{"id":"ad002555-0c70-42c7-9f28-e9aa5b31834c","arxiv_id":"2608.07836","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Batyrev's non-negativity conjecture on stringy Hodge numbers is true for Gorenstein canonical projective varieties in dimension at most 4 and false in all dimensions 5 and higher.","lead":"This paper settles the exact dimension threshold for Batyrev's non-negativity conjecture on stringy Hodge numbers: the conjecture holds in dimension at most 4 and fails in every dimension at least 5. It is a complete answer to a long-standing guiding question in mirror symmetry, motivic integration, and the McKay correspondence.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The dimension-4 half of the threshold depends on an unverified small Q-factorialization theorem ([LM26, Cor. 22.3]) in Proposition 5.1; unless it is checked, Theorem A is conditional.","rationale":"The paper makes a sharp threshold claim. The counterexample direction is explicit, parameter-free, and internally consistent; my own recomputation of E_st(X_n) confirms the u^2v^3 coefficient is -1. The remaining structural risk is Theorem A. The proof of Theorem 1.2 reduces to Proposition 5.1, and Proposition 5.1 depends on a deep local Q-factorialization statement quoted from a preprint without proof. The reader identified the same concern, and the paper itself signals the reliance in Remark 5.2. This is a load-bearing external input: without it, the inequality delta_alpha <= n_alpha is not established, so h^{2,2}_st >= 0 for fourfolds is not established. I therefore recommend conditional acceptance: the result should be accepted once [LM26, Corollary 22.3] is verified to apply to the completed local rings in question, or an independent proof of the small Q-factorialization is supplied. This is not a rejection of the mathematical claims, which may well be correct; it is a request to secure the one genuinely fragile link.","tokens_in":23180,"tokens_out":21108,"duration_ms":198663,"concrete_test":"Verify [LM26, Corollary 22.3] (arXiv:2209.08732) in the precise situation used: bS_alpha = Spec of \\widehat{O}_{Y,eta_alpha}, a 3-dimensional complete excellent normal Gorenstein terminal local C-domain. In particular confirm that all hypotheses of the corollary hold (excellent, equal characteristic zero, dualizing complex, relative MMP setup) and that the statement gives a projective Q-factorialization that is small, i.e. has no exceptional divisors, with connected fibers. If the citation checks out, recompute Proposition 5.1 from that statement. If it does not, attempt to construct the small Q-factorialization from the resolution bX_alpha of Lemma 5.3 by running the relative MMP for threefolds over bS_alpha, and test the resulting inequality delta_alpha <= n_alpha on explicit terminal Gorenstein fourfolds.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Theorem B is self-contained, and I recomputed its central numbers: the stringy E-function computation for X_n gives coefficient of u^2v^3 equal to -1, so the dimension at least 5 counterexamples are solid. The load-bearing risk is in the fourfold proof. Theorem 1.2 is proven by bounding the defect: h^{2,2}_st(Y) = I_Y + n + epsilon - sum_alpha delta_alpha >= I_Y + epsilon, where the key inequality is delta_alpha <= n_alpha in Proposition 5.1. The proof of that proposition, Step 1, invokes [LM26, Corollary 22.3] to obtain a projective Q-factorialization tau_alpha: V_alpha -> Spec of the completed local ring at the generic point of C_alpha, with no exceptional divisors, for a 3-dimensional complete excellent normal Gorenstein terminal local domain. This is not a step the authors prove or even state precisely; the entire later chain — identification of Cl tensor Q with N^1, the bound by the number of contracted curves, and the matching of those curves to discrepancy-1 divisors — depends on the existence and smallness of this Q-factorialization. If the cited corollary has unverified hypotheses or an error, delta_alpha <= n_alpha is unsupported and the fourfold theorem has no proof. The counterexamples in Theorem B do not supply evidence for Theorem A.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper determines the exact dimension threshold for Batyrev's non-negativity conjecture on stringy Hodge numbers for complex projective varieties with Gorenstein canonical singularities. Theorem A asserts non-negativity of all stringy Hodge numbers in dimension at most 4, with the fourfold case reduced to the non-negativity of h^{2,2}_{st}(Y); this is Theorem 1.2. Theorem B constructs, for every n≥1, an (n+4)-dimensional projective variety X_n = Y×(P^1)^n with Gorenstein terminal singularities, polynomial stringy E-function, and h^{2,3}_{st}(X_n)=-1, so the conjecture fails in every dimension at least 5. The core technical work is a formula for h^{2,2}_{st}(Y) in Section 4 expressing it as I_Y+n+ε−Σα δα and a bound δα≤nα in Proposition 5.1, obtained by comparing monodromy invariants of a local defect space with the divisor class group of a completed local ring.","tokens_in":23460,"tokens_out":6705,"duration_ms":59543,"significance":"If the proof is complete, this is a definitive and striking answer to a long-standing conjecture: a sharp dimension threshold of 4, with consistent positive results in low dimensions and explicit negative examples in all higher dimensions. The counterexample family in Theorem B is a clear strength: the computation is fully explicit, the resolutions and discrepancies are written down, and the offending coefficient of u^2v^3 is concretely computed; I independently checked that this coefficient is -1. The fourfold argument is also structurally appealing: it avoids fitted parameters and post hoc selections, and the main identity in Proposition 4.6 is derived from the Decomposition Theorem and local-system monodromy rather than from the statement being proved. The principal caveat is that the fourfold result depends on a specific external theorem, [LM26, Corollary 22.3], whose status is discussed below; the counterexample in Theorem B is independent of that input.","major_comments":[{"comment":"The proof of the key inequality δα≤nα invokes [LM26, Corollary 22.3] to obtain a projective Q-factorialization τα: Vα→bSα with no exceptional divisors for the three-dimensional complete excellent normal Gorenstein terminal local domain bOα. This is a load-bearing step: the identification of Cl(bOα)⊗Q with N^1(Vα/bSα)Q, the counting of Kα-irreducible contracted curves, and the final matching of those curves to discrepancy-1 divisors all use the existence and smallness of this Q-factorialization. The manuscript neither states the precise hypotheses of the cited corollary nor proves it, and [LM26] is a 2026 arXiv preprint. If that corollary is false or its hypotheses are not satisfied, Theorem 1.2 and hence the fourfold half of Theorem A have no proof. Please provide a proof of the needed statement, state its hypotheses explicitly and verify them for bOα, or explicitly mark Theorem A and Theorem 1.2 as conditional on [LM26].","section":"§5, Proposition 5.1, Step 1"},{"comment":"The formula for h^{2,2}_{st}(Y) is assembled by substituting identities into [Ola21, Remark 1.7] and by importing several decomposition-theorem formulas from [Ola21, Lemmas 9.1, 9.3, 9.5, 9.9, 10.2]. Because Remark 4.5 explicitly claims that the new identity is obtained without imposing Condition (∗) from [Ola21], the paper should itemize which of the imported results are valid without (∗) and confirm that each is applied under the hypotheses stated there. As written, a reader cannot verify Proposition 4.6 without a lemma-by-lemma check of [Ola21] against the present notation, and any hidden dependence on (∗) would affect the central fourfold claim.","section":"§4.3, Proposition 4.6"}],"minor_comments":[{"comment":"The sentence 'Since C has degree 4, c has degree 3, and their intersection is transverse, Z is equal to 12 points' should read 'Z consists of 12 points'.","section":"§2, proof of Theorem B"},{"comment":"The abstract and Theorem A state the dimension-4 result unconditionally, but the proof depends on [LM26, Corollary 22.3]; if the conditional status is retained, the wording should be adjusted so readers are not misled about the external input.","section":"Abstract and Theorem A"},{"comment":"In the formula Est(Y)=1+uv+13(uv)^2+(uv)^3+(uv)^4+u^3v^2+u^2v^3/(1+uv), the last term should be parenthesized as (u^3v^2+u^2v^3)/(1+uv) to avoid ambiguity.","section":"§2, displayed formula for Est(Y)"},{"comment":"Reference [W lo16] is listed as an arXiv preprint from 2016; if a published version exists, it should be cited instead, especially since Proposition 3.1 relies on its principalization theorem.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the external dependence on [LM26, Corollary 22.3]. I would ask the authors to include a proof or a detailed verification of that corollary, or to coordinate with Lyu and Murayama about its status. If that cannot be done, Theorem A should be stated conditionally and the abstract adjusted. Theorem B is self-contained and would stand as a valid paper on its own. I recommend major revision rather than rejection because the issue is identifiable and fixable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper answers the exact question Batyrev's non-negativity conjecture has been waiting for: the threshold is dimension 4. Theorem B gives explicit counterexamples in every dimension at least 5, and that half is self-contained and correct. I rechecked the stringy E-function of X_n and the u^2v^3 coefficient is indeed -1. The geometry is clean: a cubic hypersurface in P^5 singular along a line, resolved by blowing up the line, producing a genus-one exceptional curve, then multiplying by P^1s to clear denominators. That part deserves strong credit.\n\nThe fourfold half, Theorem A, is a long, structured argument building on Olano's work and a defect bound. The chain from the adapted resolution to the formula for h^{2,2} to the defect estimate is coherent, and I found no internal gaps in the sections I checked. Where the paper is genuinely soft is Proposition 5.1, Step 1: the bound delta_alpha <= n_alpha depends on the existence of a small Q-factorialization of the completed local ring, cited from [LM26, Cor. 22.3], a 2026 arXiv preprint. The authors apply that corollary to a three-dimensional complete excellent normal Gorenstein terminal local domain. They do not re-prove it, and they do not spell out the full list of hypotheses they verify. If that corollary has an error or an unstated hypothesis, the fourfold theorem loses its proof. The counterexample half is unaffected.\n\nIs this a flaw? Not exactly—relying on a recent preprint is normal, and the result may well be true. But it is load-bearing, and the community should know where the weight sits. I would want the authors to include the precise statement and verification of the hypotheses, or a short proof, before the fourfold claim is treated as settled.\n\nOverall: this is a significant result, honestly argued, with the citation pattern legitimate. The paper deserves a serious referee, and I would take it to reading group. My recommendation: send to peer review, with explicit instructions to check the LM26 dependence.","headline":"Batyrev's threshold is exactly dimension 4/5, but the fourfold half leans on one unverified external Q-factorialization theorem that needs scrutiny.","tokens_in":23989,"tokens_out":1984,"would_cite":true,"duration_ms":18387,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14E18","14B05","14J33","14C30"],"pacs":[],"model":"deepseek-v4-flash","headline":"Batyrev's stringy Hodge-number conjecture is decided: true through dimension 4, false at 5.","keywords":["Batyrev conjecture","stringy Hodge numbers","stringy E-function","Gorenstein canonical singularities","terminal singularities","dimension threshold","mirror symmetry","motivic integration"],"falsifier":"Exhibit a three-dimensional complete terminal Gorenstein local singularity whose divisor class group has rank larger than the number of discrepancy-$1$ divisors appearing on any resolution; that would break the bound $\\delta_\\alpha\\le n_\\alpha$ and with it the fourfold theorem. For the counterexample half, recompute the coefficient of $u^2v^3$ in $E_{st}(Y\\times\\mathbb{P}^1)$ for the paper's explicit cubic $Y$ and check that it equals $-1$.","tokens_in":22975,"feed_emoji":"📐","tokens_out":17232,"duration_ms":128760,"temperature":0.7,"pith_summary":"Stringy Hodge numbers are invariants that attach Hodge-theoretic information to singular varieties, computed from a log resolution and discrepancy data, and Batyrev conjectured that they are all nonnegative whenever the defining stringy $E$-function is a polynomial. The paper determines exactly when that conjecture holds: it is true for every complex projective variety with Gorenstein canonical singularities of dimension at most four, and it fails in every dimension at least five. On the positive side, the fourfold proof reduces the conjecture to the single middle coefficient $h^{2,2}_{st}$ and shows that any negative contribution is controlled by the number of discrepancy-$1$ divisors, forcing $h^{2,2}_{st}\\ge 0$. On the negative side, the counterexamples are explicit varieties $X_n = Y\\times(\\mathbb{P}^1)^n$, where $Y$ is a cubic hypersurface in $\\mathbb{P}^5$ with a singular line and $n\\ge 1$, and each $X_n$ has polynomial stringy $E$-function but $h^{2,3}_{st}(X_n)=-1$.","feed_headline":"Stringy Hodge conjecture holds to dimension 4, fails at 5","feed_subtitle":"Nonnegativity is proved for all fourfold singularities, with explicit counterexamples in every dimension above.","key_machinery":"The argument runs on two distinct mechanisms. For the fourfold theorem, the paper builds an adapted log resolution $f:X\\to Y$ whose exceptional divisors over the one-dimensional singular locus are separated into horizontal components (dominating singular curves) and vertical components (mapping to points). For each singular curve $C_\\alpha$, a local defect space $\\Delta_{\\alpha,x}$, canonically identified with $H^2(L_x,\\mathbb{Q})$ for the link $L_x$ of a general hyperplane slice, measures possible negative contributions to $h^{2,2}_{st}$. The key identity is $h^{2,2}_{st}(Y)=I_Y+n+\\epsilon-\\sum_\\alpha\\delta_\\alpha$, and the central bound $\\delta_\\alpha\\le n_\\alpha$ compares the monodromy invariants of the defect with the number of discrepancy-$1$ divisors over $C_\\alpha$; the bound is proved by translating $\\delta_\\alpha$ into the rank of the divisor class group of the completed local ring at the generic point of $C_\\alpha$ and using a small projective $\\mathbb{Q}$-factorialization. For the counterexamples, the mechanism is a single genus-one curve: blowing up the singular line of the cubic $Y$ in $\\mathbb{P}^5$ leaves one exceptional divisor $D$, the blow-up of $\\mathbb{P}^3$ along the genus-one curve $C=V(q_0,q_1)$, and the off-diagonal terms $u^3v^2$ and $u^2v^3$ survive multiplication by $(\\mathbb{P}^1)^n$ to give $h^{2,3}_{st}(X_n)=-1$.","core_discovery":"The paper proves an exact threshold. Theorem A: if $Y$ is a complex projective variety with at worst Gorenstein canonical singularities, $\\dim Y\\le 4$, and $E_{st}(Y;u,v)$ is a polynomial, then every stringy Hodge number $h^{p,q}_{st}(Y)$ is nonnegative. The only coefficient not already handled in the fourfold case is the middle one, and the paper proves $h^{2,2}_{st}(Y)\\ge 0$ even when $E_{st}(Y;u,v)$ is not a polynomial. Theorem B: for every $n\\ge 1$, the $(n+4)$-dimensional variety $X_n=Y\\times(\\mathbb{P}^1)^n$, where $Y=\\{s q_0(x)+t q_1(x)+c(x)=0\\}\\subset\\mathbb{P}^5$ for general quadratic forms $q_0,q_1$ and a general cubic $c$, has Gorenstein terminal singularities and polynomial stringy $E$-function, yet $h^{2,3}_{st}(X_n)=-1$. Hence the conjecture is true in dimension at most 4 and false in every dimension at least 5.","pith_inferences":["The paper leaves implicit that the terminal counterexamples also test the cohomological reformulations discussed in its introduction: if stringy Hodge numbers were actually dimensions of cohomology groups, nonnegativity would be automatic, so the negative coefficient points toward a failure of purity rather than a failure of cohomological representability.","Although not drawn out, the product construction shows that nonnegativity is not preserved under products with $\\mathbb{P}^1$: $E_{st}(Y\\times\\mathbb{P}^1)$ can become a polynomial while $E_{st}(Y)$ remains rational, so verifying the conjecture on low-dimensional bases cannot be bootstrapped to products.","A direct next computation would be the full signed Hodge diamond of $X_1=Y\\times\\mathbb{P}^1$; the paper displays the polynomial $E_{st}(X_1)$ explicitly, so checking whether $-1$ is the only negative coefficient is an immediate calculation."],"forward_implications":["All projective Gorenstein canonical fourfolds with polynomial stringy $E$-function satisfy the conjecture, and their $u^2v^2$ coefficient is nonnegative even without the polynomiality assumption.","The counterexamples are Gorenstein terminal, so failure of nonnegativity is not produced by particularly severe singularities.","Every dimension at least 5 contains a counterexample, ruling out any higher-dimensional regime where the conjecture could resume holding.","The known three-dimensional case, the classical two-dimensional case, and the new fourfold theorem together identify dimension 4 as the last dimension where the conjecture survives."],"supporting_citations":[{"why":"This is the foundational reference: it defines the stringy $E$-function and stringy Hodge numbers, states the non-negativity conjecture, and provides the invariance and duality results used throughout.","marker":"[Bat98]"},{"why":"This supplies the reduction of the fourfold case to $h^{2,2}_{st}$, the nonnegativity theorem for $h^{p,1}_{st}$, and the intersection-formula framework that the paper adapts.","marker":"[Ola21]"},{"why":"This establishes the three-dimensional case of the conjecture, which Theorem A uses as its low-dimensional input.","marker":"[SV07]"},{"why":"This gives the crepant modification to terminal singularities that lets the fourfold argument assume terminality.","marker":"[BCHM10]"},{"why":"This supplies standard facts about terminal and rational singularities and discrepancies that the local analysis uses.","marker":"[KM98]"},{"why":"This contains the load-bearing corollary on the existence of a small projective $\\mathbb{Q}$-factorialization for three-dimensional complete terminal Gorenstein local domains, used in Proposition 5.1.","marker":"[LM26]"},{"why":"This relates divisor class groups of local rings and their strict henselizations to the monodromy invariants of the defect space.","marker":"[BF84]"},{"why":"This provides the exact sequences identifying the local defect space with link cohomology and fixing its Hodge type.","marker":"[FL24]"}],"fun_headline_variants":["Batyrev's nonnegativity: true in dim ≤4, false in dim ≥5","Exact cutoff: stringy Hodge numbers stay nonnegative up to dimension 4","Stringy Hodge numbers turn negative in every dimension above 4","Dimension threshold proven: Batyrev holds at 4, fails at 5","Counterexamples in all dimensions ≥5 to Batyrev's nonnegativity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The fourfold half of the theorem depends on a cited theorem, not re-proved here, that every three-dimensional complete terminal Gorenstein local singularity admits a projective resolution of its divisor class group with no exceptional divisors; if that theorem fails, the dimension-four proof collapses.","fun_headline_variants_meta":{"raw":{"variants":["Batyrev's nonnegativity: true in dim ≤4, false in dim ≥5","Exact cutoff: stringy Hodge numbers stay nonnegative up to dimension 4","Stringy Hodge numbers turn negative in every dimension above 4","Dimension threshold proven: Batyrev holds at 4, fails at 5","Counterexamples in all dimensions ≥5 to Batyrev's nonnegativity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000393,"raw_usage":{"total_tokens":2007,"prompt_tokens":827,"completion_tokens":1180,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":443,"completion_tokens_details":{"reasoning_tokens":1074}},"tokens_in":443,"tokens_out":1180,"duration_ms":9098,"temperature":1.0,"reasoning_tokens":1074,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:26:59.729905+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a three-dimensional complete terminal Gorenstein local singularity whose divisor class group has rank larger than the number of discrepancy-$1$ divisors appearing on any resolution; that would break the bound $\\delta_\\alpha\\le n_\\alpha$ and with it the fourfold theorem. For the counterexample half, recompute the coefficient of $u^2v^3$ in $E_{st}(Y\\times\\mathbb{P}^1)$ for the paper's explicit cubic $Y$ and check that it equals $-1$.","supporting_citations":[],"review_version":2}