for loci such as $k$-fold ramification cycles or spin Hurwitz loci whenever analogous polar I-functions exist.","Beyond the paper, reading the I-functions as universal classes of contracted components suggests an operator-valued graph sum in which the correspondence becomes a change of basis on the infinite wedge; the paper establishes the numerical version of this statement.","Beyond the paper, the recursion should produce graph-sum formulas for the hyperelliptic loci $Hyp_{g,2,0}$ for every genus, a case the paper leaves open; the resulting genus-two formula can be checked against existing computations."],"forward_implications":["The ELSV formula and the Gromov–Witten/Hurwitz correspondence hold as equalities of cycle classes, so they apply to arbitrary descendent insertions and to a moving target curve, not only to point insertions on a fixed curve.","The degree-one specialization gives an effective way to compute the Torelli pullback of $A_1\\times A_{g-1}$ over $\\mathcal{M}_g$, reproducing the known compact-type values $24\\lambda_2$ in genus three and $20\\lambda_3$ in genus four.","The degree-two specialization relates double-ramification cycles $DR_g(2,-2)$ to hyperelliptic cycles $Hyp_g$, giving a recursion that computes high-genus hyperelliptic cycles from lower-genus ones and from I-functions.","The numerical invariants in the formula — Hurwitz numbers, Fulton–MacPherson integrals, and Hodge integrals — are all explicitly computable, so the theorem yields concrete enumerative numbers for arbitrary insertions."],"supporting_citations":[{"why":"Constructs the $\\epsilon$-unramified master space with perfect obstruction theory, proves properness, and supplies Proposition 6.19 analysing fixed-component obstruction theory.","marker":"[Nes24]"},{"why":"Provides the entangled-tail residue theorem used to derive the residue identity (3.3) on the master space.","marker":"[Zho22]"},{"why":"Virtual localization formula, the basis for computing I-functions and wall-crossing residues.","marker":"[GP99]"},{"why":"Gives the analytic-continuation operator formulas and unstable conventions for I-functions and Hodge integrals used in the numerical sections.","marker":"[OP06a]"},{"why":"Establishes the Gromov–Witten/Hurwitz correspondence and completed cycles that Theorem 8.1 recovers.","marker":"[OP06b]"},{"why":"Supplies the ELSV formula that Section 6 re-derives by taking the minimal-power coefficient of the $(0,1)$ wall-crossing formula.","marker":"[ELSV01]"},{"why":"Introduces admissible covers, the compactification underlying the Hurwitz cycles in the formula.","marker":"[HM82]"},{"why":"Introduces stable maps and their virtual fundamental classes, defining the Gromov–Witten side.","marker":"[Kon95]"},{"why":"Earlier wall-crossing framework for $\\epsilon$-admissibility without source markings that the present proof extends to cycles.","marker":"[Nes22]"},{"why":"Provides the Torelli-pullback computations in genera 3 and 4 that the degree-one specialization reproduces on the compact-type locus.","marker":"[COP24]"}],"fun_headline_variants":["Stable-map cycles decompose via admissible covers","Cycle wall-crossing unifies GW and Hurwitz","ELSV gets cycle-theoretic refinement","Admissible-cover graphs split stable maps"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on a residue identity on the master space whose derivation invokes the $(0,1)$ wall-crossing, while that $(0,1)$ wall-crossing is proved by applying the same formula; unless this circular dependency is broken by an independent argument, the main theorem is not established.","fun_headline_variants_meta":{"raw":{"variants":["Stable-map cycles decompose via admissible covers","Cycle wall-crossing unifies GW and Hurwitz","ELSV gets cycle-theoretic refinement","Admissible-cover graphs split stable maps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000234,"raw_usage":{"total_tokens":1491,"prompt_tokens":933,"completion_tokens":558,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":549,"completion_tokens_details":{"reasoning_tokens":502}},"tokens_in":549,"tokens_out":558,"duration_ms":6077,"temperature":1.0,"reasoning_tokens":502,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:50:44.350148+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Verify both sides of Theorem 2.7 as explicit classes in the tautological ring for degree $d=2$ maps to the relative $\\mathbb{P}^1$ with profile $((2),(2))$ at source genus $g=4$, a case beyond the paper's table; any nonzero difference in a graph coefficient would settle that the cycle-valued equality is false.","supporting_citations":[],"review_version":1}