{"id":"295ae7c8-ea2d-40ac-a8a1-d2f7e898975e","arxiv_id":"1908.03691","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The equivariant Gromov-Witten potentials of local P^1×P^1 are proven to lie in an explicit finitely generated ring and to satisfy a holomorphic anomaly equation, via Givental-Teleman graph sums.","lead":"This paper proves that certain curve-counting invariants attached to a standard three-dimensional shape in algebraic geometry are organized by a small explicit list of generators, and that they obey a clean derivative rule called a holomorphic anomaly equation. It is one of the first fully worked two-parameter examples in a research program that previously had mostly one-parameter cases.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 3.2's R-matrix ring memberships rest on unverified q-derivative computations and an induction step; the finite generation and anomaly proofs depend on this algebra.","rationale":"The reader's weakest_assumption identifies exactly the same point: Corollary 3.2's ring memberships, and particularly the asserted computations of q-derivatives of M, L, and the norm, are the unverified algebraic foundation for both main theorems. My read of the paper confirms that this is the most load-bearing spot: the architecture of the proof is otherwise standard Givental-Teleman graph summation, and if Corollary 3.2 is correct, the formal structure of the finite generation and anomaly equation arguments is coherent. The paper also has independent support for part of the claim: Section 6 gives a Feynman diagram proof that (R_k)^1 lies in 1/(λ²L+μ²M)^{3k} Q[L,M,λ,μ]_{deg=8k}, i.e., in G_{3k,8k}, and Example 6.4 confirms the k=1 case explicitly. This makes the gap narrower but does not close it: the H1, H2, and H1H2 components, and the induction using their q-derivatives, remain asserted rather than demonstrated. The degree-counting equation in the proof of Theorem 4.1 also appears not to balance as printed, which is consistent with the reader's remark about garbled bookkeeping. Because these are missing computations rather than a demonstrated contradiction, the appropriate verdict remains CONDITIONAL; since that is already the reader's verdict, no adjustment is needed.","tokens_in":20816,"tokens_out":9826,"duration_ms":97565,"concrete_test":"Compute the k=2 R-matrix components directly from the recursion in Lemma 3.1, starting from the explicit (R_1)^1 formula in Example 6.4. First verify symbolically that q1 d/dq1 M_αβ, q2 d/dq2 M_αβ, q1 d/dq1 L_αβ, and q2 d/dq2 L_αβ all lie in G_{1,4} by solving the differentiated defining equations for M_αβ and L_αβ. Then check whether (R_2)^1, (R_2)^{H1}, (R_2)^{H2}, and (R_2)^{H1H2} satisfy the precise memberships asserted in Corollary 3.2, including the denominator pattern X/((1+Ī_11)(Ĩ_22+Ĩ_22(q2,q1))) for the H1H2 component. If any membership fails, the induction in Corollary 3.2 is broken and the proofs of Theorems 4.1 and 5.2 no longer go through.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing point is Corollary 3.2: its memberships for (R_k)^1, (R_k)^{H1}, (R_k)^{H2}, and (R_k)^{H1H2} feed every vertex, tail, and edge contribution in the graph-sum proof of Theorem 4.1 and are also used in Proposition 5.1 to compute d/dX of the R-matrix and bivector. The proof of Corollary 3.2 says: \"By formula (12), we can compute q1 d/dq1 ||e_αβ||/||e_αβ||, q2 d/dq2 ||e_αβ||/||e_αβ|| ∈ G_{2,6}; and q1 d/dq1 M_αβ, q2 d/dq2 M_αβ, q1 d/dq1 L_αβ, q2 d/dq2 L_αβ ∈ G_{1,4}. Then together with induction on the behavior of R_k, we can get this corollary.\" Formula (12) is only the norm identity ||e_αβ|| = 1/(2√(−2λ²L_αβ−2μ²M_αβ)). It does not directly give the memberships for the q-derivatives of M_αβ and L_αβ; those require solving the linear system obtained by differentiating M²−λ² = q1(2(M+L))² and L²−μ² = q2(2(M+L))², and that computation is not displayed. The induction on R_k additionally needs a q-derivative-stability statement for R_k^1, R_k^{H1}, and R_k^{H2} inside the G-index classes; no such lemma is stated. If any of these memberships fails, the degree count in Theorem 4.1 collapses and the R-derivative identities in Proposition 5.1 acquire extra terms, so neither finite generation nor the holomorphic anomaly equation follows as written. This is an internal algebraic gap, not a disagreement with the surrounding literature.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves two structural results for the equivariant Gromov-Witten theory of the local Calabi-Yau threefold K_{P^1 x P^1}: finite generation of the genus-g potentials at the mirror point in the ring G[P1,P2,P3,P4,X], with deg_X F_g <= 3g-3, and a holomorphic anomaly equation expressing d/dX F_g as a quadratic expression in lower-genus correlators. The proof uses the mirror I-function, the quantum differential equation, the Givental R-matrix, and the Givental-Teleman graph-sum formula. A final section gives an oscillatory-integral and Feynman-diagram representation of the first column of the R-matrix.","tokens_in":21197,"tokens_out":6022,"duration_ms":60964,"significance":"If the main theorems are correct, this is a nontrivial two-Kaehler-parameter example of finite generation and the holomorphic anomaly equation, fitting into the program of [3,5,8,9,10,11,15]. The overall architecture is coherent: conditional on the R-matrix ring memberships in Corollary 3.2, the graph-sum degree count in Theorem 4.1 is clean, and the bivector derivative computation in Proposition 5.1 is elegant. The paper also gives a concrete combinatorial handle on the R-matrix via Feynman diagrams and identifies interesting specializations of the equivariant parameters. Its main weakness is that the load-bearing algebraic premise, Corollary 3.2, is left as a proof sketch with key computations not displayed.","major_comments":[{"comment":"Corollary 3.2 is the load-bearing statement for both main theorems, but its proof is only a sketch. The assertion that q1 d/dq1 ||e_alpha beta||/||e_alpha beta|| and q2 d/dq2 ||e_alpha beta||/||e_alpha beta|| lie in G_{2,6}, and that the q-derivatives of M_alpha beta and L_alpha beta lie in G_{1,4}, is said to follow from formula (12). However, formula (12) is only the norm identity ||e_alpha beta|| = 1/(2 sqrt(-2 lambda^2 L_alpha beta - 2 mu^2 M_alpha beta)). The memberships for the derivatives of M_alpha beta and L_alpha beta require differentiating the algebraic system M^2 - lambda^2 = q1(2(M+L))^2, L^2 - mu^2 = q2(2(M+L))^2 and solving a linear system; that computation is not displayed. In addition, the induction 'on the behavior of R_k' needs a q-derivative stability statement for the components (R_k)^1, (R_k)^{H1}, and (R_k)^{H2} within the stated G-index classes, and no such induction lemma is formulated. Since Corollary 3.2 feeds all vertex, edge, and tail contributions in Theorem 4.1 and the R-derivative identities in Proposition 5.1, this gap is load-bearing: if any of these memberships fails, the degree count collapses and the anomaly equation acquires extra terms.","section":"Corollary 3.2"},{"comment":"The graph-sum proof of Theorem 4.1 depends on exact ring memberships for the edge contributions V_k, in particular on the statement that each V_k lies in Q[z,w]_deg=k tensor G_{3(k+1),8(k+1)+3} tensored with the displayed Q-span containing X/((1+Ibar_11)^2(...)). This membership is quoted from Corollary 3.2 and the sentences following it, but it is not independently verified. The proof then concludes deg_X Cont_Gamma F_g <= |E(Gamma)| and cites |E(Gamma)| <= 3g-3 for stable graphs; this final bound is plausible but should be stated with the usual stability inequalities. More importantly, the degree bound 'in the polynomial expression of F_g' requires a definition, because the generators P1,...,P4,X may satisfy relations in the ambient function field; the theorem should specify that there exists a polynomial expression of X-degree at most 3g-3.","section":"Theorem 4.1"},{"comment":"The proof of Proposition 5.1 is too terse. The sentence that the first four derivative equations 'just follow from Lemma 3.1 and Corollary 3.2' hides the actual computation of d/dX of rational coefficients such as X/((1+Ibar_11)(Itilde22(q1,q2)+Itilde22(q2,q1))) and Itilde22(.,.)/(Itilde22(q1,q2)+Itilde22(q2,q1)). The claimed identities d/dX R^{H1}=0, d/dX R^{H2}=0, and d/dX R^{H1H2} = -z/(Itilde22(q1,q2)+Itilde22(q2,q1))(R^{H1}+R^{H2}) require nontrivial cancellations among these derivative terms; they are not shown. The final bivector identity, which is the input to the holomorphic anomaly equation, depends structurally on this exact coefficient. If the derivative computation yields any additional total-derivative term in X, the graph-sum differentiation in Theorem 5.2 will not produce the stated right-hand side. The derivation is plausible, but the missing computation is exactly the part that the stress-test identifies, and it must be supplied or referenced to a verifiable source.","section":"Proposition 5.1"}],"minor_comments":[{"comment":"The notation I^{i;lambda^2}_{22a}(q1,q2) and I^{i;mu^2}_{22a}(q1,q2) is used in Lemmas 2.1-2.4 without an explicit definition of the superscripts and subscripts; please define these series precisely.","section":"Lemma 2.1"},{"comment":"There are many typographical errors, including 'I fucntion', 'Feymann', 'Lebniz's rule', and inconsistent hyphenation in 'Kaehler'; the manuscript would benefit from a careful proofreading pass.","section":"Throughout"},{"comment":"The derivative d/dX is a formal derivative with respect to the generator X of the ring, not an ordinary partial derivative in q1 and q2; this should be stated explicitly when the holomorphic anomaly equation is introduced.","section":"Theorem 1.1"},{"comment":"Corollary 3.2 refers to 'formula (12)', which appears later in Section 6; a forward reference is acceptable, but the norm identity should be restated or numbered earlier to make the proof self-contained.","section":"Corollary 3.2"}],"recommendation":"major_revision","confidential_remarks":"The main theorems are very plausible conditional on the R-matrix ring memberships of Corollary 3.2, and I did not find a circularity problem: the holomorphic anomaly equation is derived, not assumed. The central issue is purely computational but load-bearing. I would expect the author to be able to fill the gap in a revision by displaying the differentiation of the algebraic branch equations and the induction lemma for the R-matrix components; if that computation is supplied, the paper would be a solid contribution to the two-parameter local Calabi-Yau literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick read of Xin Wang's paper. The genuine news: it extends the finite-generation and holomorphic anomaly equation program to a two-parameter local Calabi-Yau, K_{P1xP1}, using the Givental-Teleman R-matrix and graph sums. If the key algebra holds, the main theorem is new and valuable: F_g at the mirror point lies in an explicit finitely generated ring G[P1,P2,P3,P4,X], with deg_X ≤ 3g−3, and the X-derivative satisfies a clean quadratic HAE. The formal derivation of the HAE from Proposition 5.1 by Leibniz's rule is clean, and the Feynman-diagram/oscillatory-integral section gives a concrete combinatorial handle, including a worked example for the first column of R_1 that is a useful check. The paper is honest about scope: it is equivariant, and Remark 6.6 says the non-equivariant limit is not obtained.\n\nThe soft spot is real and load-bearing. Corollary 3.2 asserts that the R-matrix components lie in specific graded pieces of the ring G. The proof says \"by formula (12), we can compute\" the q-derivatives of ||e_αβ||, M_αβ, L_αβ and get the claimed memberships, then \"by induction\" on R_k. Formula (12) only gives the norm in terms of M and L; it does not by itself deliver the q-derivative memberships for M and L, which require differentiating the branch equations M²−λ² = q1(2(M+L))² and L²−μ² = q2(2(M+L))² and solving a linear system. That computation is not shown, and the induction step for R_k needs a q-derivative-stability lemma that is not stated. Since every vertex, tail, and edge contribution in Theorem 4.1 uses these memberships, and Proposition 5.1's derivative formulas use them too, a failure would collapse both the finite generation and the HAE. This is an internal algebraic gap, not a disagreement with the literature, and I don't see a contradiction. But a referee cannot accept the paper without seeing that algebra.\n\nTwo smaller points: the introduction says the example was studied by Lho from different perspectives but never spells out exactly which parts are new versus overlapping; that should be clarified. The degree-counting identity near (8) that looks garbled actually checks out once you use the dimension constraint, so I would not worry about it.\n\nBottom line: the paper deserves a serious referee, and the referee's first job is to verify or refute Corollary 3.2. It is not ready as is, but the architecture is sound and the missing pieces are explicit computations, not a speculative framework. I would bring it to a reading group if someone wants to check those identities, and I would cite it after the algebra is confirmed, not before.","headline":"The paper is a genuine two-parameter extension of the finite-generation/HAE program whose main theorem rests on an unproved R-matrix membership claim—worth refereeing, but the referee must demand the missing algebra.","tokens_in":21947,"tokens_out":5306,"would_cite":false,"duration_ms":51869,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14N35","53D45"],"pacs":[],"model":"deepseek-v4-flash","headline":"For every genus g ≥ 2, the equivariant Gromov–Witten potential of $K_{\\mathbb{P}^1\\times\\mathbb{P}^1}$ at the mirror point lies in a finitely generated ring and obeys a holomorphic anomaly equation.","keywords":["equivariant Gromov-Witten invariants","local P^1 x P^1","finite generation","holomorphic anomaly equation","R-matrix","Feynman diagram expansion","quantum differential equation","mirror symmetry"],"falsifier":"Take the explicit formula for $(R_1)^1_{\\alpha\\beta}$ given in Example 6.4 and expand the first several coefficients in $q_1,q_2$; check directly whether $(R_1)^1_{\\alpha\\beta}\\in G_{3,8}$, i.e. whether the numerator lies in the degree-8 polynomial space in $M_{\\alpha\\beta},L_{\\alpha\\beta},\\lambda,\\mu$ divided by $\\big(\\lambda^2L_{\\alpha\\beta}+\\mu^2M_{\\alpha\\beta}\\big)^3$. Then repeat the recursion of Lemma 3.1 for $k=2$; a single monomial with the wrong numerator degree or denominator power would falsify Corollary 3.2 and hence Theorem 1.1.","tokens_in":20415,"feed_emoji":"📐","tokens_out":6376,"duration_ms":65038,"temperature":0.7,"pith_summary":"This paper proves two structural facts for the equivariant Gromov–Witten theory of the local Calabi–Yau threefold $K_{\\mathbb{P}^1\\times\\mathbb{P}^1}$, the total space of $\\mathcal{O}(-2,-2)$ over $\\mathbb{P}^1\\times\\mathbb{P}^1$. For every genus $g\\ge 2$, after evaluating at the mirror point, the genus-$g$ potential $F_g$ belongs to a finitely generated ring built from four auxiliary functions and one distinguished generator $X$, and the degree of $X$ is at most $3g-3$. The same theorem gives a holomorphic anomaly equation: differentiating $F_g$ with respect to $X$ produces a universal quadratic expression in lower-genus correlators. If the theorem is correct, all higher-genus invariants of this model are recursively determined from low-genus data, and the theory displays the same finiteness that quasi-modularity provides in one-parameter examples.","feed_headline":"GW potentials of local P^1×P^1 are finitely generated","feed_subtitle":"For every genus g≥2 the equivariant potential lies in a ring of five generators, and its X-derivative is a universal quadratic expression.","key_machinery":"The argument runs through the R-matrix of the quantum differential equation. The ring $G=\\bigoplus_{k\\ge 0}G_{k,3k}$ is assembled from homogeneous polynomials in the eigenvalue functions $M_{\\alpha\\beta},L_{\\alpha\\beta},\\lambda,\\mu$ and monomials in $\\lambda^2L_{\\alpha\\beta}+\\mu^2M_{\\alpha\\beta}$. Corollary 3.2 asserts that the entries of the R-matrix lie in these graded pieces, with the $H_1H_2$ component carrying the generator $X$ explicitly. Proposition 6.5 supplies the underlying combinatorial proof: an oscillatory-integral and Feynman-diagram expansion in which each diagram contributes a rational function in $\\lambda^2L_{\\alpha\\beta}+\\mu^2M_{\\alpha\\beta}$ with numerator degree $8k$. Feeding these memberships into the higher-genus graph-sum formula for semisimple Frobenius manifolds gives the finite generation statement, and differentiating the R-matrix and the edge bivector $V$ with respect to $X$ gives the anomaly equation.","core_discovery":"The central claim is Theorem 1.1: for $g\\ge 2$, with $\\tau(q_1,q_2)=I_1/I_0$ the mirror map from the twisted I-function, the genus-$g$ equivariant Gromov–Witten potential satisfies $F_g(\\tau(q_1,q_2))\\in G[P_1,P_2,P_3,P_4,X]$, and the degree of $X$ in such a polynomial expression is at most $3g-3$. Moreover, the derivative with respect to $X$ is exactly $$\\frac{d}{dX}F_g = -\\frac12\\big(\\tilde I_{22}(q_1,q_2)+\\tilde I_{22}(q_2,q_1)\\big)\\Big(\\sum_{g_1+g_2=g}\\langle\\!\\langle H_1+H_2\\rangle\\!\\rangle_{g_1,1}\\langle\\!\\langle H_1+H_2\\rangle\\!\\rangle_{g_2,1}+\\langle\\!\\langle H_1+H_2,H_1+H_2\\rangle\\!\\rangle_{g-1,2}\\Big).$$ Here $\\langle\\!\\langle\\cdot\\rangle\\!\\rangle_{g,n}$ denotes genus-$g$, $n$-marked equivariant correlators. The coefficient and the correlator combination are built from I-function data and lower-genus invariants, so the equation closes among the potentials themselves.","pith_inferences":["Because the R-matrix Feynman expansion is purely combinatorial, the same proof scheme should give finite generation and an anomaly equation for other local surfaces with two Kähler parameters, such as local Hirzebruch surfaces; the paper explicitly mentions this as future work.","The five generators $P_1,P_2,P_3,P_4,X$ are likely special functions of $q_1,q_2$ belonging to a known ring of (quasi-)Jacobi or elliptic forms; identifying that ring would convert the theorem into a concrete quasi-modularity statement.","The non-equivariant specializations $\\lambda=0$ or $\\mu=0$ discussed in Remark 6.6 suggest a route to extract numerical predictions for the ordinary Gromov–Witten potential of local $\\mathbb{P}^1\\times\\mathbb{P}^1$ and to test whether finite generation survives the limit.","The anomaly equation provides an efficient numerical check: starting from genus-0 and genus-1 data, one can compute the first few $F_g$ from the recursion and compare them with direct curve-count localization, which would independently verify the structural claim."],"forward_implications":["All genus-$g$ potentials of this model can be computed recursively from a finite set of five generators, with the bound $\\deg_X F_g\\le 3g-3$ providing a termination criterion.","The holomorphic anomaly equation determines $F_g$ from lower-genus potentials up to a function of the other four generators; the degree bound fixes the remaining ambiguity.","The R-matrix membership in Corollary 3.2 is a statement about twisted I-function data, so the finite-generation result should transfer to any target whose two-variable I-function has the same structure.","All $q_1,q_2$-derivatives of $F_g$ are encoded in the single $X$-derivative together with the relations among $P_1,P_2,P_3,P_4,X$, collapsing the full differential system of the potentials into one anomaly equation.","The theorem upgrades the infinite-dimensional polynomial algebra of invariants to a finite-dimensional master ring $G[P_1,P_2,P_3,P_4,X]$, which is the natural two-parameter analogue of quasi-modularity."],"supporting_citations":[{"why":"Supplies the equivariant virtual fundamental class used to define the Gromov–Witten invariants of $K_{\\mathbb{P}^1\\times\\mathbb{P}^1}$.","marker":"[12]"},{"why":"Provides quantum Riemann–Roch and Birkhoff factorization, the basis for the twisted I-function and the quantum differential equation.","marker":"[2]"},{"why":"Gives the twisted genus-zero Gromov–Witten computation used to write down the I-function of the $\\mathcal{O}(-2,-2)$ twist.","marker":"[1]"},{"why":"Establishes the relation between the fundamental solution $S$ and the R-matrix, which is the starting point of the R-matrix recursion.","marker":"[7]"},{"why":"Supplies the semisimple reconstruction theorem expressing $F_g$ as a sum over stable graphs in terms of R-matrix data.","marker":"[14]"},{"why":"Provides the graph-sum formula with $ψ$-class coefficients that the proof uses to assemble the finite-generation statement.","marker":"[13]"},{"why":"Gives Wick's theorem, the tool by which the oscillatory integral expansion is converted into the Feynman-diagram representation of the R-matrix.","marker":"[4]"}],"fun_headline_variants":["Finite generation and anomaly for GW of K_{P^1×P^1}","Five generators for equivariant GW of local P^1×P^1","Holomorphic anomaly equation for equivariant GW of local P^1×P^1","Equivariant GW potentials of K_{P^1×P^1} are finitely generated"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on Corollary 3.2's graded memberships: the assertions that $q_i\\frac{d}{dq_i}\\|e_{\\alpha\\beta}\\|/\\|e_{\\alpha\\beta}\\|\\in G_{2,6}$ and $q_i\\frac{d}{dq_i}M_{\\alpha\\beta},q_i\\frac{d}{dq_i}L_{\\alpha\\beta}\\in G_{1,4}$, stated by formula (12) without a displayed computation; if any of these fails, the degree count in Theorem 4.1 collapses and the $X$-derivatives in Proposition 5.1 acquire extra terms.","fun_headline_variants_meta":{"raw":{"variants":["Finite generation and anomaly for GW of K_{P^1×P^1}","Five generators for equivariant GW of local P^1×P^1","Holomorphic anomaly equation for equivariant GW of local P^1×P^1","Equivariant GW potentials of K_{P^1×P^1} are finitely generated"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000692,"raw_usage":{"total_tokens":3095,"prompt_tokens":874,"completion_tokens":2221,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":490,"completion_tokens_details":{"reasoning_tokens":2128}},"tokens_in":490,"tokens_out":2221,"duration_ms":16373,"temperature":1.0,"reasoning_tokens":2128,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:09:48.642246+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the explicit formula for $(R_1)^1_{\\alpha\\beta}$ given in Example 6.4 and expand the first several coefficients in $q_1,q_2$; check directly whether $(R_1)^1_{\\alpha\\beta}\\in G_{3,8}$, i.e. whether the numerator lies in the degree-8 polynomial space in $M_{\\alpha\\beta},L_{\\alpha\\beta},\\lambda,\\mu$ divided by $\\big(\\lambda^2L_{\\alpha\\beta}+\\mu^2M_{\\alpha\\beta}\\big)^3$. Then repeat the recursion of Lemma 3.1 for $k=2$; a single monomial with the wrong numerator degree or denominator power would falsify Corollary 3.2 and hence Theorem 1.1.","supporting_citations":[{"cited_title":"Li and G","cited_arxiv_id":null,"evidence_quote":"Supplies the equivariant virtual fundamental class used to define the Gromov–Witten invariants of $K_{\\mathbb{P}^1\\times\\mathbb{P}^1}$."},{"cited_title":"Coates and A","cited_arxiv_id":null,"evidence_quote":"Provides quantum Riemann–Roch and Birkhoff factorization, the basis for the twisted I-function and the quantum differential equation."},{"cited_title":"Coates, A","cited_arxiv_id":null,"evidence_quote":"Gives the twisted genus-zero Gromov–Witten computation used to write down the I-function of the $\\mathcal{O}(-2,-2)$ twist."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the relation between the fundamental solution $S$ and the R-matrix, which is the starting point of the R-matrix recursion."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the semisimple reconstruction theorem expressing $F_g$ as a sum over stable graphs in terms of R-matrix data."},{"cited_title":"Pandharipande, A","cited_arxiv_id":null,"evidence_quote":"Provides the graph-sum formula with $ψ$-class coefficients that the proof uses to assemble the finite-generation statement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives Wick's theorem, the tool by which the oscillatory integral expansion is converted into the Feynman-diagram representation of the R-matrix."}],"review_version":1}