{"id":"5b17956e-b5f0-4a90-ab1d-268fe03646e0","arxiv_id":"2509.04690","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The equivariant cohomology of the moduli space of relative quasimaps to the flag variety carries a U(gl_n)-action whose specialized summand is, up to a known category equivalence, a tilting module.","lead":"This paper constructs an action of the Lie algebra U(gl_n) on the cohomology of a compactified space of maps from a projective line to the flag variety, and identifies the resulting module with the image of a tilting module under a known category equivalence. The result connects enumerative geometry to tilting theory and gives new formulas for tilting multiplicities.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Dual Verma filtration in §5.2 is only a vector-space filtration; U(gl_n)-stability of the B-B filtration is not proved, so the tilting conclusion is unsupported.","rationale":"The reader's weakest_assumption focused on the smoothness proof (Proposition 2.1) and the implied localization step. That is a legitimate concern, but I find the U(gl_n)-stability of the filtration in Section 5.2 more directly load-bearing for the central claim: even if smoothness, the B-B stratification, and localization are fully repaired, the paper still has not shown that H_λ has a dual Verma filtration as a U(gl_n)-module. A vector-space filtration with dual Verma quotients is insufficient for the conclusion that Υ^{-1}(H_λ) is a tilting module. The same gap reappears in the dualization step, so the claimed Verma filtration is also not established. The paper's fixed-point analysis and weight computations (e.g. Proposition 2.9) are consistent, and the multiplicity calculation in Corollary 5.5 would be reasonable once the module-filtration property is supplied. Thus I do not recommend rejection: the argument is a plausible sketch with a concrete missing verification. The suggested test in the n=2 case would settle whether the B-B filtration is actually a filtration of U(gl_n)-submodules. I mark agreement as 'partial' because the reader's formal weakest_assumption is different, though their rationale explicitly mentions the missing U(gl_n)-compatibility.","tokens_in":14290,"tokens_out":11689,"duration_ms":111947,"concrete_test":"Take the smallest nontrivial case n=2, λ=(1,0), ϵ=1, w=e, and degree d=(1). Compute the U(sl_2)-action on H_λ from the correspondence formulas (9) and (13) on the fixed-point basis of M_0, and compare it with the filtration defined by the B-B open sets U_i in (17). Specifically, check whether the matrix of E (or F) is block-triangular with respect to the subspaces ker(H^*(U_i) → H^*(U_m)). If any matrix element connects a higher cell to a lower one, the filtration is not U-stable; if block-triangularity holds in this example, repeat the check for degree 2 or for w=s_1. Alternatively, verify analytically that the correspondence-induced operator E_i sends the class of an attracting cell ζ_{λ,e_f} to a combination of classes with the same e_f-coordinate and that the partial order in §5.2 is compatible with the U-action; a single counterexample suffices to disprove the claimed filtration.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.2's central inference is that H_λ has a dual Verma filtration. The construction through the open subsets U_i in (17) yields a filtration of the underlying vector space H^*(M_0): each successive quotient H^*(U_{i+1})/H^*(U_i) is identified with the cohomology of an attracting cell union, which is isomorphic to that of QM_ns and hence to a dual Verma module. But nothing in the text shows that the subspaces ker(H^*(U_i) → H^*(U_m)) are U(gl_n)-submodules. The operators E_i,F_i are defined by correspondences C^d_i on QM_rel (Section 3.2), and the proof of Theorem 3.1 only records their effect on fixed-point classes; it does not compare this action with the restriction maps defining the filtration. A filtration by dual Verma modules in category O' requires each filtered piece to be stable under the U(gl_n)-action, with successive quotients isomorphic as U(gl_n)-modules to dual Vermas. The quoted sentence 'By construction, ker(U_{i+1} → U_i) is isomorphic to a dual Verma module' supplies at most an isomorphism of graded/weight spaces after specialization. The dualization step has the same gap: Poincare duality is compatible with the U-action, but dualizing a filtration that is not known to be a module filtration does not produce a Verma filtration by submodules. Since both Verma and dual Verma filtrations are needed to apply Soergel's equivalence and conclude that Υ^{-1}(H_λ) is tilting, this is a load-bearing missing verification.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the equivariant cohomology of the moduli space of relative quasimaps from P^1 with one marked point to the flag variety, which compactifies the Laumon space QM_ns. The author extends the known U(gl_n)-action on H_T^*(QM_ns) to the whole relative quasimap space QM_rel via geometric correspondences, proves smoothness and a Bialynicki-Birula stratification of QM_rel, and then specializes equivariant parameters to a regular lowest weight. The main claim is that a certain direct summand H_{λ,w} of the specialized cohomology, obtained from the connected components of the fixed locus that meet QM_ns, carries both a Verma and a dual Verma filtration after passing to Soergel's category O', so that its inverse image under Soergel's equivalence is a tilting module of U(gl_n). The paper further computes the multiplicities of indecomposable tilting summands in terms of Bruhat paths and Kazhdan-Lusztig polynomials (Corollary 5.5).","tokens_in":14571,"tokens_out":6694,"duration_ms":65313,"significance":"If the main theorem is correct, the paper provides a geometric realization of tilting modules for U(gl_n) via the cohomology of relative quasimap spaces, extending the known realization of dual Verma modules by Laumon spaces. The explicit multiplicity formula in Corollary 5.5 is concrete and checkable, and the use of Soergel's categorical equivalence is a natural bridge between geometric representation theory and tilting theory. The paper also develops useful geometric foundations for QM_rel, including smoothness, a global quotient presentation, and a B-B stratification, which are of independent interest. However, the central proof rests on several assertions that are not fully justified, so the paper is not yet convincing as written.","major_comments":[{"comment":"The dual Verma filtration is not shown to be a filtration by U(gl_n)-submodules. The sequence of surjections H^*(U_{i+1})→H^*(U_i) comes from open inclusions of strata, and the assertion that ker(U_{i+1}→U_i) is 'isomorphic to a dual Verma module' identifies at most the underlying vector space (or weight-space graded object) with the cohomology of an attracting cell. No proof is given that the restriction maps H^*(U_{i+1})→H^*(U_i) commute with the correspondences E_i,F_i defined in Section 3.2, nor that each kernel is stable under the U(gl_n)-action. Since a tilting module is required to have Verma and dual Verma filtrations by submodules, this missing equivariance is load-bearing. The same gap affects the subsequent dualization step: even if the pairing (α,β)↦ϖ_*(α∪β) is compatible with the U-action, dualizing a filtration that is not known to be a module filtration does not produce a Verma filtration by submodules. The author should either prove the equivariance of the filtration or revise the claim.","section":"§5.2, Eq. (17)"},{"comment":"The proof of Theorem 3.1 is too sketchy to establish the U(gl_n)-action on H_T^*(QM_rel). The argument only records the effect of E_i and F_i on localized fixed-point classes, asserting that the coefficients coincide with those for QM_ns, and that the H_i weights match. It does not verify the Serre relations, the commutator [E_i,F_i], or the compatibility of the correspondence action with the restriction maps between the strata U_i used in Section 5.2. These verifications are essential because the tilting conclusion ultimately depends on the module structure, not just on weight-space dimensions. A complete proof, or a precise reduction to the Laumon-space case plus a check on all new fixed-point components, is needed.","section":"§3.2, Theorem 3.1"},{"comment":"Lemma 4.2 asserts without proof that the fixed locus (QM_rel^d)^{C^*_λ} is smooth and irreducible. This statement is used to identify I_0 as the closure of the intersection with QM_ns and hence to conclude that H_λ is a well-defined direct summand (and a submodule). The footnote about finite groups does not supply an argument for the C^* case. The smoothness and irreducibility claims need either a proof or a precise reference, since the decomposition into H_λ and H'_λ is foundational for everything that follows.","section":"§4, Lemma 4.2"},{"comment":"The proof that the center acts by constants on H_λ is incomplete. For a class supported on a fixed component (λ,eP), the action of a central element z may a priori involve contributions from other fixed components; the sentence 'the action of z only depends on λ' is asserted rather than proved. Moreover, the definition of O' requires H_λ to be locally finite for the subalgebra spanned by H_i and F_i, but this property is not verified for H_λ. The membership of H_λ in O' is necessary for applying Soergel's equivalence, so these points should be addressed explicitly.","section":"§5.1, Lemma 5.1"}],"minor_comments":[{"comment":"The heuristic 'if we fix the map on the bubbles but let the map on the parametrized P^1 vary' is too vague to convey the actual construction; it would help to state the filtration explicitly in terms of the open subsets U_i already in the introduction.","section":"§1.4"},{"comment":"In the proof, 'there is a universal curve C_N living over C^n' should presumably be 'over C^N'; the notation is inconsistent with the subsequent use of I⊂{1,...,N}.","section":"§2.3, Proposition 2.2"},{"comment":"The display '0↞H^*(U_1)↞H^*(U_2)↞...' uses the arrow direction in a nonstandard way; please clarify which maps are restrictions and which are pushforwards, and state explicitly that the meaning is a sequence of surjections H^*(U_{i+1})→H^*(U_i).","section":"§5.2, Eq. (17)"},{"comment":"The relation Σ_y n_{w,y} p_{u w°, y w°} = b_{w,u^{-1}} determines n_{w,y} by inverting the Kazhdan-Lusztig matrix; it would be useful to state the inversion formula explicitly, since the author uses it in the next display without comment.","section":"§5.3, Corollary 5.5"},{"comment":"There are several typographical and formatting issues: 'U(gl n)' in the title should be 'U(𝔤𝔩_n)', and 'indecomposible' should be 'indecomposable'.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript presents an appealing conjecture-like result that is likely to be true, but the proof of the tilting claim is not yet complete: the filtration in §5.2 is not shown to be a U(𝔤𝔩_n)-module filtration, and Theorem 3.1 is only sketched. The geometric foundations in Section 2 are promising, but several assertions (such as Lemma 4.2 and Lemma 5.1) need to be supplied with proofs or precise references. I recommend major revision rather than rejection because the gaps seem fillable and the main idea is natural and potentially significant."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the extension of the U(gl_n)-action from Laumon spaces to the full relative quasimap space, and the claim that a specialized direct summand H_λ,w is the image of a tilting module under Soergel's equivalence. That is a real step forward, and the paper deserves credit for the geometric setup: fixed locus description, tangent weights, the BB stratification, and the multiplicity formula expressed through Bruhat paths and Kazhdan-Lusztig polynomials. The formula is explicit and checkable, and the connection to category O' is well motivated. If the main theorem is correct, this is a nice bridge between quasimap enumerative geometry and tilting theory.\n\nThe soft spots are where the proofs are sketches. The load-bearing one is in Section 5.2. The filtration by the open subsets U_i gives a filtration of H_λ as a vector space, and the successive quotients are isomorphic to dual Verma modules as vector spaces (or as graded/weight spaces). But the paper never proves that the kernels ker(H^*(U_i) → H^*(U_m)) are U(gl_n)-submodules. Without that, you do not have a dual Verma filtration in category O', and the conclusion that Υ^{-1}(H_λ) is tilting does not follow. The sentence \"By construction, ker(U_{i+1} → U_i) is isomorphic to a dual Verma module\" asserts exactly what needs proof. Poincaré duality gives the Verma filtration, but again only if the filtration is compatible with the module structure.\n\nSecond, the smoothness proof in Proposition 2.1 checks vanishing of the obstruction sheaf at T-fixed points and concludes global smoothness. That implication is not automatic for a T-equivariant perfect obstruction theory on a proper DM stack. I would want a sentence justifying it, either by semicontinuity or by localization of the perfect obstruction theory. If smoothness fails, the BB decomposition and equivariant localization are in trouble.\n\nThird, the proof of Theorem 3.1 verifies relations on fixed-point classes and asserts the rest. That is likely repairable, but as written it is a sketch. None of these gaps looks like a fatal error, and the overall strategy is coherent and worth taking seriously.\n\nA good referee should ask for a proof of the U(gl_n)-stability of the BB filtration, a justification of the fixed-point smoothness check, and a full verification of the algebra relations. With those, the paper could be a strong contribution. Send it to a serious referee; do not desk reject.","headline":"Genuinely new construction of a U(gl_n)-action on relative quasimap cohomology, aimed at geometric realization of tilting modules; the main theorem is plausible but rests on two unproved structural claims that a referee should push on.","tokens_in":15162,"tokens_out":1739,"would_cite":true,"duration_ms":18210,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14N35","14M15","17B10"],"pacs":[],"model":"deepseek-v4-flash","headline":"A fixed-locus summand of relative quasimap cohomology is a tilting module","keywords":["relative quasimaps","flag variety","Laumon space","equivariant cohomology","category O","tilting modules","Bialynicki-Birula decomposition","Kazhdan-Lusztig polynomials"],"falsifier":"Work out $\\mathcal{H}_{\\lambda,w}$ for $n=2$ and the simple reflection $w=s_1$ at degree one: the number of fixed points must equal the predicted dimension of the degree-one weight space, and the length of the dual Verma filtration (the number of torus-cell strata) must equal the expected multiplicity in Corollary 5.5; any mismatch would refute the tilting identification. A second check would be to find a proper Deligne-Mumford stack with a torus-equivariant coherent sheaf vanishing at all fixed points but not at a non-fixed point, which would break the smoothness argument in Proposition 2.1.","tokens_in":14011,"feed_emoji":"📐","tokens_out":12855,"duration_ms":108659,"temperature":0.7,"pith_summary":"The paper sets out to show that the equivariant cohomology of a compact moduli space of quasimaps from $\\mathbb{P}^1$ with one marked point to the flag variety — the relative quasimap space, a compactification of the Laumon space — carries a natural action of $U(\\mathfrak{gl}_n)$, and that a certain direct summand extracted from its torus-fixed locus is, up to a known categorical equivalence, a tilting module in the BGG category $\\mathcal{O}$. The point of this is that tilting modules are central objects in representation theory, but they are usually defined algebraically and hard to realize geometrically; here the module comes from cohomology classes and its decomposition into indecomposable tilting summands is computable from Bruhat paths and Kazhdan–Lusztig polynomials. If the argument is right, relative quasimap spaces give a geometric home for tilting modules of $\\mathfrak{gl}_n$ at regular integral lowest weights.","feed_headline":"Quasimap cohomology gives tilting modules for gl(n)","feed_subtitle":"A fixed-locus summand carries both Verma and dual Verma filtrations, so it is a tilting module.","key_machinery":"The load-bearing object is the relative quasimap space $QM_{\\mathrm{rel}}$: quasimaps from a chain of $\\mathbb{P}^1$'s to the flag variety with a fixed evaluation at the marked point, which compactifies the Laumon space by allowing bubbles to form at $\\infty$. The mechanism that carries the argument is the Bialynicki-Birula decomposition (the torus-attracting cell decomposition) of the fixed locus $M_0$: each attracting set is an affine fibration over its fixed component, so the resulting stratification gives a filtration of $\\mathcal{H}_{\\lambda,w}$ whose successive quotients are dual Verma modules, and Poincaré duality for orbifolds turns this into a Verma filtration. Around this sit the geometric correspondences defining $E_i$, $F_i$, and $H_i$ on equivariant cohomology, the fixed-point description of the torus action, and the categorical equivalence $\\Upsilon$ between $\\mathcal{O}'$ and $\\mathcal{O}$ that converts 'both filtrations' into the definition of a tilting module.","core_discovery":"The paper's central claim is that the $U(\\mathfrak{gl}_n)$-action on the equivariant cohomology of the Laumon space extends to the whole relative quasimap space $QM_{\\mathrm{rel}}$, and that after specializing equivariant parameters to a regular lowest weight $\\lambda$ and isolating the summand $\\mathcal{H}_{\\lambda,w}$ supported on fixed components meeting the nonsingular locus, this summand has a filtration by dual Verma modules and, by Poincaré duality, a filtration by Verma modules. Because it is finitely generated, locally finite for the subalgebra spanned by the $H_i$ and $F_i$, and has central elements acting by scalars, it lies in the dual category $\\mathcal{O}'$; the known equivalence $\\Upsilon$ then carries it into the ordinary category $\\mathcal{O}$, where having both filtrations is exactly the definition of a tilting module. The paper computes the graded dimension of $\\mathcal{H}_{\\lambda,w}$ by counting fixed points and derives the multiplicity formula $\\Upsilon^{-1}(\\mathcal{H}_{\\lambda,w}) = \\bigoplus_{y\\in W} T(y(\\lambda)-\\rho)^{\\oplus n_{w,y}}$, where the integers $n_{w,y}$ satisfy $\\sum_y n_{w,y}\\, p_{uw^\\circ, yw^\\circ} = b_{w,u^{-1}}$ with $b_{w,u}$ the number of Bruhat paths and $p$ the Kazhdan–Lusztig polynomials evaluated at $1$.","pith_inferences":["Computing $\\mathcal{H}_{\\lambda,w}$ explicitly for $\\mathfrak{gl}_2$ would show whether the Jordan blocks of the non-semisimple Cartan action match the predicted tilting decomposition, and whether the bubble-number filtration is the geometric shadow of the Verma filtration.","The same torus-cell method, applied to the whole cohomology or to partial flag varieties, may realize tilting objects in parabolic category $\\mathcal{O}$ or in singular blocks.","The multiplicity relation inverts Kazhdan–Lusztig polynomials against Bruhat-path counts; checking this identity combinatorially for small $W$ would test the formula independently of the geometry.","The smoothness proof's dependence on fixed-point vanishing marks the most fragile step; it also suggests the construction may extend to other GIT compactifications whenever a global quotient presentation is available."],"forward_implications":["The cohomology group $\\mathcal{H}_{\\lambda,w}$ provides a geometric model of a tilting module, so the torus-cell cycles give a basis adapted to a Verma flag.","The fixed-point count formula, combined with the Bruhat-path numbers and the inverse Kazhdan–Lusztig matrix, makes the tilting decomposition explicitly computable degree by degree.","The same correspondences should act on the equivariant $K$-theory of the relative quasimap space as $U_q(\\mathfrak{gl}_n)$, with generators adjusted to avoid square roots of tautological line bundles, as the paper's Remark 3.3 indicates.","Varying the evaluation point $w(x_0)$ produces a family of tilting modules; the paper's formula organizes this family through the Bruhat graph and may lead to a categorical action on quasimap cohomology."],"supporting_citations":[{"why":"defines quasimaps and the relative compactification with its stability condition, the space studied throughout.","marker":"[8]"},{"why":"constructs the $U(\\mathfrak{gl}_n)$-action on the cohomology of the Laumon space via correspondences, the action the paper extends.","marker":"[10]"},{"why":"provides the categorical equivalence $\\Upsilon$ between category $\\mathcal{O}$ and the dual category $\\mathcal{O}'$ used to identify tilting modules.","marker":"[24]"},{"why":"supplies the definition of tilting modules and the Kazhdan–Lusztig multiplicity formula used to compute $n_{w,y}$.","marker":"[13]"},{"why":"gives the Bialynicki-Birula decomposition for Deligne-Mumford stacks, the source of the dual Verma filtration.","marker":"[2]"},{"why":"supplies the general relative-quasimap and universal-curve framework used for the global quotient presentation.","marker":"[21]"},{"why":"provides Poincaré duality for orbifolds, which converts the dual Verma filtration into a Verma filtration.","marker":"[1]"},{"why":"supplies the equivariant cohomology, localization, and Gysin sequence tools behind the stratification arguments.","marker":"[7]"}],"fun_headline_variants":["Relative quasimaps yield tilting modules for gl(n)","Cohomology of quasimap space gives gl(n) tilting modules","Equivariant cohomology of quasimaps yields tilting modules","Tilting modules from relative quasimap cohomology","Quasimap cohomology extends to gl(n) tilting modules"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that the moduli space is smooth assumes that a torus-equivariant obstruction sheaf which vanishes at every torus-fixed point must vanish everywhere on the stack; if fixed-point vanishing does not force global vanishing, the decomposition, localization, and duality steps that build $\\mathcal{H}_{\\lambda,w}$ no longer apply.","fun_headline_variants_meta":{"raw":{"variants":["Relative quasimaps yield tilting modules for gl(n)","Cohomology of quasimap space gives gl(n) tilting modules","Equivariant cohomology of quasimaps yields tilting modules","Tilting modules from relative quasimap cohomology","Quasimap cohomology extends to gl(n) tilting modules"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000621,"raw_usage":{"total_tokens":2880,"prompt_tokens":950,"completion_tokens":1930,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":566,"completion_tokens_details":{"reasoning_tokens":1837}},"tokens_in":566,"tokens_out":1930,"duration_ms":13012,"temperature":1.0,"reasoning_tokens":1837,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:29:33.364820+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Work out $\\mathcal{H}_{\\lambda,w}$ for $n=2$ and the simple reflection $w=s_1$ at degree one: the number of fixed points must equal the predicted dimension of the degree-one weight space, and the length of the dual Verma filtration (the number of torus-cell strata) must equal the expected multiplicity in Corollary 5.5; any mismatch would refute the tilting identification. A second check would be to find a proper Deligne-Mumford stack with a torus-equivariant coherent sheaf vanishing at all fixed points but not at a non-fixed point, which would break the smoothness argument in Proposition 2.1.","supporting_citations":[{"cited_title":"Stable quasimaps to git quotients.Journal of Geometry and Physics, 75:17–47, 2014","cited_arxiv_id":null,"evidence_quote":"defines quasimaps and the relative compactification with its stability condition, the space studied throughout."},{"cited_title":"Gelfand–tsetlin algebras and cohomology rings of laumon spaces.Selecta Mathematica, 17:337–361, 2011","cited_arxiv_id":null,"evidence_quote":"constructs the $U(\\mathfrak{gl}_n)$-action on the cohomology of the Laumon space via correspondences, the action the paper extends."},{"cited_title":"´Equivalences de certaines cat´ egories de g-modules.CR Acad","cited_arxiv_id":null,"evidence_quote":"provides the categorical equivalence $\\Upsilon$ between category $\\mathcal{O}$ and the dual category $\\mathcal{O}'$ used to identify tilting modules."},{"cited_title":"American Mathematical Soc., 2021","cited_arxiv_id":null,"evidence_quote":"supplies the definition of tilting modules and the Kazhdan–Lusztig multiplicity formula used to compute $n_{w,y}$."},{"cited_title":"A luna ´ etale slice theorem for algebraic stacks.Annals of mathematics, 191(3):675–738, 2020","cited_arxiv_id":null,"evidence_quote":"gives the Bialynicki-Birula decomposition for Deligne-Mumford stacks, the source of the dual Verma filtration."},{"cited_title":"Cambridge University Press, 2007","cited_arxiv_id":null,"evidence_quote":"provides Poincaré duality for orbifolds, which converts the dual Verma filtration into a Verma filtration."},{"cited_title":"Springer, 1997","cited_arxiv_id":null,"evidence_quote":"supplies the equivariant cohomology, localization, and Gysin sequence tools behind the stratification arguments."}],"review_version":2}