{"id":"e7e33881-2faf-4e59-a7d6-2aa842ac1cc9","arxiv_id":"2603.24464","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"On Shimura varieties, the top weight-graded cohomology of the reductive Borel-Serre compactification is canonically isomorphic to the intersection cohomology of the Baily-Borel compactification.","lead":"This paper proves that the intersection cohomology of a compactified Shimura variety is exactly the top-weight piece of a mixed Hodge structure coming from a different, non-algebraic compactification. That identification makes intersection cohomology behave like ordinary cohomology, with cup products, pullbacks, and cycle classes, and connects volumes of special cycles to cohomological degrees.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Surjectivity in Theorem 4.2 rests on Proposition 3.6, whose stalk computation for Lp-forms is only cited 'verbatim' from [36] and may not hold for arbitrary coefficients E; if wrong, the isomorphism becomes a mere injection.","rationale":"The reader's weakest assumption identifies Proposition 3.6 as the key unsecured step, and my analysis agrees. The reliance on Nair's unpublished preprint [24] for the Hodge-theoretic bridge is a separate concern, but it is at least a specific citation; Proposition 3.6 is a sketched argument with a vague reference to Zucker's paper, and it is the step that provides surjectivity. If it fails, the central claim weakens from an isomorphism to an injection, and all subsequent constructions—cup product, pullbacks, Gysin maps, modular intersection classes—lose their canonical source. The concrete test I propose is a direct computation of the stalk cohomology for a nontrivial coefficient system in a small example, which would either confirm the 'verbatim' citation or expose a counterexample. Until such a check is done, the conditional verdict remains appropriate.","tokens_in":22696,"tokens_out":10086,"duration_ms":93956,"concrete_test":"Perform the stalk cohomology computation of A_(p)(E) explicitly for a concrete non-trivial coefficient system on a Q-rank 1 example, e.g., G = SL_2, Γ a neat congruence subgroup, X the modular curve, E = Sym^m(C^2) with m > 0, and a boundary point corresponding to a Borel subgroup. Using the coordinates (a, z, n) and weight decomposition of Section 3.1, write a closed special form in W0C(E) near the boundary and determine the Lp condition on each weight component. Check whether the stalk cohomology of A_(p)(E) at that point equals H^i(n_P,E)_{≥0} for all p greater than some explicitly computed bound. If any negative-weight component contributes to the Lp cohomology for arbitrarily large p, Proposition 3.6 is false; if the computation succeeds, verify whether the same argument can be extended to all strata and all irreducible E.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central isomorphism Theorem 4.2 requires two ingredients: the Hodge-theoretic injection of Proposition 2.6/Theorem 4.1, and the analytic surjectivity of Theorem 3.7. The surjectivity is the part that upgrades the injection to an isomorphism, and it hinges on Proposition 3.6: for p sufficiently large, the inclusion W0C(E) ↪ A_(p)(E) is a quasi-isomorphism. The proof of Proposition 3.6 only checks stalk cohomology, claiming that the stalk of Lp-cohomology at a boundary point in X_P has contributions only from weights β ≥ 0 and is identified with H^i(n_P,E)_{≥0}, 'shown verbatim following [36, (2.2.1),(3.1.4),(3.1.7)]'. This is not a derivation. It is unclear whether [36] treats arbitrary irreducible rational coefficient systems E on all boundary strata, including non-maximal parabolics, or only the trivial local system. The estimate -pβ-2ρ_P ≤ 0 for 'arbitrarily large p' is stated without the actual spectral sequence or weight filtration that would justify it, and a negative-weight class might survive for some p if the bound p > 2ρ_P/(-β) is not uniform. If the stalk computation fails, W0C(E) → A_(p)(E) is not a quasi-isomorphism, the map W0H*(X,E) → H*_(2)(X,E) need not be surjective, and Theorem 4.2 reduces to the canonical injection GrW ↪ IH. Since the cup products, pullbacks, and modular intersection classes all use the identification IH ≅ GrW W0H*, the applications collapse without surjectivity.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a precise identification between the top weight quotient of the mixed Hodge structure on the cohomology of the reductive Borel–Serre compactification of a Shimura variety and the intersection cohomology of its Baily–Borel compactification (Theorem 1.1 and the coefficient-general Theorem 4.2). The strategy combines three ingredients: (i) Hodge-theoretic weight truncations of mixed Hodge modules, following Nair–Vaish and Morel, which supply a canonical map and an injection of the top-weight quotient into intersection cohomology (Proposition 2.6, Theorem 4.1); (ii) analytic weighted cohomology à la Goresky–Harder–MacPherson, specifically an eventual quasi-isomorphism between weight-profile-0 weighted forms and L^p forms (Proposition 3.6), from which a surjection onto L^2/intersection cohomology is derived (Theorem 3.7); and (iii) Zucker's conjecture to identify L^2 cohomology with intersection cohomology. The paper then constructs canonical cup products, pullbacks, Gysin maps, and modular intersection classes on intersection cohomology, and applies them to automorphic Lefschetz properties and to a cohomological interpretation of Kudla's Siegel–Weil formula. The overall architecture is coherent and rests on substantial established results, but two load-bearing points—the stalk computation in Proposition 3.6 and the reliance on the unpublished Nair preprint for Theorem 4.1—are not fully justified in the manuscript.","tokens_in":23110,"tokens_out":4897,"duration_ms":57319,"significance":"If the main theorems are correct, the paper gives a satisfying and useful answer to a known gap in the theory: intersection cohomology of Baily–Borel compactifications is not formally a ring or a functor, but Hodge theory provides canonical structures. The claimed canonical cup product, pullbacks, projection formula, and universal cycle classes for special cycles would resolve open questions raised by Nair–Rai and would give a cohomological interpretation of Kudla's analytic Siegel–Weil formula. These applications are meaningful and go beyond a purely formal exercise. The paper also gives credit where it is due: it uses rather than reproves deep results of Goresky–Harder–MacPherson, Zucker, Saito, Looijenga–Rapoport, Saper–Stern, Ayoub–Zucker, Nair–Vaish, and others. The main deficit is that the manuscript's own contribution—the surjectivity step and the weight comparison—contains a proof gap in the analytic quasi-isomorphism, and it delegates a substantial Hodge-theoretic theorem to an unpublished preprint. If those gaps are filled, the significance is high.","major_comments":[{"comment":"The proof of Proposition 3.6 is the analytic crux that turns the canonical injection into an isomorphism. It asserts that the stalk cohomology of L^p-forms at a boundary stratum has contributions only from weights β ≥ 0 and is identified with H^i(n_P,E)_{≥0}, but the derivation is deferred to a citation: “shown verbatim following [36, (2.2.1),(3.1.4),(3.1.7)]”. No spectral sequence, no filtration, and no uniformity argument over all roots and over arbitrary coefficients E is given. The critical estimate −pβ−2ρ_P ≤ 0 for “arbitrarily large p” is stated without derivation, and it is not evident from the cited passages that [36] treats arbitrary irreducible rational coefficient systems on all boundary strata, including non-maximal parabolics. Since Proposition 3.6 is exactly what upgrades the injection of Proposition 2.6 to the isomorphism in Theorem 4.2, the paper should supply a complete","section":"§3.3, Proposition 3.6"},{"comment":"Theorem 4.1—weighted cohomology carries mixed Hodge structures and is compatible with weight-truncated intersection complexes—is the entire Hodge-theoretic bridge between weighted cohomology and intersection cohomology. It is attributed to Nair’s unpublished preprint [24] (“Theorem 4.3.1, Proposition 4.4.1”). The manuscript does not state the cited results in detail or indicate which parts are being assumed. Because the main theorem of the paper is logically dependent on this external unpublished source, the argument is only conditionally established. The author should either prove these statements, give a complete statement with a verification, or cite a published version if one now exists.","section":"§4.1, Theorem 4.1"},{"comment":"The proof of the compatibility of the constructed cup product with Poincaré–Verdier duality relies on the commutativity of a diagram involving the dualizing complex ω_̂X[−2n] and the claim that any morphism between the two sheaves is determined by its restriction to X because Hom(Q_̂X ⊗ Q_̂X, ω_̂X[−2n]) ≅ H^{2n}_c(̂X;Q)^* is one-dimensional. For sheaves or complexes on a stratified space, extension from the open stratum is not automatic without checking the boundary strata; the sheaf Hom is not simply a hypercohomology group with a single generator. This step may be repairable, but as written it is not a full justification that the cup product agrees with the Verdier duality pairing on IC_{X^{BB}}.","section":"§4.2, Proposition 4.5"},{"comment":"The claim that for unitary subgroups SU(1,m) ⊂ SU(1,n) the restriction map lands in the inner cohomology H^m_!(X_H) uses the identification IH^m(X_H^{BB}) ≅ H^m_!(X_H), cited to [22, Proposition 6.7.3]. For arbitrary noncompact Hermitian locally symmetric varieties, this identification is not automatic in all degrees and requires assumptions on the boundary strata (e.g., isolated singularities). Since this corollary is advertised as resolving an open question in [25] for different Q-ranks, the relevant hypotheses of [22, Proposition 6.7.3] should be stated and verified for the unitary Shimura varieties under consideration.","section":"§5.1, Corollary 5.5"}],"minor_comments":[{"comment":"The text contains repeated corrupted arrows and symbols, e.g., “⮯➤⬓⬄⫸➤⬓⬄” in several isomorphisms. These should be cleaned up.","section":"Throughout, displayed formulas"},{"comment":"The proof of Lemma 3.5 is compressed; the definition of the Siegel set S_{Q,t',ω'} and the verification that its difference with μ_P(R_Q × ω) is relatively compact are sketched. This is probably repairable but would benefit from more detail.","section":"§3.2, Lemma 3.5"},{"comment":"The use of r(T) is inconsistent: it first denotes the rank of T, later r−r(T) appears as a codimension. The notation should be defined explicitly to avoid confusion with the fixed r.","section":"§5.2, notation"},{"comment":"The claim that “V_1 has no G(Q)-coinvariants by semisimplicity” is true for semisimple modules, but the relevant semisimplicity of the G(Q)-module IH^*(X^{BB}) should be stated precisely here, since the earlier reference [25, Proposition 3.6] is about a different formulation.","section":"§5.1, Lemma 5.2"},{"comment":"The reference [24] is an unpublished preprint from 2012. If it has been superseded or published, the citation should be updated; if not, the dependence should be flagged in the introduction and the preprint made available for referees.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The central claim is plausible and the architecture is compelling, but the paper currently has two load-bearing gaps: an unproved analytic stalk computation for general coefficients and a heavy reliance on an unpublished Nair preprint. These are fixable in principle but require real work, not just editing. I would not reject, because there is no indication of circularity or fitted constants; the issues are about missing verification. I would advise the editor to send the manuscript back with a request for a complete proof or a fully stated and verified use of the external results, and to check whether the published version of Nair’s work exists."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper claims a clean, useful theorem: for a neat arithmetic group and an irreducible representation E, the top-weight quotient of the mixed Hodge structure on the reductive Borel–Serre cohomology is canonically isomorphic to the intersection cohomology of the Baily–Borel compactification, with general coefficients. If true, this gives intersection cohomology canonical cup products, pullbacks, Gysin maps, and cycle classes for special cycles. That is a genuinely useful package, and the author is honest about what is new: the injection is essentially known to Nair, while the surjectivity is the analytic new step.\n\nThe paper is well-written and clearly organized. The architecture rests on major established results—GHM weighted cohomology, Zucker's conjecture, Saito's mixed Hodge modules, Nair–Vaish weightless cohomology—and the applications are real: it answers the open Lefschetz question for unitary Shimura varieties in the missing degrees and gives a cohomological interpretation of Kudla's Siegel–Weil formula. There is no circularity and no fitted parameters.\n\nThe soft spot is exactly where the stress-test note points. The surjectivity that upgrades the injection to an isomorphism depends on Proposition 3.6: for large p, the inclusion W0C(E) into L^p forms is a quasi-isomorphism. The proof only checks stalk cohomology and says the stalk computation is 'shown verbatim following [36]'. Since [36] is written for trivial coefficients in many places, it is not obvious that the argument works for arbitrary irreducible E on all boundary strata, including non-maximal parabolics. This matters because every application—cup products, pullbacks, modular intersection classes—uses the identification IH ≅ GrW W0H*. If Prop 3.6 fails in some cases, the theorem collapses to a canonical injection. This is a genuine gap in the written proof, not a demonstrated error; it is the kind of thing a careful referee should check.\n\nA second, lesser concern is that Theorem 4.1, the Hodge-theoretic bridge, is cited to Nair's unpublished 2012 preprint. The author says the injectivity is 'essentially known to Nair,' which is fair, but an unpublished preprint is load-bearing for a published claim.\n\nThe paper deserves a serious referee rather than desk rejection. Send it out, and ask the referee specifically to verify Prop 3.6 for general coefficients and to comment on the dependence on Nair's preprint.","headline":"Plausible synthesis that would give canonical operations on intersection cohomology of Shimura varieties, but the analytic surjectivity step is only sketched via a 'verbatim' citation to Zucker and needs a referee's check.","tokens_in":23562,"tokens_out":1833,"would_cite":true,"duration_ms":23304,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14G35","14C30","32S35","14F43","11F27"],"pacs":[],"model":"deepseek-v4-flash","headline":"Intersection cohomology of Shimura varieties is the top-weight quotient of the mixed Hodge structure on the reductive Borel–Serre compactification, giving it canonical cup products, pullbacks, and cycle classes.","keywords":["intersection cohomology","Shimura varieties","Baily-Borel compactification","reductive Borel-Serre compactification","mixed Hodge structures","weighted cohomology","Siegel-Weil formula","automorphic Lefschetz properties"],"falsifier":"For a Q-rank 2 Shimura variety (for instance, a Siegel modular variety) with a nontrivial irreducible coefficient system E, compute the stalk cohomology of the sheaf H^i(A_{(p)}(E)) at a point of a boundary stratum as p grows, using the local geodesic retraction model, and compare it with the truncated Lie algebra cohomology H^i(n_P,E)_{≥0}. If any weight β < 0 contributes to the stalk at arbitrarily large p, the quasi-isomorphism of Proposition 3.6 fails, and the surjectivity step collapses, leaving only an injection.","tokens_in":22574,"feed_emoji":"⚖️","tokens_out":6712,"duration_ms":72338,"temperature":0.7,"pith_summary":"The paper proves that intersection cohomology of the singular Baily–Borel compactification of a Shimura variety is exactly the top-weight quotient of the mixed Hodge structure on the reductive Borel–Serre compactification, a friendlier (though non-algebraic) compact model. This identification transplants the existing ring structure and functoriality from the model's cohomology to intersection cohomology, restoring the operations the perverse-sheaf formalism cannot canonically provide. The result yields canonical cycle classes for special cycles and their Hecke translates, universal across resolutions of the singular compactification. It also gives new automorphic Lefschetz properties for noncompact Shimura varieties and a cohomological interpretation of the analytic Siegel–Weil formula.","feed_headline":"Intersection cohomology emerges as a top-weight quotient","feed_subtitle":"A Hodge-theoretic model gives intersection cohomology canonical cup products, pullbacks, and cycle classes.","key_machinery":"The argument runs on two rails that meet at weighted cohomology. On the Hodge side, weight truncation functors on mixed Hodge modules define the 'weightless complex', whose hypercohomology maps onto intersection cohomology with image equal to the top weight quotient. On the analytic side, weighted cohomology sheaves WηC(E) are built from differential forms whose boundary strata are truncated by Lie-algebra weights; the profile η=0 computes cohomology of the reductive Borel–Serre compactification, while the profile η=−ρ computes the intersection complex. The paper shows that W0C(E) embeds into the sheaf of L^p-forms and becomes quasi-isomorphic for large p, so it surjects onto L^2-harmonic fo","core_discovery":"The central claim is a canonical isomorphism of pure Hodge structures: the graded piece GrW_{k+w} of the weight filtration on the weighted cohomology W0H^k(X,E) of the reductive Borel–Serre compactification is isomorphic to the intersection cohomology IH^k(X^BB,E) of the Baily–Borel compactification. The map is constructed Hodge-theoretically as the natural morphism from the weight-zero weighted complex to the intersection complex; injectivity follows from weight filtrations. Surjectivity is proved analytically by showing that the weight-zero complex is quasi-isomorphic for large p to the sheaf of L^p-forms, and that every L^2-harmonic form (which represents intersection cohomology) is such","pith_inferences":["The same weight-comparison strategy is likely to work for other topological compactifications of locally symmetric spaces, since the analytic surjectivity theorem stated in the paper already covers arbitrary locally symmetric spaces, not only Hermitian ones; this could produce canonical characteristic classes in more general settings.","The canonical cycle class for a special cycle may carry strictly more information than its cohomology class in ordinary cohomology: the example of a punctured modular curve shows the intersection class can be nonzero even when Chow groups vanish, suggesting a refined boundary-sensitive invariant.","If the higher-degree generating series of modular intersection classes is at most quasi-modular, the Hodge-theoretic model could be used to locate exactly where holomorphy of theta lifts fails; comparing the two generating series coefficient-by-coefficient is a concrete quantitative test.","The identification gives a new bridge between automorphic forms and algebraic cycles: the semisimple structure of intersection cohomology as a module over rational points could be read off from the weight filtration on the reductive Borel–Serre compactification, potentially transferring arithmetic applications."],"forward_implications":["Intersection cohomology of Shimura varieties acquires a canonical graded-commutative cup product that realizes the Poincaré–Verdier duality pairing, making it behave like the cohomology of a smooth projective variety.","Morphisms between Shimura varieties induce canonical pullbacks and Gysin maps on intersection cohomology satisfying the projection formula, so intersection cohomology becomes functorial.","Every Shimura subvariety and its Hecke translates define a canonical modular intersection class in intersection cohomology, universal across all resolutions of the Baily–Borel compactification and restricting to the classical cycle class in ordinary cohomology.","Automorphic Lefschetz properties extend to noncompact Shimura varieties: for unitary groups U(n,1) restricting to U(m,1), the pullback on cohomology is injective in degrees up to m, with the middle-degree image landing in inner cohomology, resolving an open question.","The analytic Siegel–Weil formula is given a cohomological reading: the generating series of degrees of modular intersection classes equals the special value of a Siegel Eisenstein series."],"fun_headline_variants":["Intersection cohomology is the top-weight quotient","Top-weight quotient gives canonical cup products","Hodge theory links weighted and intersection cohomology","Weight-zero complex model for intersection cohomology"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof of surjectivity depends on Proposition 3.6, the claim that for large p the natural inclusion of the weight-zero weighted complex into the sheaf of L^p-forms is a quasi-isomorphism, whose key stalk calculation is asserted to follow 'verbatim' from a cited source rather than derived in the paper; if this stalk identification fails for some coefficient system or boundary stratum, the theorem would reduce to a canonical injection of the top-weight quotient into intersec","fun_headline_variants_meta":{"raw":{"variants":["Intersection cohomology is the top-weight quotient","Top-weight quotient gives canonical cup products","Hodge theory links weighted and intersection cohomology","Weight-zero complex model for intersection cohomology"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000445,"raw_usage":{"total_tokens":2023,"prompt_tokens":614,"completion_tokens":1409,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":358,"completion_tokens_details":{"reasoning_tokens":1350}},"tokens_in":358,"tokens_out":1409,"duration_ms":13355,"temperature":1.0,"reasoning_tokens":1350,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T17:28:33.800074+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a Q-rank 2 Shimura variety (for instance, a Siegel modular variety) with a nontrivial irreducible coefficient system E, compute the stalk cohomology of the sheaf H^i(A_{(p)}(E)) at a point of a boundary stratum as p grows, using the local geodesic retraction model, and compare it with the truncated Lie algebra cohomology H^i(n_P,E)_{≥0}. If any weight β < 0 contributes to the stalk at arbitrarily large p, the quasi-isomorphism of Proposition 3.6 fails, and the surjectivity step collapses, leaving only an injection.","supporting_citations":[],"review_version":1}