{"id":"b9850057-f40c-44ec-9630-338cd7595e45","arxiv_id":"2501.12964","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The double EPW cube of a general Gushel-Mukai fourfold is the MRC quotient of the Hilbert scheme of twisted cubics, and it admits a Lagrangian covering family.","lead":"This paper proves a long-standing conjecture: for a general Gushel-Mukai fourfold, the double EPW cube, a six-dimensional hyperkähler manifold, is the maximal rationally connected quotient of the Hilbert scheme of twisted cubics on the fourfold. It also gives a new example of O'Grady's conjecture by covering the double EPW cube with Lagrangian subvarieties built from twisted cubics on hyperplane sections.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified.","rationale":"I read the paper as proving two main statements: the double EPW cube of a general GM fourfold is the MRC quotient of the Hilbert scheme of twisted cubics, and it admits a Lagrangian covering family. The proof is built on a categorical bridge: twisted cubics C map to pr_X(I_C(H)) in Ku(X), whose moduli space M^X_{σ_X}(1,-1) is birational to the double EPW cube via KKM22. The load-bearing ingredients are: (i) the detailed geometry of τ-, ρ-, and σ-cubics in Section 3; (ii) the computations of the projection objects and their self-Ext groups in Section 4; (iii) the deformation from very general to general fourfolds in Theorem 4.11; (iv) the dominance and P1-fiber statement in Theorem 5.5; and (v) the hyperplane-section construction of Lagrangian families in Theorem 5.13. I checked these steps for internal consistency. The τ-cubic fiber computation in Theorem 5.5 is valid: for a τ-cubic C, any τ-cubic with the same projection is a residue cubic of a fixed residue cubic C'', giving exactly P1; ρ-cubics have projection objects depending only on a residue line, so their image has dimension at most 3 and cannot affect a general fiber of a 6-dimensional target. The deformation argument in Theorem 4.11 is the softest point: it explicitly uses the open substack of σ-semistable objects, while the claimed conclusion is stability. However, stability is also an open condition, and very general points are stable for every stability condition, so the same argument with the stable open substack produces a nonempty Zariski open set of GM fourfolds for which all non-σ twisted cubics have stable projections. This is a standard fix and does not change the verdict. The main theorems depend on external results from FGLZ24 and KKM22, but those dependencies are clearly stated and are not circular. I therefore agree with the reader's ACCEPT verdict.","tokens_in":45269,"tokens_out":36825,"duration_ms":416197,"concrete_test":"Re-run the deformation proof of Theorem 4.11 with the open substack of σ-stable objects in place of the σ-semistable stack: define U^s as the preimage of the stable open substack, let Z^s be the complement of |U^s| in |H_{X/W}|, and check that p(Z^s) misses all very general points so that the same constructible-density argument produces an open dense S⊂W over which every non-σ twisted cubic has σ-stable projection. If this modified argument fails, Theorem 4.11 would only establish semistability, and the rational map to the hyperkähler moduli space would not be justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No load-bearing gap found in the central argument. The proof of Theorems 1.1 and 1.3 is coherent: the geometric classification of twisted cubics in Section 3, the Ext computations and stability results in Section 4, and the construction of the dominant P1-fibered map and Lagrangian covering family in Section 5 fit together. The weakest point, also flagged by the reader, is the deformation argument in Theorem 4.11: literally it proves membership in the open stack of σ-semistable objects, while the statement needs σ-stability. This is a presentational gap rather than a fatal flaw, because the stable locus is likewise open in the universally-gluable stack, and very general fibers are stable; re-running the same constructible-density argument with the stable open substack yields an open dense locus on which the relative projection objects are σ-stable. The reliance on the companion papers FGLZ24 and KKM22 is explicit and standard, and I found no internal inconsistency in the uses made of them.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies twisted cubics on Gushel-Mukai fourfolds and uses them to obtain two new results about double EPW cubes. It classifies twisted cubics into τ-, ρ-, and σ-types, computes the projections of their ideal sheaves into the Kuznetsov component, proves stability of these projection objects for general X, and shows that the projection functor induces a dominant rational map from Hilb^{3t+1}_X to the Bridgeland moduli space M^X_{σ_X}(1,-1) with P^1 general fibers. Combining this with the birationality result of KKM22, the authors identify the double EPW cube as the MRC quotient of the Hilbert scheme of twisted cubics. They then construct a Lagrangian covering family on the double EPW cube using Hilbert schemes of twisted cubics on GM threefolds. The appendices reconstruct several classical examples categorically and prove smoothness and irreducibility of the Hilbert scheme of twisted cubics on GM threefolds.","tokens_in":45374,"tokens_out":15674,"duration_ms":175640,"significance":"If correct, Theorem 1.1 confirms a conjecture of Iliev and Manivel and places double EPW cubes in the established family of hyperkähler manifolds obtained as MRC quotients of Hilbert schemes of low-degree rational curves on Fano fourfolds. Theorem 1.3 provides a new instance of O'Grady's conjecture on Lagrangian covering families and adds to the short list of hyperkähler manifolds known to admit such families. The paper's strengths are its detailed geometric classification of twisted cubics, the explicit Ext computations that underpin the stability results, and the categorical framework that uniformly treats classical examples such as Fano varieties of lines, LLSvS eightfolds, and double EPW sextics. The reliance on companion papers is transparent, and the main argument is internally coherent.","major_comments":[{"comment":"The proof of Theorem 4.11 establishes σ-semistability rather than σ-stability. The final step shows the inclusion φ_X(H_X) ⊂ |M^X_{σ_X}(1,-1)|, and by the notation of §2.4 together with [BLM+21, Lemma 21.12] this is the stack of σ-semistable objects, while the theorem statement asserts stability. This does not undermine the main results, since the rational map in Theorem 5.5 only needs semistability and the stable locus is open inside the universally-gluable stack, so the same constructible-density argument applied to the stable open substack yields an open dense locus on which the relative projection objects are σ-stable. I would ask the authors either to make this stability-openness argument explicit or to weaken the theorem to semistability.","section":"§4, Theorem 4.11"}],"minor_comments":[{"comment":"The proof treats the cases where C and C' are both ρ-cubics and both τ-cubics, but not the mixed τ/ρ case. Since the ρ-locus has dimension at most two inside a general fiber, the birationality conclusion in Theorem 5.13 is not endangered, but the lemma as stated should either cover the mixed case or be weakened to the statement actually used.","section":"§5.3, Lemma 5.12"},{"comment":"The display 'Hτ = Hilb^{3t+1}_X' should be read as 'Hτ is open and dense in Hilb^{3t+1}_X'; as written it could suggest equality of schemes, which is not what the proof establishes.","section":"§5.1, Proposition 5.3"},{"comment":"The statement 'j :֒→ X' is missing the source; it should read 'j : Y ֒→ X'.","section":"§2.3, Proposition 2.4"},{"comment":"In the proof of Theorem 5.5, the sentence 'As dim Hilb^{3t+1}_X ≥ 7 and pr only contracts curves in Hτ, we deduce...' would be clearer if it also noted that the local P^1-fiber description rules out any additional component of dimension greater than 7 mapping dominantly to M^X_{σ_X}(1,-1).","section":"§5.2, Theorem 5.5"}],"recommendation":"minor_revision","confidential_remarks":"The paper depends on several companion papers, especially FGLZ24 and KKM22. The reliance is explicit and the uses appear consistent, but if any of these papers are not yet in final published form, the editor may wish to confirm that their statements are available to the community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a strong paper. It proves that for a general GM fourfold the double EPW cube is the MRC quotient of the Hilbert scheme of twisted cubics, and it produces a Lagrangian covering family on the cube, a new example for O'Grady's conjecture. The main theorems are new and the proof hangs together.\n\nThe genuinely new content is the systematic geometry of twisted cubics: the three-type classification (tau, rho, sigma), the homological characterizations, the residue-cubic relation, and the explicit projections into the Kuznetsov component. The dimension counts are consistent, and the P1-fiber description of the rational map is convincing. Appendix B gives a useful standalone result: smoothness and irreducibility of Hilb^{3t+1}_Y for a general GM threefold. Appendix A's uniform reconstruction of classical Lagrangian families is a nice bonus.\n\nSoft spots, in proportion. The main one is Theorem 4.11: the deformation argument as written establishes membership in the open stack of sigma-semistable objects, while the statement needs sigma-stability. The stress-test note is right that this is fixable—the stable locus is open in the universally gluable stack and very general fibers are stable, so the same constructibility argument works—but the text should say so. It is a presentational gap, not a fatal flaw.\n\nThe reliance on FGLZ24 and KKM22 is heavy, but explicit and standard. The paper does not hide its dependencies, and self-citation is legitimate here because the cited results are the tools the paper builds on. I did not find a circular step or a post-hoc exclusion. Several technical lemmas are hard to verify line-by-line, but the internal consistency of the computations is good.\n\nIf I had to bet, the two main theorems are correct. This is the kind of paper that deserves a serious referee—not a desk reject—and with minor clarifications (especially around semistability vs stability) it should be accepted.\n\nRecommendation: engage with it; send it to a competent referee, and expect a minor-revision outcome.","headline":"This paper confirms the Iliev–Manivel conjecture for general GM fourfolds, and its main argument is coherent; the only real weakness is a presentational gap between semistability and stability in the deformation step.","tokens_in":45962,"tokens_out":2094,"would_cite":true,"duration_ms":22866,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14F08","14J42","14J45","14D20","14D23"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a general Gushel–Mukai fourfold X, the double EPW cube is the maximal rationally connected quotient of the Hilbert scheme of twisted cubics, and it admits a Lagrangian covering family.","keywords":["double EPW cubes","Gushel–Mukai fourfolds","twisted cubics","Bridgeland moduli spaces","Kuznetsov components","hyperkähler manifolds","Lagrangian covering families","maximal rationally connected quotients"],"falsifier":"Find a general GM fourfold $X$ and a twisted cubic $C\\subset X$ that is not a $\\sigma$-cubic for which $pr_X(I_C(H))$ is strictly $\\sigma_X$-semistable with respect to a generic stability condition; Theorem 4.11 would then fail and the rational map $\\mathrm{Hilb}^{3t+1}_X\\dashrightarrow M^X_{\\sigma_X}(1,-1)$ would not be defined on the dense $\\tau$-cubic locus. Alternatively, exhibit a point of $M^X_{\\sigma_X}(1,-1)$ whose fiber under the dominant rational map has dimension different from 1, contradicting the claimed $\\mathbb{P}^1$ general fibers.","tokens_in":45043,"feed_emoji":"","tokens_out":11957,"duration_ms":105153,"temperature":0.7,"pith_summary":"This paper establishes that the double EPW cube $\\widetilde{C}_X$, a six-dimensional hyperkähler manifold attached to a general Gushel–Mukai fourfold $X$, is the maximal rationally connected quotient of the Hilbert scheme $\\mathrm{Hilb}^{3t+1}_X$ of twisted cubics on $X$. A twisted cubic is a closed one-dimensional subscheme with Hilbert polynomial $3t+1$ with respect to the polarization $H$. The result confirms a long-standing expectation that sporadic hyperkähler manifolds built from EPW data arise as curve-counting quotients, in the same way that Fano varieties of lines and the eightfold associated with a cubic fourfold do. The paper also proves that $\\widetilde{C}_X$ carries a Lagrangian covering family, giving a new example of the conjecture that every projective hyperkähler manifold is covered by Lagrangian subvarieties. The route passes through the Kuznetsov component of $X$: twisted cubics project to stable objects of class $\\Lambda_1-\\Lambda_2$, and the projection map has $\\mathbb{P}^1$ fibers.","feed_headline":"Twisted cubics on GM fourfolds give the double EPW cube","feed_subtitle":"A new proof identifies the six-dimensional hyperkähler manifold as the MRC quotient of the Hilbert scheme of twisted cubics.","key_machinery":"The load-bearing mechanism is the Kuznetsov component $\\mathrm{Ku}(X)$: a K3 category appearing in the semiorthogonal decomposition of $D^b(X)$, equipped with the family of stability conditions $\\mathrm{Stab}^\\circ(\\mathrm{Ku}(X))$ and the rank-two numerical lattice generated by $\\Lambda_1,\\Lambda_2$. The paper classifies twisted cubics into $\\tau$-, $\\rho$-, and $\\sigma$-cubics by how they sit inside the Grassmannian $\\mathrm{Gr}(2,5)$, then computes the projections $pr_X(I_C(H))$ for each type. For $\\tau$-cubics the projection has self-Ext algebra $\\mathbb{C}\\oplus\\mathbb{C}^6[-1]\\oplus\\mathbb{C}[-2]$ and is stable; moreover a $\\tau$-cubic and its residue cubic have isomorphic projections, which is exactly what makes the general fiber of the map $\\mathrm{Hilb}^{3t+1}_X\\dashrightarrow M^X_{\\sigma_X}(1,-1)$ a $\\mathbb{P}^1$. This categorical description turns the Hilbert scheme into a rational $\\mathbb{P}^1$-fibration over a hyperkähler sixfold, and the same fiber structure, restricted to hyperplane sections, supplies the Lagrangian covering family.","core_discovery":"For a general GM fourfold $X$, the projection functor from the derived category of $X$ to its Kuznetsov component $\\mathrm{Ku}(X)$ sends the ideal sheaf $I_C(H)$ of a twisted cubic $C\\subset X$ that is not a $\\sigma$-cubic to a $\\sigma_X$-stable object of class $\\Lambda_1-\\Lambda_2$ in the Bridgeland moduli space $M^X_{\\sigma_X}(1,-1)$. This defines a dominant rational map $\\mathrm{Hilb}^{3t+1}_X \\dashrightarrow M^X_{\\sigma_X}(1,-1)$ whose general fibers are $\\mathbb{P}^1$; because hyperkähler manifolds are not uniruled, the six-dimensional moduli space is the maximal rationally connected quotient of the Hilbert scheme. A birational comparison, already available, identifies $M^X_{\\sigma_X}(1,-1)$ with the double EPW cube $\\widetilde{C}_X$, giving the paper's main theorem. The same correspondence, restricted to twisted cubics contained in hyperplane sections (GM threefolds), produces a dominant family of Lagrangian subvarieties of $\\widetilde{C}_X$, hence a Lagrangian covering family.","pith_inferences":["Because the proof uses only openness and deformation of the relevant moduli stacks, the MRC-quotient identification should extend from the general locus to any smooth ordinary GM fourfold whose Kuznetsov component has the same numerical lattice; this is a natural deformation-theoretic extension the paper does not state.","The $\\mathbb{P}^1$-family of residue cubics of a $\\tau$-cubic suggests that $\\mathrm{Hilb}^{3t+1}_X$ is birational to a $\\mathbb{P}^1$-bundle over $\\widetilde{C}_X$, meaning the Hilbert scheme should admit an explicit two-step contraction similar to the eightfold construction; the paper does not work out this birational model.","If the Lagrangian covering family has the expected cohomological consequences, the Lefschetz standard conjecture for $\\widetilde{C}_X$ should follow from the general theory of Lagrangian-covered hyperkähler manifolds; the paper does not address this.","A direct boundary test would be to run the same projection computation on a Hodge-special GM fourfold containing a plane; the classification of $\\sigma$-cubics suggests the map should fail there, delimiting exactly how general 'general' must be."],"forward_implications":["If the main theorem is right, the double EPW cube $\\widetilde{C}_X$ becomes a curve-counting invariant: its birational class and period are governed by twisted cubics on $X$, matching the role of conics for double EPW sextics.","The theorem confirms the conjectured analogy with cubic fourfolds: just as the eightfold associated with a cubic fourfold is the MRC quotient of twisted cubics on that fourfold, the double EPW cube is the same quotient for a GM fourfold.","The existence of a Lagrangian covering family on $\\widetilde{C}_X$ makes the double EPW cube a new confirmed case of the conjecture that every projective hyperkähler manifold is covered by Lagrangian subvarieties.","The covering family is built from GM threefold hyperplane sections: for a general smooth hyperplane section $Y\\subset X$, the Hilbert scheme of twisted cubics on $Y$ maps birationally onto a Lagrangian subvariety of $\\widetilde{C}_X$, giving an explicit geometric description of those Lagrangians.","The same categorical method uniformly reconstructs Lagrangian covering families for Fano varieties of lines on cubic fourfolds, eightfolds from twisted cubics on cubic fourfolds, and double (dual) EPW sextics, as shown in the paper's appendix."],"supporting_citations":[{"why":"Shows that the double EPW cube $\\widetilde{C}_X$ and the Bridgeland moduli space $M^X_{\\sigma_X}(1,-1)$ have the same period point and are birational, transferring the quotient statement to $\\widetilde{C}_X$.","marker":"[KKM22]"},{"why":"Constructs the family of stability conditions $\\mathrm{Stab}^\\circ(\\mathrm{Ku}(X))$ and proves that $M^X_{\\sigma_X}(a,b)$ is a projective hyperkähler manifold for coprime $a,b$.","marker":"[PPZ22]"},{"why":"Supplies the Lagrangian family on $M^X_{\\sigma_X}(1,-1)$ and the stability criterion (its Lemma 4.12) used to prove stability of the projected ideal sheaves.","marker":"[FGLZ24]"},{"why":"Provides the homological classification and Ext computations for low-degree curves on GM fourfolds that the authors adapt to twisted cubics.","marker":"[GLZ24]"},{"why":"Identifies $K_{\\mathrm{num}}(\\mathrm{Ku}(X))=\\mathbb{Z}\\Lambda_1\\oplus\\mathbb{Z}\\Lambda_2$ for non-Hodge-special $X$, the lattice input for the moduli space.","marker":"[KP18]"},{"why":"States the expectation that double EPW cubes are MRC quotients of twisted-cubic Hilbert schemes and gives the conic/EPW-sextic analogue the paper extends.","marker":"[IM11]"},{"why":"Provides the cubic-fourfold model: the eightfold is the MRC quotient of twisted cubics on a cubic fourfold, the pattern this paper transfers to GM fourfolds.","marker":"[LLSvS17]"},{"why":"Gives the general construction of stability conditions on Kuznetsov components of GM varieties used to define $\\mathrm{Stab}^\\circ$.","marker":"[BLMS23]"}],"fun_headline_variants":["Twisted cubics on GM fourfolds give double EPW cube","Double EPW cube is MRC quotient of twisted cubic Hilbert scheme","GM fourfold cubics produce double EPW cube as MRC quotient","Twisted cubics yield hyperkähler cube via MRC quotient"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes $X$ is a general Gushel–Mukai fourfold, so its Kuznetsov component has numerical Grothendieck group exactly $\\mathbb{Z}\\Lambda_1\\oplus\\mathbb{Z}\\Lambda_2$ and the standard family of stability conditions is well behaved; if that categorical input failed for some non-Hodge-special fourfold, the projection map from the Hilbert scheme to the Bridgeland moduli space would not be defined.","fun_headline_variants_meta":{"raw":{"variants":["Twisted cubics on GM fourfolds give double EPW cube","Double EPW cube is MRC quotient of twisted cubic Hilbert scheme","GM fourfold cubics produce double EPW cube as MRC quotient","Twisted cubics yield hyperkähler cube via MRC quotient"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000769,"raw_usage":{"total_tokens":3392,"prompt_tokens":917,"completion_tokens":2475,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":533,"completion_tokens_details":{"reasoning_tokens":2397}},"tokens_in":533,"tokens_out":2475,"duration_ms":18190,"temperature":1.0,"reasoning_tokens":2397,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T16:34:55.656756+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a general GM fourfold $X$ and a twisted cubic $C\\subset X$ that is not a $\\sigma$-cubic for which $pr_X(I_C(H))$ is strictly $\\sigma_X$-semistable with respect to a generic stability condition; Theorem 4.11 would then fail and the rational map $\\mathrm{Hilb}^{3t+1}_X\\dashrightarrow M^X_{\\sigma_X}(1,-1)$ would not be defined on the dense $\\tau$-cubic locus. Alternatively, exhibit a point of $M^X_{\\sigma_X}(1,-1)$ whose fiber under the dominant rational map has dimension different from 1, contradicting the claimed $\\mathbb{P}^1$ general fibers.","supporting_citations":[],"review_version":1}