{"id":"372bfcd4-a0db-49d2-b077-36116e724489","arxiv_id":"2607.19251","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Kappa classes on KSBA moduli stacks are well-defined operational cohomology classes that detect the variation of a family and are Chern classes of virtual vector bundles.","lead":"Kappa classes on moduli spaces of stable pairs (KSBA spaces) are defined as operational Chow cohomology classes, generalizing the Miller–Morita–Mumford classes on curves. The paper proves their functoriality, positivity, vanishing, wall-crossing, and Chern-class structure.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the only non-self-certifying step is the [Kol23] base-change compatibility of the line bundle used to define kappa classes.","rationale":"I read the paper as a theorem-heavy companion that carefully upgrades cycle-level kappa classes to operational Chow cohomology. The main claims—well-definedness, vanishing above the variation, nonnegativity, numerical detection of normalized variation, and the virtual Chern class formula—are supported by detailed proofs that I checked for internal consistency. The normalization additivity, crepant functoriality, and Riemann–Roch arguments are structurally sound: the degree-one universal condition in Lemma 3.7 is plausible, the reduction to integral cycles in Theorem 3.8 is legitimate, and Theorem 4.2's use of singular Riemann–Roch is standard. The weakest point is exactly the one the reader identified: the f-ample Q-line bundle O_X(K_{X/M}+D) must be compatible with arbitrary base change, including non-reduced bases that arise in the operational framework. This is imported from [Kol23] and is genuinely load-bearing, but I found no reason internal to the paper to doubt it. Therefore, while this dependency warrants a check, it does not change the ACCEPT verdict.","tokens_in":21867,"tokens_out":28664,"duration_ms":327712,"concrete_test":"Verify in [Kol23, Section 8] that the Kollár condition indeed yields the base-change isomorphism g'^*L_{X/S} ≅ L_{X'/S'} for every morphism S'→S, in particular for S' = Spec k[ε]/ε^2 and for the normalization S^ν→S of a non-normal base. If the isomorphism fails for any of these, Definition 2.2 and the normalization/wall-crossing theorems break; if it holds, the central construction is sound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The paper is internally coherent as far as I can check. The most load-bearing step is Definition 2.2's dependence on Section 2.1, where the reflexive log-pluricanonical sheaf L_{X/S} is asserted to be an f-ample line bundle compatible with arbitrary base change, imported from [Kol23, Section 8]. Because operational Chow classes are tested after arbitrary—and often non-reduced—base changes, a failure of the isomorphism g'^*L_{X/S} ≅ L_{X'/S'} would make κ_r ill-defined and would invalidate Proposition 3.9 (normalization additivity) and Theorem 3.27 (wall-crossing compatibility), both of which rely on Theorem 3.8(2). I have no internal evidence that this cited condition fails, and the surrounding arguments are detailed and consistent, so I do not elevate this dependency to a demonstrated flaw.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces kappa classes κ_r on KSBA moduli stacks as elements of operational Chow cohomology, generalizing the Miller–Morita–Mumford classes. The definition uses a Q-line bundle O_X(K_{X/M}+D) obtained from a reflexive log-pluricanonical sheaf, and the main body establishes base-change compatibility, functoriality, product and normalization formulas, crepant functoriality, vanishing of kappa polynomials above the variation, nonnegativity and numerical detection of the normalized variation, chamberwise polynomiality and wall-crossing compatibility, and a Riemann–Roch formula expressing each κ_r for r≥1 as a rational multiple of a single Chern class of a virtual vector bundle.","tokens_in":22036,"tokens_out":25145,"duration_ms":248370,"significance":"This is a well-written and substantial contribution. It provides the first systematic treatment of kappa classes on singular moduli stacks and shows that they form a finite-dimensional commutative ring that encodes variation invariants. The explicit Chern-class formula in Corollary 4.11 is a strong computational tool, and the vanishing, nonnegativity, and wall-crossing results are natural and are proved in detail. The paper is transparent about its external inputs: the foundational base-change compatibility is imported from [Kol23], and positivity/wall-crossing results from [PX17] and [MZ23] are cited explicitly. The proofs are detailed enough for a careful reader to follow.","major_comments":[],"minor_comments":[{"comment":"The well-definedness of κ_r as an operational class rests on the assertion that L_{X/S} is an f-ample line bundle compatible with arbitrary base change, imported from [Kol23, Section 8]. Since operational Chow classes are tested on arbitrary—often non-reduced—base schemes, the paper should state the precise theorem from [Kol23] and explain why it applies to the universal family over the stack, including non-reduced bases. This is not a demonstrated error, but a more precise citation would remove ambiguity.","section":"Section 2.1 / Definition 2.2"},{"comment":"The notation for f^# and f_# mixes Chow homology and Chow cohomology. As written, f^#: A^k(M) → A^{k+n}(X) should be on Chow homology groups A_k(M) → A_{k+n}(X), and f_# should use the cohomology comparison isomorphisms (π^*)^{-1} and (π'^*)^{-1} rather than π_*^{-1} and π'_*. The subsequent computations in Lemma 3.23 and formula (5) are consistent with the intended definitions, but the statement of the definition needs clarification.","section":"Section 3.9, Definition 3.22"},{"comment":"In the proof that ρ^{-1}(U) ≅ U, the birationality of the ample-model contraction on every irreducible component of the target is cited to [MZ23, proof of Lemma 4.9]. A precise reference to the lemma and a brief explanation of why it applies to the reduced closures M_i and M_j would improve readability.","section":"Section 3.10, Theorem 3.27"},{"comment":"The typeset title/abstract contains apparent line-break artifacts such as 'PROPER TIES' and 'KAPP A'. Please ensure the final version has correct spacing. Also, the Introduction has a minor punctuation issue: 'the rational coefficientsa= (a 1, . . . , aq)' should read 'the rational coefficients a = (a_1, . . . , a_q)'.","section":"Throughout"}],"recommendation":"minor_revision","confidential_remarks":"This is a strong manuscript with sound central claims. The only reservations are expository: the foundational base-change compatibility from [Kol23] should be stated more precisely, and the notation in Section 3.9 needs cleaning up. No concerns about novelty, attribution, or correctness of the main theorems."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line first: this is a solid, carefully written companion paper that does what the field needs—puts kappa classes on KSBA stacks into operational Chow cohomology and proves the expected properties: base change, functoriality, normalization additivity, vanishing above variation, nonnegativity, wall crossing, and a virtual Chern-class formula. The main theorems are genuinely new relative to [Ale25], which only had cycle-level classes; the operational setup is the real upgrade.\n\nWhat I like: the paper is honest about what it imports. The authors explicitly flag the dependence on [Kol23, Section 8] for the f-ample line bundle and on [MZ23] for wall crossing. The proof of Theorem 4.2 is detailed, the induction is clear, and the Chern-class presentation of kappa classes in Section 4 is a useful piece of formal Riemann–Roch work. I could not find internal incoherence or hidden fitting. The acknowledgement about ChatGPT is fine; the references are independently cited and the mathematical arguments stand on their own.\n\nThe soft spot is exactly where the reader put it: Definition 2.2 rests on the assertion that the reflexive log-pluricanonical sheaf is a line bundle compatible with arbitrary base change, imported from [Kol23]. Operational Chow classes are tested on non-reduced base changes, so if that isomorphism fails, kappa classes are not well defined and Proposition 3.9 and Theorem 3.27 break. The paper gives no proof of that compatibility; it cites. I have no independent certification either, but this is a citation dependency, not a flaw I can point at in the text—the surrounding arguments are consistent and the source is precisely identified. The [MZ23] dependence for wall crossing is similar: if those theorems hold, Theorem 3.27 follows.\n\nFor whom: anyone working on KSBA moduli, variation, or tautological rings of higher-dimensional moduli stacks. It deserves a serious referee. The right referees are people who can check the [Kol23] base-change question and the applicability of [MZ23]. I would send it to peer review rather than desk reject.","headline":"A solid companion paper that puts kappa classes on KSBA stacks into operational Chow cohomology and proves the expected structural properties; the main risk is the imported base-change compatibility from [Kol23], which is cited rather than proved.","tokens_in":22555,"tokens_out":2684,"would_cite":true,"duration_ms":41023,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14C17","14C40","14D23","14J10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Kappa classes extend to KSBA moduli stacks, detect variation, and reduce to Chern classes","keywords":["kappa classes","KSBA moduli","operational Chow cohomology","MMM classes","variation","wall crossing","Chern classes","stable pairs"],"falsifier":"Take a KSBA family over a non-normal base whose normalization has components with different variations, and compute whether κ_r is numerically nonzero for r strictly between the normalized variation and the total variation. Theorem 3.18 predicts κ_r ≡ 0 for r > normalized variation and non-zero for r ≤ normalized variation; any counterexample would falsify the numerical-triviality criterion. Alternatively, compute the virtual bundle E_{1,m} for a low-genus moduli space of stable curves and check whether the formula κ_1 = c_1(E_{1,m})/(N^{n+1}) reproduces the standard λ_CM class; a mismatch wou","tokens_in":21713,"feed_emoji":"🧮","tokens_out":5198,"duration_ms":51399,"temperature":0.7,"pith_summary":"This paper establishes that kappa classes can be defined as operational Chow cohomology classes on KSBA moduli stacks—the higher-dimensional analogues of the moduli spaces of stable curves—and that they behave like their curve-theoretic ancestors. The central results are: kappa classes are compatible with base change, products, normalization, and crepant maps; every homogeneous polynomial in them of codimension greater than the variation of the family vanishes; κ₁ detects the total variation, while the largest index r with κ_r numerically nonzero equals the normalized variation; and for r ≥ 1, κ_r is, up to a rational factor, the r-th Chern class of a single virtual vector bundle built from pushforwards of powers of the relative canonical bundle. If correct, these classes give a finite-dimensional commutative ring of invariants that encodes geometric variation and is concretely computable.","feed_headline":"Kappa classes detect variation on moduli of stable pairs","feed_subtitle":"They vanish above the variation, wall-cross compatibly, and reduce to single Chern classes.","key_machinery":"The argument runs through operational Chow cohomology on Deligne–Mumford stacks: for a flat proper family of relative dimension n, the Gysin pushforward f^!(c) = f_*(c · [f]) turns a degree n+r class on the total space into an operational class of codimension r on the base. The needed Q-line bundle Λ = O_X(K_{X/M}+D) is obtained as (1/N) of an f-ample line bundle L = i_* ω^{⊗N}_{U/S}(ND|_U), whose base-change compatibility is imported from Kollár's boundedness results. The Chern-class formulas come from Grothendieck–Riemann–Roch for singular varieties (Baum–Fulton–MacPherson) combined with Newton identities: if the lower Chern characters of a perfect complex vanish, the r-th Chern class is (","core_discovery":"Definition 2.2 sets κ_r = f^! c_1(O_X(K_{X/M}+D))^{r+n}, where f is the universal KSBA family of relative dimension n and f^! is the Gysin pushforward in operational Chow cohomology; here O_X(K_{X/M}+D) is a Q-line bundle on the total space coming from an f-ample reflexive log-pluricanonical line bundle. The paper proves this class is a well-defined operational class, and derives a suite of structural properties: base-change functoriality, a product formula with multiplicativity of the kappa series, additivity under normalization, descent to the coarse moduli space, vanishing of all kappa polynomials above the variation (Theorem 3.12), nonnegativity on effective cycles with strict positivity","pith_inferences":["If the construction of K_{X/M}+D can be extended beyond Kollár's setting—for example to moduli of pairs with worse-than-slc singularities or to non-flat universal families—the same operational framework would presumably define kappa classes there, provided a suitable Q-line bundle with base-change compatibility exists.","The Chern-class expression suggests an effective computational route for concrete moduli spaces: compute the Chern character of the pushforwards V_k, take finite differences, and extract κ_r; for moduli of surfaces or threefolds this may yield explicit intersection numbers that are otherwise hard to access.","The wall-crossing compatibility may give a way to transport kappa classes across different stability chambers, relating invariants of different GIT or log canonical models within one birational family.","The nonnegativity statement suggests that kappa classes could define a nef cone on KSBA moduli, and the Khovanskii–Teissier-style inequalities noted in Remark 3.21 hint at log-concavity properties of the sequences s_k(H), which may be testable in low-dimensional examples."],"forward_implications":["For any KSBA family over an integral base, every monomial in kappa classes of total codimension greater than the variation vanishes, so the kappa ring is finite-dimensional and nilpotent with index equal to variation+1.","κ₁ equals the first Chern class of the logarithmic CM line bundle, which is semiample and has Iitaka dimension equal to the total variation; hence the first kappa class alone measures total variation.","The largest index r with κ_r numerically non-zero is the normalized variation, so knowledge of the kappa classes determines whether any normalization component varies independently.","Each positive-degree κ_r is a rational multiple of a single Chern class of the virtual bundle E_{r,m}; in particular the kappa classes are central in operational cohomology and the kappa ring is commutative.","As boundary coefficients vary in an admissible polytope, the kappa classes are chamberwise polynomial in the coefficients, and their numerical pairings agree on faces via the wall-crossing maps."],"fun_headline_variants":["Kappa classes generalize MMM classes on KSBA moduli","Kappa classes vanish above the variation on moduli","Kappa classes detect variation, nonnegative on cycles","Kappa classes: wall-crossing, single Chern class multiples","New kappa classes on stable pairs: key properties"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The definition of kappa classes rests on the assertion that the reflexive log-pluricanonical sheaf L_{X/S} is an f-ample line bundle compatible with arbitrary base change, so that a Q-line bundle K_{X/M}+D exists on the universal family; if that base-change compatibility fails for the non-normal bases used in later arguments, the classes κ_r and the normalization additivity would collapse.","fun_headline_variants_meta":{"raw":{"variants":["Kappa classes generalize MMM classes on KSBA moduli","Kappa classes vanish above the variation on moduli","Kappa classes detect variation, nonnegative on cycles","Kappa classes: wall-crossing, single Chern class multiples","New kappa classes on stable pairs: key properties"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000189,"raw_usage":{"total_tokens":1137,"prompt_tokens":672,"completion_tokens":465,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":416,"completion_tokens_details":{"reasoning_tokens":385}},"tokens_in":416,"tokens_out":465,"duration_ms":5379,"temperature":1.0,"reasoning_tokens":385,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T12:58:05.142345+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a KSBA family over a non-normal base whose normalization has components with different variations, and compute whether κ_r is numerically nonzero for r strictly between the normalized variation and the total variation. Theorem 3.18 predicts κ_r ≡ 0 for r > normalized variation and non-zero for r ≤ normalized variation; any counterexample would falsify the numerical-triviality criterion. Alternatively, compute the virtual bundle E_{1,m} for a low-genus moduli space of stable curves and check whether the formula κ_1 = c_1(E_{1,m})/(N^{n+1}) reproduces the standard λ_CM class; a mismatch wou","supporting_citations":[],"review_version":1}