{"id":"665ca1fd-fd8a-43fc-86e4-1a31aded372e","arxiv_id":"1908.04213","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For any quasi-symmetric representation of a reductive group, the derived categories of GIT quotients form a perverse schober on the partial compactification of the stringy Kähler moduli space.","lead":"The paper builds a perverse schober, a modern categorical object that tracks how derived categories of sheaves change as a geometric quotient varies, for a broad class of group actions. It extends an earlier local system of categories to the full compactified moduli space, giving a complete wall-crossing structure.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing comparison with the [HLS16a] local system: Prop 5.1 proves an H-schober, but not that it extends the stated local system.","rationale":"The reader's weakest_assumption concerns Theorem 5.6 (finite global dimension of Λ_C) and its role in Lemma 5.7. That assumption is not the most load-bearing part of the central claim: the H-schober axioms in Definition 3.4 do not require the right adjoints γ to preserve compact objects, and the main H-schober structure on the categories E_C is obtained before passing to E_C^c. A failure of finite global dimension would affect part (2) of Proposition 5.1, but not necessarily the existence of an H-schober on E_C or the validity of Theorem 1.2. The genuinely unproved step relevant to Theorem 1.2 is the identification of the restriction of the constructed H-schober with the [HLS16a] local system. Proposition 5.1 is entirely about the H-schober axioms; the text does not verify that on the open stratum the functors φ_{C1C2} agree with the equivalences of [HLS16a]. Since an extension of a local system requires compatibility with the given monodromy, this comparison is essential. The paper otherwise gives a detailed and plausible proof of the H-schober structure, including a self-contained appendix for the main combinatorial decomposition, so the appropriate remedy is to request an explicit comparison rather than to reject the work.","tokens_in":17025,"tokens_out":36811,"duration_ms":358592,"concrete_test":"For adjacent maximal cells C1, C2 sharing a facet, let Φ: D(X^{ss,χ1}/G) → D(X^{ss,χ2}/G) be the wall-crossing equivalence defined in [HLS16a]. Using the tilting equivalences E_{C_i} ≅ D(X^{ss,χ_i}/G) from Remark 5.2, prove that the functor φ_{C1C2}: E_{C1} → E_{C2} is naturally isomorphic to the transport of Φ along these tilting equivalences, for every wall. If this comparison cannot be established, Theorem 1.2 is not proven by Proposition 5.1.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Theorem 1.2 asserts that the constructed H-schober extends the local system of triangulated categories on the SKMS established in [HLS16a]. Proposition 5.1, which is cited as the precise form of Theorem 1.2, proves that the data (E_C, δ, γ, φ_χ) is an X(T)^W-equivariant H-schober, but it never states or proves that the restriction of this H-schober to the complement of the hyperplane arrangement is equivalent, as a local system of triangulated categories, to the [HLS16a] local system. Remark 5.2 identifies the fibers for maximal cells via the tilting equivalence E_C ≅ D(X^{ss,χ}/G), but the wall-crossing functors φ_{C1C2} between adjacent maximal cells are not compared with the equivalences or monodromy defined in [HLS16a]. Without this comparison, the construction could yield a different local system on the same SKMS, in which case Theorem 1.2 would not follow from Proposition 5.1. This is an omitted proof of a stated claim, not an internal inconsistency.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs an X(T)^W-equivariant H-schober on the real affine space X(T)^W_R stratified by the hyperplane arrangement H associated by Halpern-Leistner and Sam to a quasi-symmetric representation of a reductive group. The fibers E_C are subcategories of D(X/G) generated by tilting objects P_C, with restriction functors δ and right adjoints γ, together with a tensor action of X(T)^W. Proposition 5.1 states that this data satisfies the axioms of an H-schober, that collinear faces give mutation spherical pairs, and that the same holds for compact objects. Appendix A gives a self-contained proof of the required semi-orthogonal decompositions. The paper claims in Theorem 1.2 that this extends the local system of triangulated categories on the SKMS established in [HLS16a].","tokens_in":17228,"tokens_out":3993,"duration_ms":43218,"significance":"If the extension claim is fully justified, this is a substantial contribution: it produces perverse schobers in a broad GIT setting, generalizing Donovan's punctured-disk examples and giving evidence for the Bondal–Kapranov–Schechtman program. The paper's strengths include an explicit and detailed construction, a self-contained combinatorial appendix (Appendix A) proving the needed semi-orthogonal decompositions, and explicit identification of mutation spherical pairs. The main weakness is that the relation between the constructed H-schober and the [HLS16a] local system is asserted but not proved, which leaves the central theorem's content partly open.","major_comments":[{"comment":"Theorem 1.2 claims that the local system from [HLS16a] extends to a perverse schober. Proposition 5.1, cited as the precise form of this theorem, proves that the data (E_C, δ, γ, φ_χ) is an X(T)^W-equivariant H-schober. However, the proposition never compares the restriction of this H-schober to the open stratum (X(T)_C^W \\ H_C)/X(T)^W with the local system constructed in [HLS16a]. In particular, for adjacent maximal cells C1 and C2, the functor φ_{C1C2} is shown in §5.5 to be a mutation functor, whereas the wall-crossing functors in [HLS16a] are defined via window shifts; no identification is supplied. Since the abstract and introduction make the extension of that specific local system a central claim, this is an omitted proof of a load-bearing statement. I ask the authors to add a direct comparison on the generating objects P_χ, or to state and verify a uniqueness/rigidity result for H-schobers extending a given local system on the open stratum.","section":"§5.1, Theorem 1.2 vs. Proposition 5.1"},{"comment":"The proof that γ_{C'C} preserves compact objects, which is needed for Proposition 5.1(2), relies on Theorem 5.6, quoted from [ŠVdB17, Theorem 1.6.1], that Λ_C has finite global dimension. This theorem is not reproved in the paper. The reliance is acceptable if the cited result is accepted, but the dependence should be stated explicitly; if finite global dimension failed, the right adjoint would not restrict to compact objects and the H-schober structure on E_C^c would collapse. Please indicate precisely where in [ŠVdB17] this is established and flag Proposition 5.1(2) as conditional on it.","section":"§5.3, Lemma 5.7 and Theorem 5.6"}],"minor_comments":[{"comment":"The symbol W is used both for the Weyl group (\"Let T ⊂ B ⊂ G ... Weyl group W\") and for the G-representation (\"Below W will be a finite dimensional G-representation\"). This collision is confusing in a paper about quasi-symmetric representations; please rename one of the two objects.","section":"§2"},{"comment":"The sentence ending \"forms an H-schober 2\" appears to contain a stray superscript '2' after \"schober\"; please remove it.","section":"§3.4, Example 3.11"},{"comment":"The reduction to the neighboring case is described as \"by considering those we reduce formally\"; a few more details or a pointer to a precise configuration of faces would help the reader follow the proof.","section":"§5.6, proof of property (T)"},{"comment":"The paper uses the notion of H-schober \"somewhat loosely\" and explains in Remark 3.5 that an equivariant H-schober is to be viewed as a perverse schober on the quotient stack. Since Theorem 1.2 states the result in terms of \"perverse schober\", it would be helpful to state explicitly in the theorem which version of perverse schober is meant and how the passage from X(T)^W-equivariant H-schober to perverse schober on X(T)_C^W/X(T)^W is made.","section":"§3.4 and §5.1"}],"recommendation":"major_revision","confidential_remarks":"The missing comparison with the [HLS16a] local system is the sole substantive obstacle to the central claim; the construction and technical apparatus appear sound, and Appendix A is a valuable self-contained contribution. If the authors add the required identification, I would be happy to support acceptance. The paper is within the scope of the journal and the results are of clear interest to the derived-category and GIT communities."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Solid paper. The construction of H-schobers for all quasi-symmetric representations is a genuine advance over Donovan's special cases, and the main technical claims are proved carefully. Appendix A is a real asset: it makes the needed semi-orthogonal decompositions self-contained, so the paper does not merely point at the authors' earlier work. The extra observation that adjacent chambers form mutation spherical pairs is a strong result, not something that follows formally from the H-schober axioms.\n\nThe soft spot is the one the stress-test flags. Theorem 1.2 asserts that the constructed perverse schober extends the HLS16a local system, but Proposition 5.1 only proves the H-schober axioms. It identifies the fibers for maximal cells in Remark 5.2, but it never compares the wall-crossing functors φ_{C1C2} with the equivalences that HLS16a use to define their local system. So the extension claim is not actually demonstrated in the text. I consider this a fixable gap rather than a fatal flaw: the E_C are exactly the HLS16a windows, and the φ's are the natural mutation functors between them, so a short comparison or a precise citation should close it. But as written, the main theorem is under-proved.\n\nMinor: Lemma 5.7 relies on the cited finite global dimension result from the authors' earlier work; that is acceptable background, not circularity. The deliberate choice of H-schobers over a more restrictive notion of perverse schober is handled by verifying the axioms, which the paper does.\n\nWho should read this: anyone working on derived categories of GIT quotients, noncommutative resolutions, or perverse schobers. I would send it to peer review, and I would ask the authors to make the comparison with the HLS16a local system explicit.","headline":"A genuine advance in constructing perverse schobers for quasi-symmetric GIT quotients, with one fixable gap: the comparison with the HLS16a local system is missing.","tokens_in":17773,"tokens_out":7579,"would_cite":true,"duration_ms":70057,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["13A50","53D37","32S45","16S38","18E30","14F05"],"pacs":[],"model":"deepseek-v4-flash","headline":"For quasi-symmetric representations of reductive groups, the local system of triangulated categories on the stringy Kähler moduli space extends to a perverse schober on its partial compactification.","keywords":["perverse schobers","geometric invariant theory","categorification","stringy Kähler moduli space","quasi-symmetric representations","H-schober","mutation spherical pair","semi-orthogonal decomposition"],"falsifier":"For the one-dimensional torus example described in the introduction (weights summing to zero, not all zero), one can compute the cone of the map $\\gamma_{C_1C_2}P_\\chi \\to P_\\chi$ in (5.5) for a collinear triple of cells; if that cone is not contained in $\\mathcal{E}_{C_1,C_2}$, or if the natural map in (5.4) is not an isomorphism, then axiom (T) fails. A broader check would be to find a quasi-symmetric representation for which $\\Lambda_C$ has infinite global dimension, which would contradict the imported theorem and invalidate the construction.","tokens_in":16819,"feed_emoji":"📐","tokens_out":14313,"duration_ms":126152,"temperature":0.7,"pith_summary":"This paper proves that the stringy Kähler moduli space (SKMS) attached to a quasi-symmetric representation of a reductive group supports a perverse schober: the local system of triangulated categories previously built on the open SKMS extends to the whole partial compactification. A quasi-symmetric representation is one whose weights on every line through the origin sum to zero, and the construction applies to the quotient stack $X/G$ with $X = W^*$. The extension is built cell by cell from a hyperplane arrangement in the character space, with one triangulated category per cell, so wall-crossing in geometric invariant theory is encoded as categorified perverse sheaf data. The paper also shows that at every wall the two adjacent categories form a mutation spherical pair, which is the precise categorical reflection of a wall crossing.","feed_headline":"Perverse schobers cover the stringy Kähler moduli compactification","feed_subtitle":"The triangulated categories attached to quasi-symmetric GIT representations now extend across the discriminant.","key_machinery":"The central mechanism is the H-schober, a categorical analogue of a perverse sheaf on a space stratified by a real hyperplane arrangement. The construction assigns to each cell $C$ the category $\\mathcal{E}_C = \\langle P_\\chi \\mid \\chi \\in L_C\\rangle \\subset D(X/G)$, where $P_\\chi = V(\\chi)\\otimes\\mathcal{O}_X$ and $L_C = (\\xi_C - \\rho + \\tfrac12\\Sigma)\\cap X(T)^+$ for any $\\xi_C \\in C$. The inclusions $\\delta_{CC'}$ and their right adjoints $\\gamma_{C'C}$ are the structure maps, and the $X(T)^W$-action is by twisting with characters. The proof of the schober axioms runs through explicit semi-orthogonal decompositions $\\mathcal{E}_C = \\langle \\mathcal{E}_{C,C_1}, \\mathcal{E}_{C_1}\\rangle = \\langle \\mathcal{E}_{C_1}, \\mathcal{E}_{C,C_2}\\rangle = \\cdots$ obtained from the zonotope $\\Delta = -\\rho + \\tfrac12\\Sigma$, together with the duality $\\mathbf{D}(\\mathcal{E}^c_C) = \\mathcal{E}^c_{-C}$ and the finite global dimension of $\\Lambda_C = \\mathrm{End}_{X/G}(P_C)$, which ensures the adjoints preserve compact objects.","core_discovery":"Theorem 1.2, made precise as Proposition 5.1, states that the data $((\\mathcal{E}_C)_C, (\\gamma_{C'C})_{C'C}, (\\delta_{CC'})_{C'C})$ together with the $X(T)^W$-action given by tensoring with characters defines an $X(T)^W$-equivariant H-schober on $X(T)^W_{\\mathbb{C}}$, hence a perverse schober on the quotient $X(T)^W_{\\mathbb{C}}/X(T)^W$ that extends the SKMS local system. For each cell $C$ of the hyperplane arrangement, $\\mathcal{E}_C$ is the subcategory of $D(X/G)$ generated by objects $P_\\chi = V(\\chi)\\otimes\\mathcal{O}_X$ with $\\chi$ ranging over the finite set $L_C$ of dominant weights determined by the cell. The inclusions $\\delta_{CC'}$ for $C' \\subset C$ admit right adjoints $\\gamma_{C'C} = \\mathbf{R}\\mathrm{Hom}_{X/G}(P_C,-)\\otimes_{\\mathrm{End}_{X/G}(P_C)}P_C$, and the composite wall-crossing functors $\\varphi_{C_1C_2} = \\gamma_{C'C_2}\\delta_{C_1C'}$ are equivalences across facets. In addition, for collinear cells $C_1, C, C_2$ with $C < C_1, C_2$, the pair $(\\mathcal{E}_{C_1}, \\mathcal{E}_{C_2})$ is a mutation spherical pair in $\\mathcal{E}_C$, and the same statement holds for the compact objects $\\mathcal{E}^c_C$.","pith_inferences":["Inference: because the categories and functors are determined by the finite weight sets $L_C$, chamber-to-chamber derived equivalence for a given representation is a finite combinatorial condition that could be checked by explicit computation on examples.","Inference: each mutation spherical pair should yield a spherical functor between the adjacent categories, so one expects monodromy autoequivalences around the corresponding missing divisors; the paper does not develop this monodromy picture.","Inference: the same zonotope method would produce a weaker structure for representations that are not quasi-symmetric, since the balanced condition on weights is used in the combinatorial Lemma A.10; the resulting data would likely be an H-schober with non-equivalence across some walls.","Inference: for torus examples the construction should recover, and organize, the known braid-group actions and window-shift autoequivalences of GIT quotients, providing a single categorical object that packages all wall-crossing equivalences."],"forward_implications":["Wall crossings in the SKMS become mutation functors: across each wall the two adjacent categories form a mutation spherical pair, giving a 4-periodic semi-orthogonal decomposition at the wall.","The equivariance under $X(T)^W$ means the schober descends to a perverse schober on the quotient partial compactification, exactly the extension promised by Theorem 1.2.","In every maximal chamber where the semistable locus is smooth and nonempty, the fiber $\\mathcal{E}_C$ is equivalent to $D(X^{ss,\\chi}/G)$, so all maximal chambers are derived equivalent.","When the representation is generic, $\\Lambda_C$ is a non-commutative crepant resolution of $k[X]^G$, so the family of categories realizes these NCCRs over the compactified base.","The compact-object version $\\mathcal{E}^c_C$ carries the full H-schober structure, so the result is compatible with duality and with the distinction between perfect and arbitrary complexes on stacks."],"supporting_citations":[{"why":"Constructs the SKMS hyperplane arrangement and the local system of triangulated categories on its complement that this paper extends.","marker":"[HLS16a]"},{"why":"Supplies the quasi-symmetric representation framework and the theorem that $\\Lambda_C = \\mathrm{End}_{X/G}(P_C)$ has finite global dimension, the key finiteness input.","marker":"[ŠVdB17]"},{"why":"Introduces the H-schober notion used here, the perverse-schober analogue for hyperplane arrangements.","marker":"[BKS18]"},{"why":"Provides the combinatorial description of perverse sheaves on hyperplane arrangements that H-schobers categorify.","marker":"[KS16]"},{"why":"Supplies the notions of spherical functors and spherical pairs, used to articulate the mutation structure at walls.","marker":"[KS15]"},{"why":"Gives the semi-orthogonal decompositions of GIT quotient stacks used in Appendix A to prove the schober axioms.","marker":"[ŠVdB16]"},{"why":"Treats the special punctured-disk/torus example that the present construction generalizes.","marker":"[Don18]"}],"fun_headline_variants":["Perverse schobers extend across GIT wall-crossing boundaries","Categorifying perverse sheaves on stringy Kähler moduli","Schobers on compactified stringy Kähler moduli from GIT","New perverse schobers for quasi-symmetric GIT quotients","Triangulated categories glue along the GIT discriminant"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the endomorphism ring $\\Lambda_C = \\mathrm{End}_{X/G}(P_C)$ has finite global dimension, a fact the paper imports from earlier work; if it failed, the right adjoint $\\gamma_{C'C}$ would not preserve compact objects and the H-schober structure on the compact categories would collapse.","fun_headline_variants_meta":{"raw":{"variants":["Perverse schobers extend across GIT wall-crossing boundaries","Categorifying perverse sheaves on stringy Kähler moduli","Schobers on compactified stringy Kähler moduli from GIT","New perverse schobers for quasi-symmetric GIT quotients","Triangulated categories glue along the GIT discriminant"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000851,"raw_usage":{"total_tokens":3718,"prompt_tokens":982,"completion_tokens":2736,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":598,"completion_tokens_details":{"reasoning_tokens":2641}},"tokens_in":598,"tokens_out":2736,"duration_ms":20398,"temperature":1.0,"reasoning_tokens":2641,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:47:38.792776+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For the one-dimensional torus example described in the introduction (weights summing to zero, not all zero), one can compute the cone of the map $\\gamma_{C_1C_2}P_\\chi \\to P_\\chi$ in (5.5) for a collinear triple of cells; if that cone is not contained in $\\mathcal{E}_{C_1,C_2}$, or if the natural map in (5.4) is not an isomorphism, then axiom (T) fails. A broader check would be to find a quasi-symmetric representation for which $\\Lambda_C$ has infinite global dimension, which would contradict the imported theorem and invalidate the construction.","supporting_citations":[],"review_version":1}