{"id":"dfd2c47a-d624-4c3e-96e3-528179d72479","arxiv_id":"2607.06345","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Kähler quotients of compact Hamiltonian torus manifolds undergo explicit bimeromorphic transformations across walls, enabling comparison of Kähler classes and computation of Riemann–Roch numbers of singular quotients.","lead":"The paper shows how Kähler quotients of torus actions transform when the moment map value crosses a wall, extending VGIT to non-projective Kähler manifolds. It matters for symplectic and algebraic geometers studying singular reductions and quantization.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"Proposition 4.3's dimension claim (dim_C G_b = 1) is the load-bearing step; its proof via Atiyah's theorem requires careful scrutiny of the dim Φ(G·b) = dim(T·b) identification.","rationale":"The reader correctly identified Proposition 4.3 as a critical step and the triviality of the generic infinitesimal isotropic subgroup as a key assumption. However, the reader's framing of the concern is slightly off: the assumption about trivial generic isotropy is standard and well-justified (it ensures the quotient has expected dimension). The more precise load-bearing concern is the logical step inside Proposition 4.3's proof: going from 'int F ⊂ Φ(G·b)' to 'dim Φ(G·b) = d−1'. This step uses Atiyah's theorem dim Φ(G·b) = dim(T·b), but the dimension conclusion about Φ(G·b) requires that int F is not merely contained in Φ(G·b) but determines its dimension. In practice, this is very likely correct: for a wall point b with dim T_b > 0, the orbit closure image Φ(G·b) is a face of Φ(G·b') for a generic point b' in the same orbit, and by the convexity structure, this face should have dimension exactly d−1 when b lies on a codimension-1 wall. The paper's argument is essentially sound but compresses this step. The overall structure of the paper is coherent: Theorem 3.6 (proper modification) follows cleanly from standard GIT-type arguments, Theorem 4.7 builds on Proposition 4.3 in a natural way, and the applications (Theorem 7.4 on rational desingularization, Theorem 7.7 on Riemann-Roch) follow logically. The gradient flow argument in Lemma 6.3 Step 2 is a standard technique and the computation checks out. The sheaf-theoretic arguments in Lemma 7.8 and Lemma 7.10 use appropriate machinery (Grothendieck spectral sequences, Stein space techniques). The self-citation [67] for cohomologically symplectic structures is grounded in [35]. The conditionality of the reader's verdict is reasonable given the complexity, but I see no issue that would warrant changing the verdict. The concern I identify is a gap in rigor rather than a likely error, and it could be closed by a more detailed argument using the face structure of moment polytopes for torus actions.","tokens_in":48445,"tokens_out":1171,"duration_ms":946845,"concrete_test":"Construct an explicit example with T = (S^1)^2 acting on a compact Kähler manifold X of dimension ≥ 4, where the moment body has a 1-dimensional wall F separating two 2-dimensional subpolytopes. Find a point b ∈ Φ^{-1}(ϵ) with dim T_b > 0 on the wall and verify directly whether dim T_b = 1 or dim T_b ≥ 2. If dim T_b ≥ 2 occurs, Proposition 4.3 fails and Theorem 4.7's fiber description needs revision. A concrete test case: take X = CP^3 with a suitable (S^1)^2-action and compute the stabilizer dimensions at wall points explicitly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Theorem 4.7) depends critically on Proposition 4.3, which asserts that for b ∈ Φ^{-1}(ϵ) with dim T_b > 0 and [G·b] ∈ B, one has dim_C G_b = 1 (equivalently dim T_b = 1). The proof argues: (1) ϵ ∈ Φ(G·b) and by Lemma 3.9, int F ⊂ Φ(G·b), so dim Φ(G·b) = dim F = d−1; (2) by Atiyah [5, Theorem 2(c)], dim Φ(G·b) = dim(T·b), giving dim T_b = 1. The concern is whether step (1) is fully justified. Lemma 3.9 shows X^{ss}(Φ_ϵ) = X^{ss}(Φ_λ) for λ ∈ int F, which means b ∈ X^{sss}(Φ_λ) for all λ ∈ int F, i.e., int F ⊂ Φ(G·b). But does int F being an open subset of Φ(G·b) guarantee that dim Φ(G·b) = d−1? This requires that int F (which is (d−1)-dimensional as a relatively open subset of the wall) is actually open in the affine span of Φ(G·b). If Φ(G·b) had dimension ≥ d, then int F could be a subset of its boundary rather than its interior, and the dimension conclusion would fail. The argument implicitly assumes that the dimension of Φ(G·b) equals the dimension of the smallest affine subspace containing int F, which is d−1. This is plausible but the logical step from 'int F ⊂ Φ(G·b)' to 'dim Φ(G·b) = d−1' is not spelled out rigorously. If dim Φ(G·b) could be larger than d−1, then dim T_b > 1, the identity component G_b^0 would have dimension > 1, and the fiber description in Theorem 4.7 (quotient of weighted projective space by finite group) would break down, as the quotient would involve a higher-dimensional torus rather than C*.","agreement_with_reader":"partial"},"referee_report":{"model":"glm-5.2","summary":"This paper studies how Kähler quotients of Hamiltonian torus actions on compact Kähler manifolds vary as the moment map value crosses walls in the moment polytope. The main structural result (Theorem 4.7) describes the fibers of natural proper modifications between regular and singular quotients as quotients of weighted projective spaces by finite groups, with a codimension formula relating the dimensions of the two fibers. This feeds into a Kähler class comparison theorem (Theorem 6.4, a broken-line-segment result extending Duistermaat–Heckman to critical values) and applications to desingularization, algebraic dimension invariance, and Riemann–Roch numbers. The proofs proceed through standard tools: the Hilbert–Mumford numerical function, Atiyah's convexity theorem, the Holomorphic Slice Theorem, Carrell–Sommese decompositions, and Leray spectral sequences.","tokens_in":48665,"tokens_out":3792,"duration_ms":200049,"significance":"The paper provides a detailed analytic treatment of wall-crossing for Kähler quotients that complements the algebro-geometric VGIT framework of Dolgachev–Hu and Thaddeus. A notable strength is that the authors fix the Kähler form and vary only the moment map value, which yields a more transparent parameter space and removes the need for the 'truly faithful cell' condition in the torus case (Proposition 4.3, Remark 4.4). The Kähler class comparison (Theorem 6.4) and the Riemann–Roch invariance (Corollary 7.6, Theorem 7.7) are concrete, falsifiable results. The proof of Lemma 6.3 contains a detailed local computation relating Kähler potentials across the wall via gradient flow, which is a useful technical contribution. Theorem 5.1 provides an explicit example of the Strong Factorization Conjecture for bimeromorphic maps of compact Kähler manifolds under quasi-free assumptions.","major_comments":[{"comment":"Proposition 4.3 (p. 18–19): The proof's logical step from 'int F ⊂ Φ(G·b)' to 'dim Φ(G·b) = d−1' is not fully spelled out. The inclusion int F ⊂ Φ(G·b) only gives dim Φ(G·b) ≥ d−1. To conclude equality, one needs the additional observation that the hypothesis dim T_b > 0 implies dim(T·b) ≤ d−1, and then by Atiyah [5, Theorem 2(c)] (dim Φ(G·b) = dim(T·b)) one gets dim Φ(G·b) ≤ d−1. Combining the two inequalities yields dim Φ(G·b) = d−1. This is a small but load-bearing gap in the presentation, since Proposition 4.3 underpins the fiber description in Theorem 4.7. The authors should add one sentence making this argument explicit.","section":null}],"minor_comments":[{"comment":"Section 2, p. 6: The assumption that the generic infinitesimal isotropic subgroup is trivial (regular values of Φ are non-empty) is stated but its necessity for the main results is not discussed. A brief remark on where this assumption enters would help the reader.","section":null},{"comment":"Lemma 3.8, Case 1 (p. 13–14): The argument that P ∩ Φ(G·x) ≠ ∅ leading to P ⊂ Φ(G·x) uses the uniqueness of the local minimum of Φ_v. The logic is correct but dense; a sentence clarifying that the contradiction arises because P ⊂ H⁻ while Φ_v ≥ ⟨η, v⟩ forces all of X into H⁺ ∪ Π_F would improve readability.","section":null},{"comment":"Theorem 4.2, Eq. (4.1) (p. 17): The codimension formula uses codim_C(B_i, X^{ss}_ϵ // C*). It would help to note that B_i here is a connected component of X^{C*} ∩ Φ⁻¹(ϵ) viewed inside the quotient, consistent with the notation in Theorem 4.7.","section":null},{"comment":"p. 21, line after Eq. (4.5): The notation 'f_{ξ,ϵ} (resp. f_{ξ,ϵ})' in the sentence beginning 'We consider the fibers of f_{ξ,ϵ} (resp. f_{ξ,ϵ})' appears to be a typo; the second should be f_{ζ,ϵ}.","section":null},{"comment":"Lemma 6.2 (p. 27–28): The proof uses the vanishing ČH¹(ℙ(a), R) = 0 via Dolgachev [13, Corollary 2.3.6]. It would be helpful to note that this is the standard vanishing H¹(ℙ(a), R) = 0 for weighted projective spaces, which follows from their topology (they are simply connected for the relevant cases).","section":null},{"comment":"Theorem 6.4 (p. 32–33): The statement says the curve γ(a) is 'a broken line segment at the point γ(ϵ)'. It would be clearer to state explicitly that γ is continuous and piecewise affine with a potential change of slope at a = ϵ.","section":null},{"comment":"Section 7.3, p. 38–43: The proof of Theorem 7.7 is lengthy. Lemma 7.8 in particular could benefit from a more streamlined presentation; the key ideas (extension via Riemann theorem, G-invariance by averaging) could be highlighted more prominently.","section":null},{"comment":"References: The paper cites [67] (Yang, 'Cohomologically symplectic structures on stratified spaces') with a DOI link but no volume/page numbers. The bibliographic details should be completed if available.","section":null},{"comment":"Several minor typographical issues throughout: e.g., 'truly faithful cell defined in [14]' (p. 5) should perhaps be 'the truly faithful cell condition'; 'tours' (p. 37) should be 'torus'; 'desigularization' (p. 38) should be 'desingularization'.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The reader's stress-test concern about Proposition 4.3 is legitimate as a presentation gap but does not indicate a mathematical error. The argument can be completed in two lines using the hypothesis dim T_b > 0 together with Atiyah's theorem, as noted in the major comment. The paper is a solid contribution that fits the journal's scope in complex algebraic/differential geometry. The reliance on the authors' own prior work [65, 67] for the cohomologically symplectic structure and Kähler class construction is standard in this area and does not raise circularity concerns, as those results are themselves established via independent arguments."},"author_rebuttal":{"model":"glm-5.2","summary":"We thank the referee for a careful reading and for identifying a small but important gap in the proof of Proposition 4.3. The referee's observation is correct: the inclusion int F ⊂ Φ(G·b) alone yields only dim Φ(G·b) ≥ d−1, and the reverse inequality requires the additional step through dim(T·b) ≤ d−1 and Atiyah's theorem. We will revise the proof accordingly.","responses":[{"response":"The referee is entirely correct. The current proof of Proposition 4.3 establishes that int F ⊂ Φ(G·b), which gives dim Φ(G·b) ≥ d−1, but does not explicitly justify the reverse inequality. As the referee notes, the missing step is: since dim T_b > 0 by hypothesis, the orbit T·b has dimension at most d−1 (because T has dimension d and the stabilizer T_b is positive-dimensional). By Atiyah's theorem [5, Theorem 2(c)], dim Φ(G·b) = dim(T·b), so dim Φ(G·b) ≤ d−1. Combining the two inequalities yields dim Φ(G·b) = d−1, and then dim T_b = 1 follows from dim Φ(G·b) = dim(T·b) = d−1. We will add a sentence to the proof making this chain of inequalities explicit. We are grateful for this careful observation.","revision_made":"yes","referee_comment":"Proposition 4.3 (p. 18–19): The proof's logical step from 'int F ⊂ Φ(G·b)' to 'dim Φ(G·b) = d−1' is not fully spelled out. The inclusion int F ⊂ Φ(G·b) only gives dim Φ(G·b) ≥ d−1. To conclude equality, one needs the additional observation that the hypothesis dim T_b > 0 implies dim(T·b) ≤ d−1, and then by Atiyah [5, Theorem 2(c)] (dim Φ(G·b) = dim(T·b)) one gets dim Φ(G·b) ≤ d−1. Combining the two inequalities yields dim Φ(G·b) = d−1. This is a small but load-bearing gap in the presentation, since Proposition 4.3 underpins the fiber description in Theorem 4.7. The authors should add one sentence making this argument explicit."}],"tokens_in":48058,"tokens_out":527,"duration_ms":100720,"standing_objections":[]},"desk_editor":{"model":"glm-5.2","letter":"The main thing to know: this paper extends VGIT wall-crossing for torus actions from projective varieties to general compact Kähler manifolds, and the key technical innovation — fixing the Kähler form while varying only the moment map value — genuinely removes the 'truly faithful cell' condition from Dolgachev–Hu. That removal is real and is the paper's main contribution. Theorem 4.7 (fiber description as quotients of weighted projective spaces by finite groups, plus the codimension formula) is the load-bearing structural result, and the applications to desingularization (Theorem 7.4) and Riemann–Roch invariance (Theorem 7.7, Corollary 7.6) follow from it in a natural way. The Čech-cohomological broken-line-segment description of Kähler class variation (Theorem 6.4) is also new and useful for the quantization program. The proofs are detailed and proceed through standard tools (Hilbert–Mumford, Holomorphic Slice, Leray spectral sequences, Carrell–Sommese decompositions) in a competent way. The gradient flow argument in Lemma 6.3 Step 2 is a nice piece of analysis. The self-citation to [67] for the cohomologically symplectic structure is grounded in Heinzner–Stratmann [35], so it is not circular. The stress-test concern about Proposition 4.3 — specifically whether 'int F ⊂ Φ(G·b)' forces dim Φ(G·b) = d−1 — is legitimate as a gap in exposition but not as a mathematical error. Here is why: Atiyah's convexity theorem gives dim Φ(G·b) = dim(T·b), and since b lies on the wall with dim T_b > 0, the orbit image Φ(G·b) is a proper face of the full moment polytope. The wall F is (d−1)-dimensional, and int F sits inside Φ(G·b) as a relatively open subset. Because Φ(G·b) is a convex polytope in its own affine span, containing a (d−1)-dimensional relatively open set forces that affine span to be (d−1)-dimensional. So the conclusion dim T_b = 1 follows. The logic is sound; the paper just doesn't spell out the intermediate step about affine spans explicitly. A referee should ask for one or two clarifying sentences here. The assumption that the generic infinitesimal isotropic subgroup is trivial (regular values non-empty) is standard and clearly stated; it is not a soft spot. The sheaf-theoretic work in Section 7 (Lemma 7.8 on the isomorphism L_0 ≅ (f_{β,0})_* L_β, and Lemma 7.10 on vanishing higher direct images) is intricate and worth checking line by line, but the strategy is sound and the use of Riemann extension plus the averaging argument for T-invariance is correct. This paper is for symplectic and complex geometers working on reduction, wall-crossing, and quantization. It deserves a serious referee who can verify the analytic details in Sections 6–7.","headline":"Genuine extension of VGIT wall-crossing to the Kähler analytic category; the load-bearing dimension claim in Proposition 4.3 holds up but deserves a careful referee check.","tokens_in":49550,"tokens_out":763,"would_cite":true,"duration_ms":97272,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"glm-5.2","headline":"Wall-crossing gives bimeromorphic maps between Kähler quotients","keywords":[],"falsifier":"If one could exhibit a compact Kähler Hamiltonian torus-manifold where the regular value set of the moment map is non-empty but some wall point has a complex stabilizer of dimension greater than one, the fiber description in Theorem 4.7 would fail and the codimension formula would not hold.","tokens_in":48423,"feed_emoji":"🔺","tokens_out":920,"duration_ms":123634,"temperature":0.7,"pith_summary":"The paper studies what happens to Kähler quotients of torus actions when the moment map value crosses a wall in the moment polytope. The central result (Theorem 4.7) shows that when two subpolytopes are separated by a wall, there exist natural proper modifications between the corresponding Kähler quotients and the singular quotient on the wall, sharing the same center, with fibers that are quotients of weighted projective spaces by finite groups, and whose dimensions satisfy a codimension formula. This structural theorem enables a comparison of Kähler classes across walls (Theorem 6.4), showing the Kähler class traces a broken line segment in Čech cohomology. As applications, the authors prove that singular nondegenerate Kähler quotients admit partial rational desingularizations obtained by shifting the moment map value, that all nondegenerate quotients share the same algebraic dimension and Riemann–Roch numbers, and that the quantization commutes with reduction principle extends to singular quotients in the integral case.","feed_headline":"Shifting the moment map resolves singular Kähler quotients","feed_subtitle":"Wall-crossing between subpolytopes yields bimeromorphic maps with explicit fibers, enabling class comparison and Riemann–Roch invariance.","key_machinery":"The Hilbert–Mumford numerical function M_Φ, the plus/minus decomposition of Carrell–Sommese for C*-actions, Atiyah's convexity theorem for moment map images of orbit closures, the Stein factorization and Zariski's Main Theorem for proper modifications, and the Leray spectral sequence for cohomological comparison","core_discovery":"The key discovery is that wall-crossing for Kähler quotients of torus actions is governed by a pair of proper modifications with a common center, whose fibers are explicitly describable as finite quotients of weighted projective spaces. The critical intermediate result (Proposition 4.3) shows that for a point on the wall, the complex stabilizer has dimension one, which forces the fiber structure and enables the codimension formula. This explicit fiber description, combined with the Duistermaat–Heckman theorem extended to the wall, yields the broken-line behavior of Kähler classes and the invariance of Riemann–Roch numbers under shift desingularization.","pith_inferences":[],"forward_implications":["The broken-line formula for Kähler classes provides a concrete computational tool for tracking how the cohomology class of the reduced symplectic form changes when passing through singular quotients, extending the Duistermaat–Heckman theorem to the Kähler analytic setting.","The partial rational desingularization via moment map shifting offers a canonical procedure for resolving non-orbifold singularities of Kähler quotients, which is directly relevant to extending geometric quantization results to singular reduced spaces.","The invariance of Riemann–Roch numbers across all nondegenerate quotients means that quantization commutes with reduction can be verified on any convenient representative quotient, including smooth ones obtained by shifting.","The explicit factorization into blow-ups and blow-downs for quasi-free actions with finite center provides a concrete class of bimeromorphic maps between compact Kähler manifolds satisfying the Strong Factorization Conjecture.","The constancy of algebraic dimension across nondegenerate quotients suggests that bimeromorphic geometry of the quotient is essentially determined by the moment body interior, independent of the specific level chosen."],"fun_headline_variants":["Wall-crossing for Kähler quotients yields bimeromorphic maps via proper modifications","Singular Kähler quotients desingularize by shifting the moment map across walls","Proposition 4.3: one-dimensional stabilizers force fiber structure at wall crossings","Kähler classes follow broken-line behavior across subpolytope walls","Riemann–Roch numbers invariant under shift desingularization of singular quotients"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The paper assumes throughout that the generic infinitesimal isotropic subgroup of the torus action is trivial, meaning the regular value set of the moment map is non-empty. If the moment map had no regular values, the subpolytope decomposition and all subsequent structural results would break down. Additionally, the proof of Proposition 4.3—that the complex stabilizer of a wall point has dimension exactly one—relies on the specific structure of torus actions and Atiyah's Conv","fun_headline_variants_meta":{"raw":{"variants":["Wall-crossing for Kähler quotients yields bimeromorphic maps via proper modifications","Singular Kähler quotients desingularize by shifting the moment map across walls","Proposition 4.3: one-dimensional stabilizers force fiber structure at wall crossings","Kähler classes follow broken-line behavior across subpolytope walls","Riemann–Roch numbers invariant under shift desingularization of singular quotients"]},"model":"glm-5.2","effort":"high","cost_usd":0.0,"raw_usage":{"total_tokens":733,"prompt_tokens":642,"completion_tokens":91,"prompt_tokens_details":null},"tokens_in":642,"tokens_out":91,"duration_ms":38084,"temperature":1.0,"reasoning_tokens":null,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-08T08:56:09.475114+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"If one could exhibit a compact Kähler Hamiltonian torus-manifold where the regular value set of the moment map is non-empty but some wall point has a complex stabilizer of dimension greater than one, the fiber description in Theorem 4.7 would fail and the codimension formula would not hold.","supporting_citations":[],"review_version":1}