{"id":"fe6967ab-071d-48ec-a684-ac6978f827f1","arxiv_id":"2505.03986","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every finite group action on a cubic threefold singular along a line, plane, or the chordal curve is linearizable; actions on cubics singular along a conic are generally not linearizable, though all such actions are unirational.","lead":"This math paper studies whether finite group symmetries of cubic threefolds with non-isolated singularities can be straightened out into linear symmetries of projective space. It finds that for most singularity types every finite group action can be linearized, but for cubics singular along a conic most actions cannot.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5.1's proof relies on a false generic-freeness claim: an involution in PGL2 fixes a line in the Veronese surface S, so the S-action is not generically free and Proposition 2.1 cannot be applied as stated.","rationale":"The reader identified the introduction's fixed-point linearizability claim as the weakest assumption. That claim is actually sound for the non-cone cubic threefolds under consideration: if a finite group G fixes a singular point p, projection from p gives a G-equivariant birational map X ⇢ P3, because X has multiplicity exactly 2 at p (X is not a cone) and a general line through p meets X in one residual point. Thus the reduction to fixed-point-free actions is valid. The real load-bearing flaw is in the linearization proofs: the paper repeatedly invokes Proposition 2.1 with the claim that a certain base action is generically free, but for the Veronese surface S in Theorem 5.1 (and the P2-factor in Theorem 6.2) the action of any involution fixes a line, so generic freeness fails. This is a concrete, checkable false statement, not a mere omission. Since Theorem 5.1 is one of the paper's principal results and its proof depends on this assertion, the central claim is not established as written. A rigorous fix would require either proving linearizability for non-generically-free base actions via an alternative argument or restricting the theorem and supplying direct constructions for involution-containing groups. Because the submitted proof is invalid for most finite subgroups in that case, the appropriate verdict is REJECT in its current form.","tokens_in":13192,"tokens_out":18976,"duration_ms":172543,"concrete_test":"Take G = C2 ⊂ PGL2 acting on the chordal cubic via the involution induced by (x1:x2) ↦ (x2:x1). Compute the induced action on the Veronese surface S = P2: with coordinates (z0,z1,z2) for binary quadrics, the involution sends (z0,z1,z2) to (z2,z1,z0). The fixed locus is the line {z0 = z2}. A general point on this line is fixed by the nontrivial element, so the action is not generically free. Since the proof of Theorem 5.1 explicitly requires generic freeness on S to apply Proposition 2.1, this single check refutes the stated proof. Then check whether an alternative G-equivariant birational map X ⇢ P3 exists for this G; if none is supplied, Theorem 5.1 is unproved for involutions.","verdict_should_be":"REJECT","load_bearing_attack":"The proof of Theorem 5.1 (chordal cubic) asserts: 'For all such G, the induced action on S is generically free' (Section 5). This is false for any finite subgroup G containing an involution, e.g., G = C2. An involution in PGL2 induces on the Veronese surface S ≅ P2 (the symmetric square representation) a linear transformation with eigenvalues 1, -1, 1, so its fixed locus is a line in S. Any point of that fixed line is fixed by the nontrivial element, so the action on S is not generically free. Consequently, the hypothesis of Proposition 2.1's linearizability statement fails, and the no-name lemma argument does not go through for the P1-bundle X̃ → S. The same issue occurs in the proof of Theorem 6.2, where 'acts linearly and generically freely on P2' is claimed for the P2-factor of the birational model P1×P2; the action on that P2 has fixed lines for involutions. These are not mere gaps in computation but demonstrably false statements. The theorems may still be true, but as written the proofs are invalid for all finite subgroups containing an involution, which includes most of the groups listed in Theorem 5.1.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies finite group actions on cubic threefolds in P^4 whose singular locus has dimension at least one. It classifies the possible automorphism groups for the four singularity types (line, conic, plane, rational normal quartic) and proves G-linearizability in the line, plane, and chordal cases, non-linearizability for most conic cases, and G-unirationality for all cases. The methods combine normal forms, blowups, projections, the no-name lemma, and twists over nonclosed fields.","tokens_in":13493,"tokens_out":19493,"duration_ms":178080,"significance":"If the results are correct, they provide a fairly complete equivariant birational classification for this class of rational threefolds, extending the authors' earlier work on isolated singularities. The negative result for conic singularities (Theorem 4.3) is particularly interesting as a counterpoint to the isolated-singularity case. The paper also gives explicit normal forms and automorphism group lists that could be useful for moduli and arithmetic questions. However, the reliance on asserted automorphism group classifications and on a reduction whose proof is omitted means the significance is conditional on filling those gaps.","major_comments":[{"comment":"The claim that an action fixing a singular point of X is linearizable is used to reduce all four singularity types to the fixed-point-free case, but no proof or reference is provided. Since Theorems 3.1, 4.5, 5.1, and 6.2 assert statements for arbitrary finite subgroups, the authors need to justify this reduction. The natural argument via projection from the fixed point should be written out (or a reference given), because without it the proofs do not cover groups that fix a singular point.","section":"Section 1, second paragraph"},{"comment":"The automorphism group classifications are asserted after 'direct computation' with the computations largely omitted. These classifications are load-bearing: they are used to reduce the linearizability arguments to the listed groups. The authors should include the computations, at least in an appendix or supplementary material, or provide a clear algorithmic description that allows verification. As written, a reader cannot check that the lists are complete or that the stated generators are correct.","section":"Propositions 3.2, 3.3, 4.1, 4.2, 6.1"},{"comment":"The proof invokes an undefined 'Condition (A)' and asserts that the induced action on S is generically free and that the P1-bundle is G-linearized. Generic freeness is indeed true (the induced PGL2 action on the Veronese surface is faithful), but the lifting of the action to the vector bundle (the content of Condition (A)) is not established for all finite subgroups G, and the connection to Proposition 2.1 is too terse. Please expand this proof. The stress-test concern that involutions violate generic freeness is a misunderstanding: a fixed line in S does not prevent generic freeness, since the generic point of S is not on that line; the same clarification applies to the proof of Theorem 6.2.","section":"Section 5, proof of Theorem 5.1"}],"minor_comments":[{"comment":"There are several typos, including 'EQUIV ARIANT' in the header and 'group s' in the introduction.","section":"General"},{"comment":"The matrix display has entries like 'b 2' and 'a 2' where superscripts appear to be missing; these should be b^2, a^2, c^2, and d^2.","section":"Equation (6.2)"},{"comment":"The term 'Condition (A)' is used without a definition in this paper; it should be either defined or an explicit reference to a previous paper should be given.","section":"Section 5"},{"comment":"The proof is extremely brief and relies on [6, Proposition 2.6] without explaining how the hypotheses of that proposition are satisfied; a few more sentences would help the reader.","section":"Theorem 4.3"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: the conic case (Theorem 4.3) is a genuine new result, and the unirationality theorem (4.5) goes through cleanly. But the proofs of Theorems 5.1 and 6.2 contain a false claim about generically free actions, so those two theorems are unsupported as written. The normal-form and automorphism computations are extensive and look right, but they are largely asserted via 'direct computation'—a minor issue by comparison.\n\nThe stress-test note is correct. For any involution in PGL2, the induced action on the Veronese surface S ≅ P2 has eigenvalues 1, -1, 1 on Sym^2, so it fixes a line pointwise. Hence the assertion in Section 5 that 'for all such G, the induced action on S is generically free' is false for every finite subgroup containing an involution, which includes all the non-cyclic groups relevant there. Proposition 2.1's linearizability claim explicitly requires generic freeness on the base, so the no-name lemma argument does not apply. The same problem occurs in Section 6: the action on the P2 factor fixes a line for any element with eigenvalue -1, so it is not generically free for, say, C2. This is not a gap in one edge case; it breaks the stated proofs of two main theorems.\n\nThe theorems may still be true. For the chordal cubic, the structure of the Schwarzenberger bundle or other methods might give linearizability, but the paper would need to supply a correct argument. The automorphism-group classifications (Propositions 3.2, 3.3, 4.1, 4.2, 6.1) are plausible and probably checkable, but the 'direct computation' leaves a lot out. Minor note: the introduction's reduction to actions fixing no singular points is asserted without proof; for the line and conic cases it excludes some components, and a reader shouldn't have to take it on faith.\n\nBottom line: I'd send this to referees, but with a clear warning that the generic-freeness claim must be fixed before acceptance. The conic result alone is worth a serious look. I'd wait for a corrected version before citing the chordal- or plane-singularity linearizability results.","headline":"Main results are plausible and one is new, but the proofs of the chordal cubic and plane-singularity theorems rest on a false generic-freeness assertion and need repair.","tokens_in":14007,"tokens_out":4373,"would_cite":false,"duration_ms":43334,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14E07","14J30","14J50","14L30"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that every finite group action on a cubic threefold with singular locus a line, plane, or chordal quartic is linearizable, while conic-singular cubics mostly admit non-linearizable actions.","keywords":["cubic threefold","non-isolated singularities","linearizable actions","G-unirationality","finite group actions","chordal cubic","no-name lemma","equivariant birational geometry"],"falsifier":"Exhibit a finite subgroup of the automorphism group of one of the classified conic-singular cubics that fixes a singular point and whose action is not linearizable; this would refute the asserted reduction and force the theorems to be restricted to actions with no fixed singular points.","tokens_in":13018,"feed_emoji":"📐","tokens_out":12087,"duration_ms":104561,"temperature":0.7,"pith_summary":"The paper studies finite group actions on cubic threefolds whose singular locus has positive dimension but is not an isolated point. Such threefolds come in four families: singular along a line, along a conic, along a plane, or along a rational normal quartic (the chordal cubic). The paper proves that every finite group action is linearizable — equivariantly birational to a linear action on projective three-space — in the line, plane, and chordal cases, and that in the conic case most finite actions are not linearizable, with even dihedral groups providing explicit obstructions. It also proves that every finite action on all four families is $G$-unirational, meaning a linear representation of the group dominates the threefold equivariantly. The contrast with the isolated-singularity story, where rational K-unstable cubics had linearizable actions, shows that non-isolated singularities introduce genuinely new non-linearizable equivariant behavior.","feed_headline":"All finite actions linearize on three singular cubic types","feed_subtitle":"Only conic-singular cubics host non-linearizable dihedral group actions.","key_machinery":"The case analysis rests on the classification of non-cone cubic threefolds with non-isolated singularities, which reduces to four normal forms according to the singular locus: a line, a conic, a plane, or a rational normal quartic. For each form the paper identifies the invariant subvariety that carries the argument — the singular line, the plane spanned by the conic, the singular plane itself, or the secant-variety structure of the chordal cubic. The workhorse is the no-name lemma, a theorem saying that a generically free linear group action on a vector bundle over a projective space is equivariantly birational to a trivial bundle; this turns projective bundles over linearizable bases into linearizable varieties. A second lemma converts $G$-unirationality into a statement about twists over nonclosed fields, so an invariant unirational subvariety with rational points forces the whole threefold to be $G$-unirational. In the conic case, unprojection — rewriting the cubic by introducing one new coordinate to eliminate an invariant plane — produces intersections of two quadrics in which the non-linearizable dihedral actions can be recognized by the elliptic curve they stabilize.","core_discovery":"The central claim, stated as Theorems 3.1, 5.1, 6.2, 4.3, and 4.5, is a near-complete dichotomy for finite group actions. If the singular locus is a line, a plane, or the chordal quartic, every finite subgroup of the automorphism group acts linearizably: after an equivariant birational transformation the threefold becomes a projective bundle over a linearizable base, and the no-name lemma trivializes the bundle. If the singular locus is a conic, the story splits: the paper constructs models as intersections of two quadrics and shows that, for even-order dihedral subgroups generated by a specific rotation and swap, the action is not linearizable, because a Klein four subgroup fixes a cubic surface while a residual involution fixes a smooth elliptic curve. At the same time, every finite action on every one of the four families is $G$-unirational, since the singular plane or line supplies a $G$-invariant unirational subvariety whose twisted forms have rational points. The chordal cubic shows a further subtlety: each finite subgroup linearizes, but the full $\\mathrm{PGL}_2$ automorphism group does not.","pith_inferences":["The paper establishes a sharp separation between $G$-unirationality, which holds for every finite action, and linearizability, which fails for conic dihedral actions; a testable expectation is that non-linearizable finite actions on rationally connected threefolds often come from subgroups with an elliptic curve in their fixed locus.","The remaining open linearizability questions in the conic family are concentrated in finitely many parameter values; computing whether the finite groups not covered by Theorem 4.3 admit equivariant birational maps to $\\mathbb{P}^3$ would complete the classification.","If the asserted reduction that actions fixing a singular point are linearizable ever fails, the theorems would still apply to the fixed-point-free actions explicitly analyzed, but the statements quantifying over all finite subgroups would need to be narrowed."],"forward_implications":["Every finite group action on a cubic threefold singular along a line is linearizable (Theorem 3.1).","Every finite group action on a cubic threefold singular along a plane is linearizable (Theorem 6.2).","Every finite subgroup action on the chordal cubic is linearizable, although the full automorphism group action is not (Theorem 5.1 and Remark 5.2).","For conic-singular cubics, the even dihedral subgroups generated by $a=\\zeta_n$ and the swap are not linearizable, so the full automorphism group action is not linearizable (Theorem 4.3).","Every finite subgroup action on any non-cone cubic threefold with non-isolated singularities is $G$-unirational (Theorem 4.5)."],"supporting_citations":[{"why":"The classification of singular cubic 3-folds supplies the four normal forms (line, conic, plane, chordal quartic) on which the case analysis rests.","marker":"[19]"},{"why":"Gives the criterion equating $G$-unirationality with unirationality of all twists, used in Proposition 2.3 and Theorem 4.5.","marker":"[12]"},{"why":"Supplies the no-name lemma that turns generically free linear actions on vector bundles over projective space into trivial bundles, the engine of the linearization proofs.","marker":"[13]"},{"why":"Earlier work on singular cubic threefolds supplies the fixed-elliptic-curve criterion used to prove non-linearizability in the conic case.","marker":"[6]"},{"why":"Shows cubic hypersurfaces with rational points are unirational, completing the $G$-unirationality argument for twists.","marker":"[15]"},{"why":"Companion work on singular cubic threefolds supplies the unprojection technique rewriting cubics as intersections of two quadrics.","marker":"[5]"},{"why":"Theorem on connected algebraic groups acting on three-dimensional Mori fibrations rules out an equivariant birational map for the full automorphism group of the chordal cubic.","marker":"[3]"},{"why":"Arithmetic study of quadric-surface fibrations motivates the conic case and its open rationality questions.","marker":"[9]"}],"fun_headline_variants":["Finite group actions on most singular cubics are linearizable","Cubic threefolds: linearizable except for dihedral on conic singularities","Non-isolated cubic singularities: finite actions linearize, barring one family","Dihedral actions break linearizability on conic-singular cubics","Near-complete dichotomy for finite actions on singular cubic threefolds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes without proof that any finite group action fixing a singular point of the cubic is linearizable; the reduction to the fixed-point-free case for all four singularity types depends on this claim.","fun_headline_variants_meta":{"raw":{"variants":["Finite group actions on most singular cubics are linearizable","Cubic threefolds: linearizable except for dihedral on conic singularities","Non-isolated cubic singularities: finite actions linearize, barring one family","Dihedral actions break linearizability on conic-singular cubics","Near-complete dichotomy for finite actions on singular cubic threefolds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000945,"raw_usage":{"total_tokens":3953,"prompt_tokens":783,"completion_tokens":3170,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":399,"completion_tokens_details":{"reasoning_tokens":3072}},"tokens_in":399,"tokens_out":3170,"duration_ms":22178,"temperature":1.0,"reasoning_tokens":3072,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:40:38.717432+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a finite subgroup of the automorphism group of one of the classified conic-singular cubics that fixes a singular point and whose action is not linearizable; this would refute the asserted reduction and force the theorems to be restricted to actions with no fixed singular points.","supporting_citations":[{"cited_title":"Yokoyama","cited_arxiv_id":null,"evidence_quote":"The classification of singular cubic 3-folds supplies the four normal forms (line, conic, plane, chordal quartic) on which the case analysis rests."},{"cited_title":"Duncan and Z","cited_arxiv_id":null,"evidence_quote":"Gives the criterion equating $G$-unirationality with unirationality of all twists, used in Proposition 2.3 and Theorem 4.5."},{"cited_title":"Hajja and M","cited_arxiv_id":null,"evidence_quote":"Supplies the no-name lemma that turns generically free linear actions on vector bundles over projective space into trivial bundles, the engine of the linearization proofs."},{"cited_title":"Cheltsov, Yu","cited_arxiv_id":null,"evidence_quote":"Earlier work on singular cubic threefolds supplies the fixed-elliptic-curve criterion used to prove non-linearizability in the conic case."},{"cited_title":"Koll´ ar","cited_arxiv_id":null,"evidence_quote":"Shows cubic hypersurfaces with rational points are unirational, completing the $G$-unirationality argument for twists."},{"cited_title":"Blanc, A","cited_arxiv_id":null,"evidence_quote":"Theorem on connected algebraic groups acting on three-dimensional Mori fibrations rules out an equivariant birational map for the full automorphism group of the chordal cubic."}],"review_version":1}