{"id":"02c7e249-987f-4f05-aff6-b866e4e85002","arxiv_id":"2506.22640","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For moduli spaces of pointed admissible A-covers, the rational homology in each fixed degree and genus is finitely generated over a new combinatorial category, with generation degree at most g+5i.","lead":"This paper proves that the rational homology groups of moduli spaces of admissible covers, in a fixed degree and genus, stabilize as the number of marked points grows, with an explicit generation bound of g+5i. This implies the corresponding rank generating functions are rational with explicit poles, improving earlier bounds for the moduli space of curves.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 5.2 appears false: for A=V4, X={1,2,3}, ℓ=(e1,e2,e1+e2), transition data are trivial while M^A_{0,3}(ℓ) has two components; the §5 module action is then not well-defined.","rationale":"The reader's weakest assumption identifies the same lemma, so I agree with that judgment. The concern is not merely that [BCK24, Prop 2.4] is cited without proof: the invariant stated in Lemma 5.2 is structurally unable to distinguish components whenever the monodromy labels generate A, and the V4 example shows such components exist. The X=3 case is a clean test because the target moduli is a point, so there is no braid monodromy to connect the two orbits. The example uses only the paper's own definitions: source marked points are part of the data (Definition 2.2), and transition data are quotiented by all labels (Definition 5.1). A second consequence is that Definition 6.1's V0,0 disagrees with the geometric H0 in this example, so even the base case of the induction fails. Other potential issues, such as the purity assertion in Proposition 6.4, are secondary: if the action is not well-defined, the induction in Theorem 6.6 never gets off the ground. I therefore recommend REJECT, conditional only on this finite calculation confirming the two orbits.","tokens_in":21214,"tokens_out":38665,"duration_ms":445987,"concrete_test":"Run the finite computation: for A=V4, X={1,2,3}, ℓ=(e1,e2,e1+e2), fix C=P1 with branch points 0,1,∞ and take the pointed V4-cover E={v^2−u^2=1}. List the 8 triples of lifts (one over each branch point) and the V4-action (u,v)↦(±u,±v); count the orbits, predicting 2. Independently compute the invariant of Definition 5.1 and Lemma 5.2 for representatives of both orbits; it is constant. If the orbit count is 2 (and the two objects are non-isomorphic), Lemma 5.2 and the base case of §6.1 are false and the module structure in §5 cannot be defined. If the orbit count is 1, my reading of Definition 5.1 or of the cover is wrong and the concern is resolved.","verdict_should_be":"REJECT","load_bearing_attack":"Section 5 constructs the ^FWS^op_A-module action via gluing maps φ_{(f,g)*}, and well-definedness is justified solely by Lemma 5.2: a connected component of M^A_{0,X}(ℓ_X) is said to be determined by transition data b_{x1,x2} ∈ A/⟨ℓ_X(x)|x∈X⟩. This invariant is too coarse when the labels generate A. Take A=V4, X={1,2,3}, ℓ=(e1,e2,e1+e2). The quotient A/⟨ℓ_X⟩ is trivial, so every cover has the same b-data. For a fixed target (P1,0,1,∞), a cover is the conic v^2−u^2=c with deck group acting by (u,v)↦(±u,±v); over each branch point the fiber has two points, so choosing one lift at each branch point gives 8 triples. The diagonal V4-action is free (joint stabilizer ⟨e1⟩∩⟨e2⟩∩⟨e1+e2⟩={0}), hence there are exactly 2 orbits. These orbits are non-isomorphic pointed covers, so the zero-dimensional stack M^A_{0,3}(ℓ) has two connected components, while Lemma 5.2 puts them in one. Consistently, Definition 6.1 defines V0,0(X,ℓ_X)=Q for this object, but the geometric H_0 is Q^2, contradicting the stated coincidence. Since the gluing fibers f^{-1}(y)⊔{y} have at least three points, this failure is load-bearing: without Lemma 5.2 the maps (f,g)_* are not well-defined and Theorem B does not follow.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces combinatorial categories FWSA and ^FWSA whose objects are A-labelled finite sets, and proposes to make the rational homology groups H_i(M^A_{g,X}(ℓ_X);Q) into a ^FWS^op_A-module by gluing genus-zero admissible covers. The main structural result, Theorem B, asserts finite generation of this module in degree at most g+5i; Theorem A and Corollaries C and D then derive rationality and pole bounds for the associated rank generating functions. The proof is inductive: the suspension map of Proposition 4.4 is compared with geometric gluing maps, Harer's vcd bound supplies surjectivity in high degree, and the component classification of genus-zero admissible covers quoted as Lemma 5.2 is used to make the gluing maps well-defined.","tokens_in":21610,"tokens_out":18207,"duration_ms":215299,"significance":"If correct, the paper would provide a substantial strengthening of the effective representation stability results for M_{g,n} and the first effective finite-generation theorem for homology of moduli spaces of pointed admissible A-covers, with an appealing and conceptually simple inductive strategy. The category-theoretic framework is natural, the algebraic sections are developed in detail, and the paper is transparent about its reliance on [BCK24] for the component classification. However, the central construction of the module structure depends on a component classification that I believe is false; the main theorems are therefore not established in the present form.","major_comments":[{"comment":"Lemma 5.2 is false. Let A=V4, X={1,2,3}, and ℓ_X=(e1,e2,e1+e2). Since the labels generate A, the quotient A/⟨ℓ_X(x)|x∈X⟩ is trivial, so every genus-zero cover has the same transition data b_{x1,x2}=0. The underlying unpointed V4-cover of P1 branched at three points is unique and connected; a pointed cover is a choice of one of the two lifts over each of the three branch points. There are 8 such choices, and the diagonal V4-action on them is free because ⋂_x⟨ℓ_X(x)⟩=0, so there are exactly two isomorphism classes. Since M_{0,3} is a point, these two classes are two distinct connected components of M^A_{0,3}(ℓ_X). Thus condition (2) of Lemma 5.2 holds for covers lying in different components, contradicting condition (1).","section":"§5, Lemma 5.2"},{"comment":"Because Lemma 5.2 is false, the connected component T_y of M^A_{0,f^{-1}(y)⊔y}(ℓ_X|f^{-1}(y), -ℓ_Y(y)) is not determined by the transition datum g|_{f^{-1}(y)}. This failure is load-bearing for the module action. Concretely, let X={1,2,3,4}, Y={y,z}, ℓ_X(1)=e1, ℓ_X(2)=e2, ℓ_X(3)=ℓ_X(4)=0, ℓ_Y(y)=e1+e2, ℓ_Y(z)=0, and let f:X→Y be the surjection with f^{-1}(y)={1,2} and f^{-1}(z)={3,4}. For any pointing g, the space T_y is M^A_{0,3}(e1,e2,e1+e2), which has two connected components, although all transition data are trivial. Hence the gluing map φ_{(f,g)*}, and therefore the homology map (f,g)_*, is not well-defined independently of the omitted choices. This invalidates the construction of the ^FWS^op_A-module structure on which Theorem B, Theorem A, and Corollaries C and D rest.","section":"§5, Definition 5.3 and equation (7)"},{"comment":"The claimed identification of V0,0 with H_0(M^A_{0,X}(ℓ_X);Q) for |X|≥3 is also false. For the example A=V4, X={1,2,3}, ℓ_X=(e1,e2,e1+e2), the set S0,0(X,ℓ_X)=(A/⟨ℓ_X(x)|x∈X⟩)^X/A is a single point, so V0,0(X,ℓ_X)=Q. But M^A_{0,3}(ℓ_X) has two connected components, so H_0(M^A_{0,3}(ℓ_X);Q)=Q^2. Since the quotient map q: eV0→V0,0 and the base case of the induction in Theorem 6.6 depend on this identification, the inductive proof of finite generation is not geometrically justified.","section":"§6.1, Definition 6.1"}],"minor_comments":[{"comment":"The symbol P is used in condition (3) before a source curve P has been named; the condition should refer to the named source curve E or the source curve should be named P consistently.","section":"§2, Definition 2.1(3)"},{"comment":"The text says \"morphism in ^FWS^op_A\" where the morphism is an arrow in ^FWS_A; this is harmless but should be corrected for clarity.","section":"§5, before equation (7)"},{"comment":"The notation M^A_{g,X}(ℓ_X) is used for both the compact admissible-cover stack and for its interior in the Borel–Moore exact sequence. Please clarify the two uses and state explicitly why H_i of the compact stack is pure of weight 2i.","section":"§6, Proposition 6.4"},{"comment":"The index of summation in the definition of I_{i,g} is written as (i1,g1)+(i2,g2)=(g,i); this should be (i,g).","section":"§6, Theorem 6.6 proof, equation (11)"}],"recommendation":"reject","confidential_remarks":"The counterexample to Lemma 5.2 is explicit and can be checked directly: the zero-dimensional moduli space M^A_{0,3}(e1,e2,e1+e2) for A=V4 has two reduced points while the lemma predicts one. Since the same triple of labels appears as a gluing fiber in an actual morphism of ^FWS_A, the failure is not an isolated edge case of the classification; it breaks the well-definedness of the module action. The authors would need either a correct component classification with enough data to define the gluing maps, or a different construction of the module action, before the main theorems can be assessed. I would be willing to look at a substantially revised version that addresses this point."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline is simple: the central construction looks unsound. Theorem B rests on Lemma 5.2, which claims the connected component of a point in M^A_{0,X}(ℓ_X) is determined by the pairwise transition data b_{x1,x2} in A/⟨ℓ_X(x)⟩. That quotient is too coarse. Take A=V4, X={1,2,3}, ℓ=(e1,e2,e1+e2). The labels generate A, so the quotient is trivial and every cover has the same b-data. But a genus-zero V4-cover branched at three points is a conic; over each branch point the fiber has two points, so there are 8 choices of one lift per branch point, and the diagonal V4 action on these triples is free. Hence there are two non-isomorphic pointed covers, i.e., two connected components. Lemma 5.2 puts them in one. Definition 6.1 then gives V_{0,0}(X)=Q for this object while H_0 is Q^2. Since the gluing maps φ_{(f,g)*} are only well-defined if the connected component Ty is uniquely determined by (f,g), the ^FWS^op_A-module structure is not well-defined. The authors cite [BCK24, Prop 2.4] for the lemma, but either the citation is wrong or the proposition is being applied outside its hypotheses.\n\nWhat is genuinely good: the inductive framework is novel and attractive. The categories FWSA and ^FWSA are a natural extension of Sam–Snowden, and the argument via τ_{i,g} using Harer's vcd and Deligne purity is a real idea that would explain a linear bound. The algebra sections (3,4) are carefully written. The claimed improvement from O(gi+i^2) to g+5i is the kind of result people want.\n\nThe soft spots beyond Lemma 5.2: Proposition 6.4 has a compressed step identifying the pure part of boundary homology; it may be fixable but needs unpacking. The paper also relies on several cited results without proof, which is normal but becomes a problem when one of them is wrong.\n\nBottom line: this paper should go to peer review because the error is specific and the approach may be salvageable, but as it stands the main theorem is not established. A serious referee should catch this counterexample or an equivalent one. The authors need to either fix the classification of components (perhaps adding the actual lift data to the invariant) or change the construction of the module action.\n\nFor a reading group, it is a useful case study in how a beautiful framework can hinge on a small but false structural lemma.","headline":"The paper's main theorem rests on a false component classification for genus-zero admissible covers; the module action is not well-defined, though the inductive strategy is promising.","tokens_in":22103,"tokens_out":8768,"would_cite":false,"duration_ms":94578,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14H10","14H30","14D23","20C30"],"pacs":[],"model":"deepseek-v4-flash","headline":"For fixed abelian group A, the rational homology of moduli spaces of pointed admissible A-covers is a finitely generated module over a combinatorial category, generated in degree at most g+5i.","keywords":["representation stability","admissible covers","moduli spaces","combinatorial categories","finite generation","rational generating functions","mapping class group","algebraic stacks"],"falsifier":"Exhibit two genus-zero admissible $A$-covers of the same labelled set $(X,\\ell_X)$ whose transition data $b_{x_1,x_2}$ agree for every pair but that lie in different connected components of $M^A_{0,X}(\\ell_X)$; such a pair would invalidate Lemma 5.2 and with it the well-definedness of the $\\widehat{FWS}_A^{op}$ action. A direct check for a small group such as $A=\\mathbb{Z}/2$ with four marked points would settle the foundation.","tokens_in":21034,"feed_emoji":"📐","tokens_out":12245,"duration_ms":102435,"temperature":0.7,"pith_summary":"This paper proves a representation stability result for the rational homology of moduli spaces of pointed admissible $A$-covers, where $A$ is a finite abelian group and the number of marked points grows. The main theorem gives the sequence of homology groups $H_i(M^A_{g,X}(\\ell_X); \\mathbb{Q})$ the structure of a finitely generated module over a combinatorial category of finite sets labelled by $A$, with generators living on sets of size at most $g+5i$. A direct consequence is that the generating function recording the dimensions of these homology groups is rational, with poles constrained to the set $\\{-1, -\\frac12, \\ldots, -\\frac{1}{|A|^2(g+5i)}\\}$. The result also covers unpointed admissible covers, and in the trivial-group case it improves previous stability bounds for the Deligne–Mumford compactification $\\overline{M}_{g,n}$ from quadratic to linear in genus and homology degree. If correct, it shows that the homology of these moduli spaces in each degree is determined by a finite amount of data on small marking sets.","feed_headline":"Admissible-cover homology is fixed by data on at most g+5i points","feed_subtitle":"Their homology is a finitely generated module; rank generating functions are rational with predicted poles.","key_machinery":"The load-bearing device is the combinatorial category $\\widehat{FWS}_A$, a generalization of the category of finite sets and surjections: objects are $A$-labelled finite sets, and a morphism from $(X,\\ell_X)$ to $(Y,\\ell_Y)$ is a surjection $f: X \\to Y$ with the sum-of-labels condition $\\sum_{x\\in f^{-1}(y)} \\ell_X(x) = \\ell_Y(y)$, together with an arbitrary pointing function $g: X \\to A$ that twists the induced map. The homology groups form a functor on the opposite category via gluing maps that attach a canonical genus-zero admissible cover to the fibres of a surjection; the connected component of that genus-zero cover is fixed by transition data (Lemma 5.2), which makes the action well-defined. Finite generation is proved by induction on $(i,g)$ using a surjective gluing map $\\tau_{i,g}$ built from suspension operations and convolution tensor products on the category; the degrees at which $\\tau_{i,g}$ is surjective are controlled by the known bound on the virtual cohomological dimension of the mapping class group, and purity of the weight filtration turns the geometric surjection into a homology surjection. This mechanism converts a geometric stabilization statement into a combinatorial finite-generation statement whose Hilbert-series consequences come from the representation theory of categories of finite sets and surjections.","core_discovery":"The paper's central claim is Theorem B: for fixed homology degree $i$ and genus $g$, the assignment $(X, \\ell_X) \\mapsto H_i(M^A_{g,X}(\\ell_X); \\mathbb{Q})$ is a finitely generated module over the opposite of the combinatorial category $\\widehat{FWS}_A$, generated in degree at most $g+5i$ (and in degree 1 when $(i,g)=(0,0)$). Objects of $\\widehat{FWS}_A$ are finite sets $X$ with a labelling $\\ell_X: X \\to A$; morphisms are compatible surjections with an additional pointing twist, and the action on homology comes from gluing a fixed genus-zero admissible cover onto a given cover. The paper derives Theorem A: the rank generating function $\\sum_n \\dim H_i(M^A_{g,n}; \\mathbb{Q})\\, t^n$ is rational of the form $P(t)/\\prod_{j=1}^{(g+5i)|A|^2} (1+jt)^{d_j}$, and Corollaries C and D give finite generation and rational multivariate generating functions for the unpointed spaces $\\mathrm{Adm}^A_{g,X}(\\ell_X)$. The authors emphasize that the degree bound $g+5i$ is linear, improving the earlier quadratic bound for the moduli space of curves $\\overline{M}_{g,n}$.","pith_inferences":["A direct verification of the cited component classification for genus-zero admissible covers would settle the foundation of the main theorem; the inductive argument itself is otherwise self-contained, so the $g+5i$ bound would survive refinements elsewhere.","The linear bound suggests that the homology of $M^A_{g,X}(\\ell_X)$ can be presented by generators of a topological nature (for instance, decorated stable graphs) numbering at most about $g+5i$; finding such presentations could yield closed formulas for the rank generating functions for small $g$ and $i$.","The construction of $\\widehat{FWS}_A$ is specific to abelian $A$; a stability statement for non-abelian groups $G$ would need a classification of connected components of $M^G_{0,n}$, which the paper identifies as the natural next step. Testing the machinery on cyclic groups, where transition data can be made explicit, would be the cheapest check of the approach's scope.","Because the proof uses purity of the weight filtration, a parallel statement for compactly supported or intersection cohomology, where a different weight property governs the surjection, would test how much of the method is specific to ordinary rational homology."],"forward_implications":["For any fixed $A$, $g$, and $i$, all but finitely many of the vector spaces $H_i(M^A_{g,X}(\\ell_X); \\mathbb{Q})$ are spanned by pullbacks from smaller labelling sets, so the whole sequence is determined by data in degree at most $g+5i$.","The rank generating function of $H_i(M^A_{g,n}; \\mathbb{Q})$ is rational with poles in the set $\\{-1, -\\frac12, \\ldots, -\\frac{1}{|A|^2(g+5i)}\\}$.","The same finite generation, with degree multiplied by $|A|$, holds for the unpointed admissible-cover spaces $\\mathrm{Adm}^A_{g,X}(\\ell_X)$, giving rational multivariate generating functions with explicit denominator structure.","In the trivial-group case $A=1$, the result recovers $\\overline{M}_{g,n}$ and improves the known generation-degree bound for its homology as a module over the category of finite sets and surjections from quadratic to linear in $g$ and $i$.","The module structure also carries representation-theoretic information: after summing over labelings, the homology yields a sequence of $S_n$-representations governed by the theory of finitely generated modules over the category of finite sets and surjections."],"supporting_citations":[{"why":"Defines the moduli space $M^A_{g,X}$ of pointed admissible $A$-covers and its monodromy stratification, the objects whose homology is studied.","marker":"[JKK05]"},{"why":"Supplies Proposition 2.4, cited as Lemma 5.2, classifying connected components of genus-zero admissible covers by transition data; the well-definedness of the module action rests on it.","marker":"[BCK24]"},{"why":"Harer's bound on the virtual cohomological dimension of the mapping class group, used via Borel–Moore duality to give the degree thresholds at which the gluing map is surjective.","marker":"[Har86]"},{"why":"Establishes the Noetherian and Hilbert-series theory of $FWS_A^{op}$-modules, including the Fourier duality used to pass from labelled sets to unlabelled ones.","marker":"[SS19]"},{"why":"Provides the Hilbert-series result for $FS^{op}$-modules (Corollary 8.1.4) that turns finite generation into the rationality statement of Theorem A.","marker":"[SS17]"},{"why":"Operadic perspective on gluing maps for admissible covers, supporting the functorial structure of the $\\widehat{FWS}_A^{op}$-module.","marker":"[Pet13]"},{"why":"Earlier finite-generation result for the homology of Deligne–Mumford compactifications that the present paper improves from a quadratic to a linear degree bound.","marker":"[Tos21]"},{"why":"Gives the effective Hilbert-series bound (Theorem 2.1) used to obtain the explicit denominator form in Theorem 3.19 and Corollary D.","marker":"[Tos24]"}],"fun_headline_variants":["Admissible-cover homology fixed by at most g+5i points","Linear bound for finite generation of admissible-cover homology","Rational generating functions for ranks of admissible-cover homology","Admissible-cover homology: finitely generated with degree ≤ g+5i","Stable homology for admissible covers with linear point bound"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole construction depends on the claim, cited from a companion paper rather than proved here, that a genus-zero admissible cover is determined up to connected component by the pairwise transition data among its marked points; if that classification were wrong, the gluing maps defining the module action would not be well-defined.","fun_headline_variants_meta":{"raw":{"variants":["Admissible-cover homology fixed by at most g+5i points","Linear bound for finite generation of admissible-cover homology","Rational generating functions for ranks of admissible-cover homology","Admissible-cover homology: finitely generated with degree ≤ g+5i","Stable homology for admissible covers with linear point bound"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1778,"prompt_tokens":1019,"completion_tokens":759,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":673}},"tokens_in":635,"tokens_out":759,"duration_ms":7161,"temperature":1.0,"reasoning_tokens":673,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:01:05.806909+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit two genus-zero admissible $A$-covers of the same labelled set $(X,\\ell_X)$ whose transition data $b_{x_1,x_2}$ agree for every pair but that lie in different connected components of $M^A_{0,X}(\\ell_X)$; such a pair would invalidate Lemma 5.2 and with it the well-definedness of the $\\widehat{FWS}_A^{op}$ action. A direct check for a small group such as $A=\\mathbb{Z}/2$ with four marked points would settle the foundation.","supporting_citations":[],"review_version":1}