{"id":"4ee5d6ff-6812-4177-879c-d13a245a30ec","arxiv_id":"2510.14795","paper_version":2,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A log surface has finitely many weakly log canonical klt models, but this paper's main new contribution is a method reducing that finiteness to boundedness of polarizations.","lead":"This paper proves that a log surface has only finitely many weakly log canonical klt models up to log isomorphism, by a new reduction to boundedness of polarizations. The result itself was already known to experts via the cone conjecture; the claimed novelty is the method, which is conditional on a boundedness input.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.2, Case 2 relies on an unjustified abundance step: from K_X+B≡0 it asserts ν=κ via [Fuj12, Thm 8.1] to conclude K_X+B∼_R0; without this, isotriviality for real-coefficient boundaries fails and Theorem 4.4 collapses.","rationale":"The paper’s genuinely new contribution is Theorem 4.4, reducing finiteness of log surfaces to boundedness of polarization. Its proof rests on Theorem 3.2, which asserts isotriviality of smooth families of terminal 0-pairs after an étale base change. The real-coefficient case is the only place where the new method must go beyond the Q-coefficient techniques of [Amb05]/[Xu25]. I checked the relevant passage: the step from P≡P_K to P∼_R P_K is essential, because later the proof needs ⌊nP⌋∼nP_K and a constant linear system. The only justification offered is the line attributing ν=κ to [Fuj12, Thm 8.1]. That theorem is a semi-ampleness criterion, not an assertion that pseudo-effective log surfaces satisfy abundance; the equality ν=κ is exactly what needs proof. Since K_X+B≡0 gives ν=0, the desired conclusion would follow from abundance (κ=0 and semiampleness), but the paper does not invoke abundance explicitly. As written, this is a non-sequitur. Lemma 3.7’s comparison of cycles on fibers also requires the trivialization of the family, which is already in place after Theorem 3.6, but the text’s wording is loose. I do not see a comparably serious flaw elsewhere: Lemma 4.5’s use of Alexeev’s boundedness is appropriate for the finite-coefficient klt pairs in question; Lemma 4.8’s amplitude argument is sound; the skipped proof in §5, Case 2, concerns only the already-known finiteness statement, not Theorem 4.4. For these reasons I agree with the reader’s identification of the weakest assumption, and the appropriate verdict remains CONDITIONAL, i.e., no change from the reader’s verdict.","tokens_in":14784,"tokens_out":17885,"duration_ms":150659,"concrete_test":"Independently re-derive the step in §3, Case 2: (i) check whether [Fuj12, Theorem 8.1] actually gives ν(K_X+B)=κ(K_X+B) under the hypotheses K_X+B≡0 and B a boundary; if it only gives semi-ampleness under that equality, then (ii) supply a proof that κ(K_X+B)=0 (e.g., via the log abundance theorem for surfaces) and that semiample + numerically trivial implies ∼_R0. If the derivation cannot be completed, Theorem 3.2 fails for real boundaries and Theorem 4.4 is unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 3.2, Case 2 (real coefficients), after reducing to X=X×_C S, the argument sets −K_X=P_K+N (Zariski) and P=B−N, so P≡P_K. To promote this to P∼_R P_K (needed to make ⌊nP⌋∼nP_K and trivialize the family), the text asserts: “By the semi-ampleness theorem [Fuj12, Theorem 8.1] we have ν(K_X+B)=κ(K_X+B), that is, K_X+B∼_R0” (§3, Case 2). This is not a consequence of the stated hypotheses: K_X+B≡0 gives ν=0, but κ could be −∞ for a numerically trivial non-effective R-divisor, and [Fuj12, Thm 8.1] is a semi-ampleness criterion, not a proof that ν=κ holds automatically. One needs the log abundance theorem for surfaces, plus the fact that a semiample numerically trivial R-divisor is R-linearly trivial. The paper neither states nor proves this; the phrase “that is” conflates ν=κ with ∼_R0. Lemma 3.7’s “algebraic equivalence implies numerical equivalence” between divisors on different fibers also requires the canonical identification of the trivial family; as written it conflates cycles on the total space with cycles on a fiber. Since Theorem 4.4’s MMP argument uses Theorem 3.2 to make the terminalization family isotrivial, this gap is load-bearing for the paper’s new method, even though Theorem 1.3 itself is known via the cone conjecture.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a new approach to the finiteness of weakly log canonical (wlc) klt models of a log surface in a fixed 0-class. Its main structural result, Theorem 1.4, states that if the underlying varieties of the models form a class with bounded polarization, then only finitely many log pairs occur up to log isomorphism. The proof reduces the problem to an isotriviality statement for families of terminal 0-pairs, Theorem 3.2, and then applies an MMP argument to a terminalization family. The paper also claims Theorem 1.3, the full finiteness statement for log surfaces, as an application of Theorem 1.4 together with Alexeev's boundedness theorem and, in the K3 case, Kawamata/Sterk's results. The author notes that Theorem 1.3 is already known to experts via the cone conjecture, and presents Theorem 4.4 as the principal novelty.","tokens_in":15131,"tokens_out":21365,"duration_ms":181554,"significance":"The main idea is attractive and potentially useful: if the isotriviality theorem and the MMP reduction are correct, Theorem 4.4 gives a clean route from boundedness of polarizations to finiteness of models, and it suggests a template for higher dimensions. The paper is also careful to acknowledge that Theorem 1.3 is not new, and it clearly identifies Theorem 4.4 as the contribution. The rational-coefficient part of Theorem 3.2 is built on established machinery of Ambro and Xu. However, the real-coefficient case of Theorem 3.2 contains a load-bearing gap concerning abundance, and the application of Alexeev's boundedness theorem in §5 requires a uniformity statement that is not proved. These issues are local and potentially repairable, but as written they block acceptance.","major_comments":[{"comment":"The step from P≡P_K to P∼_R P_K (equivalently K_X+B∼_R0) is not justified. The paper asserts: \"By the semi-ampleness theorem [Fuj12, Theorem 8.1] we have ν(K_X+B)=κ(K_X+B), that is, K_X+B∼_R0.\" From K_X+B≡0 one only knows ν=0; κ can be −∞ for a numerically trivial non-effective R-divisor. Fujino's theorem is a semiampleness theorem, not an automatic equality ν=κ. One must first know that (X,B) is klt and K_X+B is nef, then use log abundance to get K_X+B semiample, and then use that a semiample numerically trivial R-divisor is R-linearly trivial. The subsequent claim that (X, B+ϵP) is klt for 0<ϵ≪1 also needs checking: if some coefficient of B is 1, adding ϵP can create coefficients >1, so the pair need not be klt. Since the rest of Case 2 depends on nP∼_R nP_K and on the basepoint-freeness of ⌊nP⌋, this gap is load-bearing for Theorem 3.2 and hence for Theorem 4.4.","section":"§3, Case 2"},{"comment":"The proof of Lemma 3.7 conflates cycles on the total space with cycles on fibers. The displayed formula \"(B_j)_s − B_j = Σ (V_s − V_{s0})\" has terms of different dimensions: (B_j)_s is a divisor on the fiber X_s, while B_j is a divisor on the total space X. Algebraic equivalence implies numerical equivalence only for cycles on a fixed ambient variety, not between divisors on different fibers. In the trivialized situation X=X×_C S the conclusion B_s≡B is true and can be proved directly from the fact that each B_i is of the form W_i×S, but the argument as written is not valid. This lemma is used to prove numerical constancy of B, so it should be rewritten carefully.","section":"§3, Lemma 3.7"},{"comment":"The application of Alexeev's theorem [Ale94, Theorem 6.9] requires a fixed ϵ>0 such that all pairs (X_α,B_α) are MRϵ-klt. The manuscript only establishes that the coefficients B_α lie in a fixed finite set Γ and that each pair is klt; it does not prove a uniform lower bound for discrepancies on the minimal resolutions. For a fixed terminal 0-pair one might be able to derive such a bound, but the paper does not supply the argument. Without a uniform ϵ, the boundedness conclusion on the class D is not justified. Since this is the step that proves finite polarization in the Calabi-Yau case B≠0, this is a load-bearing point for Theorem 1.3.","section":"§5, Case 3.1"},{"comment":"The definition of boundedness of a class of varieties (resp. pairs) includes condition (2): every fiber of the parametrizing families must itself be isomorphic to an object of the class. The proofs of Lemma 2.14 and Proposition 4.1 only verify condition (1), and condition (2) generally fails for families obtained by closing a Chow locus: the closure contains fibers that are not in the original class. The later arguments only need the weaker statement that the class is contained in a finite union of families. The definitions and the statements of Lemma 2.14 and Proposition 4.1 should be adjusted accordingly, otherwise the terminology is misleading.","section":"Definitions 2.9–2.10, Lemma 2.14, Proposition 4.1"}],"minor_comments":[{"comment":"The symbol (X,B) is used both for the fixed log surface and for the total family (X/S,B). This makes statements such as \"−K_X≡B\" and \"K_X+B∼_R0\" ambiguous: sometimes they refer to the fiber, sometimes to the total space. Please introduce separate notation, e.g. (X_0,B_0) or (X/S,B).","section":"§3, setup"},{"comment":"The claim that (φ_α)_*(A_α) is ample for a general ample divisor A_α needs justification in the presence of a nontrivial exceptional locus; as written it is asserted rather than proved.","section":"§5, Case 1"},{"comment":"The sentence \"It can be proved that there is an isomorphism φ_α:C→C_α such that φ_α∘π=π_α∘f_α\" is stated without proof. This is a standard property of the Iitaka fibration, but since it is used to control the exceptional loci, a short proof or precise reference would be appropriate.","section":"§5, Case 2"},{"comment":"The Zariski decomposition is applied to −K_X via [Băd01, Theorem 14.14]. If this is applied to the fiber X rather than the total space, that should be said explicitly; the notation currently suggests the total space, which has dimension >2 in typical use.","section":"§3, Case 2"},{"comment":"There are several typographical and referencing issues, for example the numbering mismatch between \"Theorem 3.4\" and \"Theorem 3.3\" in the proof of Case 1, and the inconsistent use of \"Theorem 4.8\" for a lemma inside Theorem 4.4. These should be cleaned up.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper's advertised novelty, Theorem 4.4, depends on the isotriviality theorem for real coefficients, and that theorem is not proved as written. The abundance gap in §3, Case 2 is the main obstacle. The author should be asked to supply a rigorous abundance argument, possibly by invoking Fujino's theorem with explicit hypotheses, and to clarify the semiampleness of the positive part. The unproved uniform ϵ in the Alexeev application also needs attention. If these points can be repaired, the paper would make a useful contribution; at present it is not ready for acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Alex, quick read of Serebrennikov's arXiv:2510.14795. First thing you should know: the headline finiteness theorem (Theorem 5.1) is not new. The author says so himself in the last paragraph of §5, citing Totaro's cone conjecture and [CL14]. That honesty is typical: the paper is transparent about what is known and what is being attempted. The genuinely new content is Theorem 4.4, a reduction: if the underlying surfaces in a fixed 0-class have bounded polarizations, then there are finitely many wlc klt models. Plus an isotriviality theorem (3.2) for families of terminal 0-pairs with real-coefficient boundaries.\n\nWhat's good: the reduction-to-polarization idea is clean and could have legs in higher dimensions. The isotriviality result for R-boundaries is a real extension beyond the Q-coefficient case handled by Ambro/Xu; Birkar's decomposition trick to split B into Q-divisors is a sensible route. The proof of Theorem 4.4 — using the terminalization family, isotriviality, and a relative MMP with a fixed exceptional locus — is inventive. The author credits external heavy machinery rather than hand-waving.\n\nWhere it's soft. The proof of Theorem 3.2, Case 2, has a genuinely confusing step. The text reads 'By the semi-ampleness theorem [Fuj12, Theorem 8.1] we have ν(K_X+B)=κ(K_X+B), that is, K_X+B∼_R 0.' The 'that is' is wrong as written. Numerically trivial plus semiample does give R-linear triviality, and Fujino's theorem is the right tool, but the argument needs to say that, not just assert it. Without that, the isotriviality proof has a hole. It is easily repairable, but a referee has a right to insist. Also, Lemma 3.7's 'algebraic equivalence implies numerical equivalence' skips over the identification of cycles on different fibers. In the trivialized family it's fine, but the notation obscures the canonical identification; it needs a rewrite. The §5, Case 2 line 'It can be proved that...' (the fibration-preserving isomorphism in the κ=1 case) is a skipped proof of a standard fact; that should be one paragraph.\n\nBottom line: the paper is an honest, worthwhile attempt at a method. The known-result caveat is disclosed, the new theorems are plausible, and the gaps are fill-in rather than fatal. I'd send it to a serious referee, not desk reject. I wouldn't cite it in my own work until the revision appears, but I'd bring it to the reading group to see if the isotriviality proof survives scrutiny.","headline":"Main finiteness theorem is known and the author says so; new value is the reduction-to-polarization method and isotriviality for real boundaries, with a fixable gap in §3.","tokens_in":15718,"tokens_out":8239,"would_cite":false,"duration_ms":67242,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14E30","14J10","14J28"],"pacs":[],"model":"deepseek-v4-flash","headline":"Finiteness of log-surface models follows from bounded polarizations.","keywords":["log surfaces","log Calabi-Yau pairs","finiteness of models","bounded polarization","weakly log canonical models","isotrivial families","minimal model program","0-classes"],"falsifier":"Find a non-isotrivial smooth projective family of terminal log Calabi-Yau surfaces with real boundary coefficients and a dense set of closed points whose fibers are log-isomorphic to a fixed pair; this would directly contradict Theorem 3.2 and invalidate the main reduction. Alternatively, exhibit an R-divisor B on a smooth projective surface with K_X+B ≡ 0 for which the proof's inference from the equality of numerical and Kodaira dimension to K_X+B ∼_R 0 fails.","tokens_in":14585,"feed_emoji":"📐","tokens_out":9396,"duration_ms":71080,"temperature":0.7,"pith_summary":"The paper proves that a log surface—a projective surface equipped with a boundary divisor—has only finitely many weakly log canonical projective models with klt singularities up to log isomorphism, confirming a conjecture in dimension two. The key novelty is a reduction: finiteness of the models follows as soon as the underlying varieties have bounded polarization, meaning a uniform bound on the degree of a very ample divisor and on the boundary coefficients. The bulk of the proof establishes that families of terminal log Calabi-Yau surfaces with a dense set of isomorphic fibers become trivial after an étale base change; combined with the minimal model program over the base, this forces all models to agree on a dense open set, and a standard induction argument yields finiteness. The author notes that the finiteness statement itself was already known to experts via the cone conjecture; the contribution is the bounded-polarization method, which is intended to generalize to higher dimensions.","feed_headline":"Bounded polarizations make log-surface models finite","feed_subtitle":"A new proof reduces finiteness of log surface models to one boundedness condition.","key_machinery":"The carrying mechanism is an isotriviality theorem for families of terminal 0-pairs: a smooth projective elementary family of terminal log Calabi-Yau surfaces whose isomodular locus is dense becomes log-isomorphic to a trivial family after an étale base change. This is combined with the log MMP run over the base: the MMP contracts the exceptional divisors of the terminalization, and a lemma shows each step preserves the triviality of the family, so the end product is a single model with the same isomorphism type as every fiber over a dense open set.","core_discovery":"The central claim is Theorem 1.4: for a fixed birational class of log surfaces (technically, a 0-class under crepant birational equivalence), if the set D of underlying varieties of all weakly log canonical klt models has bounded polarization, then the set C of models is finite up to log isomorphism. The proof embeds C into finitely many elementary projective families using bounded polarization, then shows that the terminalizations of the models in each family all become log-isomorphic to one fixed log surface after an étale base change; as a result the models themselves are also isomorphic on a dense open set, and a standard induction argument gives finiteness. The author states that the pr","pith_inferences":["If the suspected gap in the real-coefficient isotriviality proof is patched, the same strategy could extend to higher-dimensional klt Calabi-Yau pairs whose models have bounded polarization, sidestepping the cone conjecture.","The bounded-polarization hypothesis may be the right uniform parameter to isolate, in any dimension, the 'finite-model' case of the finiteness conjecture; one could test it on known families of elliptic or fibred Calabi-Yau threefolds.","The paper's separation of the known finiteness result from its new method suggests that future work should be judged by whether the bounded-polarization reduction can be made to work independently of the cone conjecture in higher dimensions.","A natural test case: families of log surfaces with real boundary coefficients whose isomodular locus is dense but where the boundary has irrational slopes; Theorem 3.2 claims these are still isotrivial, which would be a surprising rigidity if true."],"forward_implications":["If the bounded-polarization reduction is correct, finiteness of weakly log canonical klt models for log surfaces follows without invoking the cone conjecture.","The same reduction offers a template for proving finiteness of minimal models in higher dimensions: establish bounded polarization, then prove an analogous isotriviality statement.","For the K3 case (B=0), the method suggests a purely algebraic route to finiteness, replacing the geometric cone-theorem input with a bounded-polarization argument.","The isotriviality theorem implies any family of terminal log Calabi-Yau surfaces carrying a dense set of isomorphic fibers must be étale-locally trivial, a rigidity property of such families."],"fun_headline_variants":["Log surface models finite under bounded polarizations","Finiteness theorem for log surfaces proven","Bounded polarizations curb log surface variety","Only finitely many log canonical models exist","Klt log surfaces admit finitely many models"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof's load-bearing premise is that every smooth projective family of terminal log Calabi-Yau surfaces with real-coefficient boundaries and a dense set of isomorphic fibers becomes trivial after an étale base change; if that isotriviality theorem fails, the reduction to bounded polarization collapses.","fun_headline_variants_meta":{"raw":{"variants":["Log surface models finite under bounded polarizations","Finiteness theorem for log surfaces proven","Bounded polarizations curb log surface variety","Only finitely many log canonical models exist","Klt log surfaces admit finitely many models"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000537,"raw_usage":{"total_tokens":2317,"prompt_tokens":547,"completion_tokens":1770,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":291,"completion_tokens_details":{"reasoning_tokens":1703}},"tokens_in":291,"tokens_out":1770,"duration_ms":11338,"temperature":1.0,"reasoning_tokens":1703,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T09:30:46.947098+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a non-isotrivial smooth projective family of terminal log Calabi-Yau surfaces with real boundary coefficients and a dense set of closed points whose fibers are log-isomorphic to a fixed pair; this would directly contradict Theorem 3.2 and invalidate the main reduction. Alternatively, exhibit an R-divisor B on a smooth projective surface with K_X+B ≡ 0 for which the proof's inference from the equality of numerical and Kodaira dimension to K_X+B ∼_R 0 fails.","supporting_citations":[],"review_version":1}