{"id":"cc713d65-148e-4583-8468-720780adbf88","arxiv_id":"1908.07510","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For every Lagrangian fibration of a projective hyper-Kähler manifold, the perverse filtration equals the monodromy weight filtration of an associated type III degeneration.","lead":"This paper proves that two seemingly different filtrations on the cohomology of hyper-Kähler manifolds coincide: the perverse filtration from a Lagrangian fibration equals the monodromy weight filtration from a carefully chosen degeneration. A smart generalist might read it because it unifies two major tools in algebraic geometry, giving a structural result previously known only for K3 surfaces.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim rests on unverified application of Soldatenkov's degeneration theorem: the h from the signature argument is not checked to satisfy [12]'s polarization/period-domain hypotheses, so the P=W equality depends on an external existence result.","rationale":"The reader's weakest-assumption analysis identifies the same load-bearing point: the proof depends on Soldatenkov's external degeneration theorem and on the existence of a suitable integral class h. The paper's signature argument for h is mathematically sound given b2(M) ≥ 4 and the signature computation for V_ρ^⊥. The remaining doubt is not an internal inconsistency but an unverified match between the h constructed here and the hypotheses of [12, Theorem 4.6]. Since relying on a cited theorem is normal mathematical practice, and no concrete counterexample or failure of the cited theorem has been identified, the concern does not warrant changing the reader's ACCEPT verdict. The recommended concrete test would settle whether the concern actually lands; absent a positive failure, the verdict should remain unchanged.","tokens_in":5497,"tokens_out":48738,"duration_ms":1071756,"concrete_test":"Independently check the exact hypotheses of [12, Theorem 4.6]. For a concrete Lagrangian fibration (e.g., an elliptic K3 fibration or a Hilbert scheme of a K3 carrying a fibration), choose rational η, β, ρ, compute V_ρ^⊥, pick a primitive integral h, and verify whether (N_{β,ρ}, x) is a nilpotent orbit with x in the period-domain component of M and with h of type (1,1) for the constructed family. If yes, the concern is resolved; if no, Theorem 3 requires additional hypotheses.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 3 requires an actual type III projective degeneration with monodromy N_{β,ρ}. The proof delegates this to Soldatenkov [12, Theorem 4.6]. In Section 10, the only condition explicitly verified is existence of an integral class h ∈ H^2(M,Z) with q(h)>0 and q(h,β)=q(h,ρ)=0; the signature argument indeed produces such h from V_ρ^⊥. But the paper does not verify that this h satisfies the full hypotheses of [12, Theorem 4.6]: in particular, that h is a polarization, i.e. of type (1,1) for a period point x in the relevant period domain component, and that the pair (N_{β,ρ}, x) is an admissible nilpotent orbit. The sentence 'These nilpotent orbits eventually provide the required degeneration through global Torelli' is exactly where the load is carried. If [12] requires h to be a (1,1) polarization of the nearby fibers, the two-line signature argument in Section 10 is insufficient, and the equality P_k = W_{2k} would not be established. If [12]'s theorem only needs q(h)>0 and N h=0, the concern dissolves. This is the most load-bearing step because the entire comparison of filtrations in Section 11 is conditional on the existence of the degeneration.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a P=W theorem for Lagrangian fibrations of projective hyper-Kähler manifolds. For any Lagrangian fibration π:M→B, it constructs a type III projective degeneration f:M→Δ whose general fiber is deformation equivalent to M and whose monodromy weight filtration satisfies P_kH^*(M,Q) = W_{2k}H^*_{lim}(Q) = W_{2k+1}H^*_{lim}(Q) under the natural identification H^*(M,Q)≅H^*_{lim}(Q). The proof combines the sl2×sl2 splitting of the perverse filtration from [9] with the Looijenga–Lunts–Verbitsky algebra, constructs a nilpotent operator N_{β,ρ} from a relative ample class β and an auxiliary positive class ρ, realizes this operator as logarithmic monodromy using Soldatenkov's degeneration theorem [12], and then matches the two weight decompositions through the representation theory of so(5,C). The paper also derives applications to the 'Perverse = Hodge' dimensional equality and to the multiplicativity of the perverse filtration.","tokens_in":5777,"tokens_out":14126,"duration_ms":151734,"significance":"If the cited degeneration theorem is applied correctly, this is a clean and conceptually satisfying proof of the expected P=W phenomenon for Lagrangian fibrations of compact hyper-Kähler manifolds. The argument is elegant and concise, and the explicit so(5,C) matching in Section 11 is a particularly nice part of the paper. The authors are careful to credit prior work, and the honest remark in Section 5 that the argument is not logically independent of [9] is appropriate. The theorem also gives a unified explanation of the previously known 'Perverse = Hodge' equality and the multiplicativity of the perverse filtration.","major_comments":[],"minor_comments":[{"comment":"The proof of Theorem 3 delegates the existence of the degeneration to [12, Theorem 4.6], but the manuscript does not state the precise hypotheses of that theorem. Since this is the only step that connects the algebraically defined N_{β,ρ} to an actual degeneration, I recommend quoting the full statement of [12, Theorem 4.6] and explicitly verifying all of its hypotheses, including the role of the integral class h and any restrictions on b2(M).","section":"Section 10"},{"comment":"The existence of an integral h with q(h)>0 and q(h,β)=q(h,ρ)=0 is stated in two sentences. In the case b2=4, the orthogonal complement of Vρ has signature (1,0), so a brief comment on why this positive-definite rank-one rational subspace contains an integral class of positive square would make the argument fully transparent.","section":"Section 10"},{"comment":"The identification of the so(5,C) weight decomposition V^{i,j} with the decomposition P^{i,j}_C is asserted without derivation. A short explanation using H = H_η + H_β and −√−1 K_{23} = H_η − H_β would improve readability and make the matching of the two filtrations easier to follow.","section":"Section 11"},{"comment":"The reference to Soldatenkov [12] is listed as 'to appear'; if a final journal and year are available, they should be added.","section":"References"}],"recommendation":"minor_revision","confidential_remarks":"The main theorem is plausible and the proof is a careful assembly of existing results. The principal reservation is the heavy dependence on [12, Theorem 4.6]; if possible, the editor may wish to have the Soldatenkov theorem checked by a second referee. The self-citation of [9] is not a circularity concern because the P=W equality itself is not assumed as input."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper proves a real open case of P=W: for any Lagrangian fibration of a projective hyper-Kähler manifold, the perverse filtration coincides with the monodromy weight filtration of an explicit type III degeneration. Before this, that identity was known only for K3 surfaces, so the novelty is substantial. The paper is short and carefully assembled from two main ingredients: the canonical splitting of the perverse filtration from Shen–Yin [9], and Soldatenkov's construction of type III degenerations [12]. It also recovers Perverse=Hodge and multiplicativity of the perverse filtration. The authors are transparent that their proof is not logically independent of [9] because two of them are co-authors; I take that as honesty rather than a flaw.\n\nI checked the representation-theoretic step in Section 11. The matching of the perverse decomposition with the monodromy weight decomposition via the so(5)-action is explicit and correct. The conclusion P_k = W_{2k} = W_{2k+1} follows once the degeneration with monodromy N_{β,ρ} exists.\n\nThe one soft spot is the construction of that degeneration. Section 10 delegates to [12, Theorem 4.6] and only verifies the existence of an integral class h with q(h)>0 and q(h,β)=q(h,ρ)=0. The signature argument indeed produces such h. The stress-test concern is whether [12] additionally requires h to be a polarization of type (1,1) at the relevant period point, and whether the pair (N_{β,ρ},x) is an admissible nilpotent orbit. The authors write \"from the proof of [12, Theorem 4.6]\" these conditions suffice, and N h = 0 follows from the orthogonality conditions, but they do not spell out the remaining verifications. If [12]'s hypotheses are exactly as stated, the proof is complete. If a polarization condition is hidden there, the two-line argument is thin. I cannot resolve this from the text alone; it is precisely the kind of thing a referee should check. This is a minor-to-moderate concern, not a load-bearing flaw.\n\nThe citation pattern is appropriate. The reference to [9] is central, and the overlap is disclosed. The reference to [12] is the right tool. No circularity: the paper does not assume the P=W equality as an input.\n\nWho benefits: anyone working on hyper-Kähler geometry, perverse filtrations, or degenerations. It is a short note that settles a conjecture and deserves serious peer review. My recommendation: send it out with a referee who knows Soldatenkov's construction, and ask for a one-paragraph clarification of the hypotheses in [12]. I would accept contingent on that check.","headline":"A genuinely new P=W theorem for Lagrangian fibrations, proved by cleanly assembling Shen–Yin's splitting with Soldatenkov's degenerations; the only soft spot is the terse handoff to Soldatenkov's hypotheses.","tokens_in":6311,"tokens_out":3985,"would_cite":true,"duration_ms":36639,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14D07","14D06","14J42"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any Lagrangian fibration of a projective hyper-Kähler manifold, the perverse filtration coincides with the monodromy weight filtration of a type III degeneration.","keywords":["P=W conjecture","perverse filtration","monodromy weight filtration","hyper-Kähler manifolds","Lagrangian fibration","type III degeneration","limiting mixed Hodge structure","hard-Lefschetz operator algebra"],"falsifier":"Take an explicit Lagrangian fibration on a projective hyper-Kähler manifold (for instance from moduli of sheaves on a K3 surface) and check, in the Beauville–Bogomolov–Fujiki lattice, whether there is an integral class $h$ with $q(h)>0$ and $q(h,\\beta)=q(h,\\rho)=0$; if no such $h$ exists, the prescribed-monodromy degeneration used in the proof cannot be built, and the claimed identification would have to be verified by another construction.","tokens_in":5308,"feed_emoji":"🔗","tokens_out":18021,"duration_ms":163778,"temperature":0.7,"pith_summary":"The paper establishes a $P=W$ statement for projective hyper-Kähler manifolds that admit a holomorphic Lagrangian fibration: the perverse filtration on the cohomology of the fibration equals, in even-indexed degrees, the monodromy weight filtration of a suitably chosen maximally unipotent (type III) degeneration of the same manifold. The authors construct such a degeneration for every Lagrangian fibration, not just for K3 surfaces, thereby proving the general case of a conjecture that had previously been settled only there. This matters because it moves the perverse filtration, which is defined from the geometry of the fibration, into the theory of limiting mixed Hodge structures, and it explains the known 'Perverse = Hodge' dimension equalities and the multiplicativity of the perverse filtration.","feed_headline":"Perverse filtration matches monodromy weights","feed_subtitle":"A maximally unipotent degeneration makes the two filtrations agree on hyper-Kähler cohomology.","key_machinery":"The argument is carried by the Lie algebra of cohomology operators generated by all hard-Lefschetz cup-product $\\mathfrak{sl}_2$-triples, together with three classes in $H^2(M,\\mathbb{Q})$: a relative ample class $\\eta$, the pullback $\\beta$ of an ample class on the base, and a class $\\rho$ of positive square orthogonal to both. The first two provide an $\\mathfrak{sl}_2\\times\\mathfrak{sl}_2$-action whose weights split the perverse filtration; adjoining $\\rho$ promotes this to an $\\mathfrak{so}(5)$-action. The nilpotent operator $N_{\\beta,\\rho}=[L_\\beta,\\Lambda_\\rho]$ satisfies $N^3=0$ and is identified with the logarithmic monodromy of a type III degeneration supplied by a cited construction theorem. Representation theory of $\\mathfrak{so}(5)$ then shows that the weight decomposition for $N_{\\beta,\\rho}$ is a rotation of the perverse weight decomposition, producing the equality of filtrations.","core_discovery":"The central claim is Theorem 3: given any holomorphic Lagrangian fibration $\\pi: M \\to B$ with $M$ a projective hyper-Kähler manifold, there exists a projective type III degeneration $f:\\mathcal{M}\\to\\Delta$ (maximally unipotent logarithmic monodromy, $N^2\\neq 0$, $N^3=0$) whose general fibres are deformation equivalent to $M$, together with an identification of cohomology rings $H^*(M,\\mathbb{Q}) = H^*_{\\lim}(\\mathbb{Q})$ under which $P_k H^*(M,\\mathbb{Q}) = W_{2k}H^*_{\\lim}(\\mathbb{Q}) = W_{2k+1}H^*_{\\lim}(\\mathbb{Q})$ for every $k$. In other words, the perverse filtration of the fibration is exactly the evenly indexed monodromy weight filtration of a canonical degeneration. Because the identification is an isomorphism of rings, the perverse filtration inherits multiplicativity under cup product from the monodromy weight filtration.","pith_inferences":["The same $\\mathfrak{so}(5)$ bookkeeping should work for any triple of classes with the same orthogonality and positivity relations, so this version of $P=W$ is probably a property of the hard-Lefschetz operator algebra rather than of the particular fibration geometry.","The proof suggests a recipe for testing other settings: find an integral class of positive square orthogonal to two isotropic classes, build the corresponding nilpotent orbit, and compare the two weight decompositions; this is a finite lattice-and-representation-theory check in each example.","If a concrete example fails to have such an integral class, the construction used here cannot produce the degeneration, although the equality of filtrations might still hold through a degeneration with different monodromy; the paper does not address that possibility.","The methods should extend to non-projective compact hyper-Kähler manifolds or fibrations over non-projective bases as long as the period-map construction of the degeneration remains available."],"forward_implications":["The perverse filtration is multiplicative under cup product, since the monodromy weight filtration is multiplicative and the identification preserves the ring structure.","The 'Perverse = Hodge' dimension identities $\\dim \\mathrm{Gr}^P_i H^{i+j} = \\dim \\mathrm{Gr}^F_i H^{i+j}$ follow from the main equality, so perverse-filtration dimensions of Lagrangian fibrations are read off from ordinary Hodge numbers.","The conjecture for Lagrangian fibrations, previously known for K3 surfaces, is now established in all dimensions for every projective hyper-Kähler manifold carrying such a fibration.","The theorem provides a compact geometric analogue of the $P=W$ conjecture from non-abelian Hodge theory, with the monodromy weight filtration playing the role of the weight filtration on character varieties.","The identity $W_{2k}=W_{2k+1}$ forces the limiting mixed Hodge structure of the constructed degeneration to have only even weights, a structural constraint on any type III degeneration realizing a Lagrangian fibration."],"supporting_citations":[{"why":"Supplies the existence theorem for a type III degeneration with prescribed logarithmic monodromy; the external input that makes the comparison with the monodromy weight filtration possible.","marker":"[12]"},{"why":"Develops the $\\mathfrak{sl}_2\\times\\mathfrak{sl}_2$ weight decomposition that canonically splits the perverse filtration and derives the Perverse = Hodge dimension equality.","marker":"[9]"},{"why":"Computes the hard-Lefschetz operator algebra as $\\mathfrak{so}$ of the extended quadratic space, used to place $\\beta,\\eta,\\rho$ in an $\\mathfrak{so}(5)$ action.","marker":"[3]"},{"why":"Conjectures this P=W statement for Lagrangian fibrations and proves the K3 case, setting the target that Theorem 3 extends.","marker":"[4]"},{"why":"Introduces the Lie algebra generated by hard-Lefschetz $\\mathfrak{sl}_2$-triples and its weight grading, the structural home of the cohomology operators.","marker":"[7]"},{"why":"Gives the matching identification of that Lie algebra with $\\mathfrak{so}$ of the extended quadratic space.","marker":"[15]"},{"why":"Supplies the explicit relations for the $\\mathfrak{so}(5)$ action used in the final weight comparison.","marker":"[14]"},{"why":"Identifies the nilpotent operator $N_{\\beta,\\rho}$ with a commutator inside the Lie algebra, connecting the monodromy candidate to the $\\mathfrak{so}(5)$ action.","marker":"[6]"},{"why":"Identifies nearby-fibre cohomology with logarithmic de Rham hypercohomology and matches the cup product with the monodromy weight filtration, yielding the multiplicativity consequence.","marker":"[13]"}],"fun_headline_variants":["P=W for Lagrangian fibrations","Type III degeneration proves P=W","Matching filtrations: perverse and monodromy","Maximally unipotent degeneration aligns filtrations","Hyper-Kähler P=W: perverse equals monodromy weight"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on a cited theorem that builds a type III degeneration with prescribed logarithmic monodromy, and that theorem needs an integral second-cohomology class $h$ of positive square orthogonal to the two isotropic classes; if no such class exists, the degeneration is not produced and the equality of filtrations is not established by this argument.","fun_headline_variants_meta":{"raw":{"variants":["P=W for Lagrangian fibrations","Type III degeneration proves P=W","Matching filtrations: perverse and monodromy","Maximally unipotent degeneration aligns filtrations","Hyper-Kähler P=W: perverse equals monodromy weight"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001235,"raw_usage":{"total_tokens":4984,"prompt_tokens":769,"completion_tokens":4215,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":385,"completion_tokens_details":{"reasoning_tokens":4138}},"tokens_in":385,"tokens_out":4215,"duration_ms":31948,"temperature":1.0,"reasoning_tokens":4138,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:06:47.140772+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an explicit Lagrangian fibration on a projective hyper-Kähler manifold (for instance from moduli of sheaves on a K3 surface) and check, in the Beauville–Bogomolov–Fujiki lattice, whether there is an integral class $h$ with $q(h)>0$ and $q(h,\\beta)=q(h,\\rho)=0$; if no such $h$ exists, the prescribed-monodromy degeneration used in the proof cannot be built, and the claimed identification would have to be verified by another construction.","supporting_citations":[{"cited_title":"Soldatenkov, Limit mixed Hodge structures of hyperk¨ ahler manifolds, Mosc","cited_arxiv_id":null,"evidence_quote":"Supplies the existence theorem for a type III degeneration with prescribed logarithmic monodromy; the external input that makes the comparison with the monodromy weight filtration possible."},{"cited_title":"Topology of Lagrangian fibrations and Hodge theory of hyper-K\\\"ahler manifolds","cited_arxiv_id":"1812.10673","evidence_quote":"Develops the $\\mathfrak{sl}_2\\times\\mathfrak{sl}_2$ weight decomposition that canonically splits the perverse filtration and derives the Perverse = Hodge dimension equality."},{"cited_title":"Torus fibers and the weight filtration","cited_arxiv_id":"1908.05110","evidence_quote":"Conjectures this P=W statement for Lagrangian fibrations and proves the K3 case, setting the target that Theorem 3 extends."},{"cited_title":"Looijenga and V","cited_arxiv_id":null,"evidence_quote":"Introduces the Lie algebra generated by hard-Lefschetz $\\mathfrak{sl}_2$-triples and its weight grading, the structural home of the cohomology operators."},{"cited_title":"Verbitsky, Cohomology of compact hyper-K¨ ahler manifolds and its applications, Geom","cited_arxiv_id":null,"evidence_quote":"Gives the matching identification of that Lie algebra with $\\mathfrak{so}$ of the extended quadratic space."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the explicit relations for the $\\mathfrak{so}(5)$ action used in the final weight comparison."},{"cited_title":"Kurnosov , A","cited_arxiv_id":null,"evidence_quote":"Identifies the nilpotent operator $N_{\\beta,\\rho}$ with a commutator inside the Lie algebra, connecting the monodromy candidate to the $\\mathfrak{so}(5)$ action."},{"cited_title":"Steenbrink , Limits of Hodge structures, Invent","cited_arxiv_id":null,"evidence_quote":"Identifies nearby-fibre cohomology with logarithmic de Rham hypercohomology and matches the cup product with the monodromy weight filtration, yielding the multiplicativity consequence."}],"review_version":1}