gpt-6.2-sol high: cf-geometry-modular
Reviewer record, not the publication verdict. The full response includes unaccepted proposals and source limitations. The curated report states what was adopted. Initial packets used a conservative 2010 cutoff; focused packets used the verified 2012 journal received date. The focused Fable answer needed a metadata-only derivative; its original failure and partial source access remain recorded.
Completed visible assessment, preserved as a readable check record. It is a model assessment, not a formal certificate. This document contains no private reasoning stream.
The literal conclusions for ordinary flags, standard-torus Schubert varieties, and the two endpoint-inclusive Lefschetz representations survive the local defects identified here. Lagrangian flag stability has a repairable proof, but the printed fake-degree formula fails already in degree one. The modular section has genuine false identifications: the special-linear adjoint is not always simple, and in characteristic two the symplectic Lie algebra is the symmetric-tensor submodule, not the symmetric-square quotient. The most consequential failure is Theorem 8.4 at p=2: the claimed integral F_2-vector-space abelianization has a Z/4 quotient. Theorem 8.3's parity lists are consistent with the Boolean model, but its proof omits the map conditions. No later theorem is used to close these gaps. The historical test uses the recorded 2010-08-07 first-arXiv fallback, not a verified journal received date; v3 is the 2013 author text. Dates of earlier references are supported here only by the manuscript's bibliography. Where an earlier preprint might matter, the particular theorem and version still need dated evidence. Source-only review cannot eliminate possible later knowledge from model weights, so external results have been treated as conditional rather than validated by recollection. The independent matrix and exterior-power arguments are separate checks, not evidence that the original cited versions contained the required results.
G1
Checked statement: Section 7, shared-counter items 7.1–7.7: reversed-map stability for ordinary and Lagrangian flags, the tableau formula, multiplicity stability for equivariant Schubert cohomology, the two virtual Lefschetz statements, and the diagonal coinvariant question.
Returned assessment: repairable_gap
7.1 uses the restriction maps for the inclusions of flag varieties and the right Weyl action in the compact model. These are consistent; the last tautological line restricts to a trivial line. The span of the degree-d basis monomials not involving the last variable maps isomorphically onto the smaller space, where d is polynomial degree and cohomological degree is 2d. In the larger ring a monomial of degree d has at most d positive exponents. For n>d, place its r positive exponents in the last r positions among the first n: each exponent is at most d-r+1, less than the smallest of those positions. Thus the chosen section generates under S_{n+1}. This checks both reversed-map conditions, not forward type preservation. Given 7.2, the eventual tableau bijection proves multiplicity stability: after the last descent the row numbers cannot increase, so if n>max(S)+|λ|, some label in this final string is in the top row and hence n is in that row. Use max(∅)=0. The supplementary uniformity assertion is also justified: each degree-d monomial orbit has an S_{n-r} fixing the inactive variables, r≤d, so the branching rule forces first-row length at least n-d for every constituent, and the coinvariant piece is a quotient. Only finitely many padded labels then need to stabilize. 7.3 has a false statistic as printed, but no counterexample to its stability conclusion. 7.4 claims only multiplicity stability, for which no map-level assertion about Tymoczko's basis is necessary. With the standard diagonal torus and cohomological grading corrected, its argument needs the finite fixed lower Bruhat interval and stable multiplicities in polynomial pieces. The two Lefschetz theorems, 7.5 and 7.6, are literally true for a simpler reason than claimed: both order complexes are equivariantly cones, so both unreduced Lefschetz representations are trivial. The printed appeals to Stanley do not prove the intended nontrivial proper-part assertions. Item 7.7 is a question, not a theorem. Its footnote reports a later answer, not an original proof.
Premises used
- Borel [Bo1], available_from: 1953; date_evidence: manuscript bibliography, Ann. of Math. 57, 115–207; date_basis: cited publication year; version: cited journal article, not inspected. Needed: rational coinvariant presentation for the complete type-A and type-C flags, with generators the degree-two tautological classes and the respective permutation/signed permutation action. No integral presentation is needed.
- Fulton [Fu], available_from: 1997; date_evidence: manuscript bibliography, LMS Student Texts 35; date_basis: cited edition year; version: cited book, not inspected. Prop. 10.3 is invoked for the type-A coinvariant presentation and monomial basis 0≤j_k<k; Prop. 10.7 for Bruhat incidence for standard type-A Schubert varieties. Incidence plus the subword description, not mere preservation of comparisons between old permutations, gives the fixed lower interval.
- Reutenauer [Re], available_from: 1993; date_evidence: manuscript bibliography, LMS Monographs New Series 7; date_basis: cited edition year; version: cited book, not inspected. Thm. 8.8 as invoked in 7.2 is the major-index count for the rational S_n coinvariant piece of polynomial degree d and a defined padded Specht label. Its proof is also cited for the graded decomposition of the polynomial ring as coinvariants times symmetric invariants. These uses do not require a statement about stability.
- Fulton–Harris [FH], available_from: 1991; date_evidence: manuscript bibliography, GTM 129; date_basis: cited edition year; version: cited first course, not inspected. Eq. 4.42 is the ordinary characteristic-zero S_n branching rule used for the finite-support bound; iteration is enough. Geck–Pfeiffer [GP], available_from: 2000; date_evidence: manuscript bibliography, LMS Monographs New Series 21; date_basis: cited edition year; version: cited book, not inspected. Lemma 6.1.3 is invoked elsewhere in CF-v3-TeX for the characteristic-zero bipartition induction rule; only induction with a trivial second factor, hence horizontal-strip Pieri in the positive partition, is needed for the alternate 7.3 check below.
- Stembridge [Ste], available_from: 1989; date_evidence: manuscript bibliography, Pacific J. Math. 140, 353–396; date_basis: cited publication year; version: cited journal article, not inspected. Thm. 5.3 is cited for the fake degrees of the hyperoctahedral coinvariant algebra in polynomial grading. The exact statistic and tableau conventions of that theorem were not inspected; the counterexample is to the manuscript's transcription.
- Tymoczko [Ty], available_from: 2008 by the cited publication; date_evidence: manuscript bibliography, Amer. J. Math. 130, 1171–1194, also listing arXiv:math/0604578; date_basis: cited journal year, without relying on the preprint; version: 2008 journal article identified, not inspected. Thm. 1.1 is used for the graded S_n-module decomposition for the indicated Weyl action on standard-torus equivariant cohomology of type-A Schubert varieties: a polynomial-ring copy shifted by the cell grading for each lower Bruhat element. No type-preserving maps, arbitrary torus, or general semisimple version is required.
- Stanley [Sta], available_from: 1982; date_evidence: manuscript bibliography, J. Combin. Theory Ser. A 32, 132–161; date_basis: cited publication year; version: cited journal article, not inspected. Theorems 4.3 and 6.4 are invoked for descent representations of rank selections in the Boolean lattice and cross-polytope face lattice. The intended application needs the proper-part/reduced Lefschetz conventions, actual ranks, and the type-B terminal descent when appropriate. Neither result is needed for the literal cone statements.
- Retrospective only: [CEF1], available_from: April 2012 as dated in the bibliography; date_evidence: arXiv:1204.4533 bibliography entry; date_basis: recorded preprint month; version: cited preprint, exact revision not supplied. Section 3 is the manuscript's reference for the affirmative answer to 7.7 and the multidiagonal generalization. It is after the fallback cutoff and is not used in this audit.
Missing inputs
- The cited publications are not supplied. In particular, check the exact Stembridge statistic and the Stanley rank, descent and sign conventions against the cited editions before attributing a corrected formula to those authors.
- The exact action and grading formulation of Tymoczko Thm. 1.1 remains an imported premise. The role of the standard diagonal torus can be checked internally; the paper cited for the Weyl action cannot be.
- The packet gives native-line bands, but the pasted TeX does not display individual line numbers. Within those bands the locators identify the exact labels, environments, and offending lines by their TeX text; no finer absolute line numbers are asserted.
Scope changes
- Section numbering is shared: 7.2 is the nested tableau theorem, 7.5 and 7.6 are unlabeled Lefschetz theorems, and 7.7 is \label{question:coinv}, a question. The [CEF1] footnote is retrospective.
- The statements about flag cohomology use reversed-map stability; the equivariant Schubert and virtual Lefschetz statements ask only for multiplicity stability. No strong or simple stability is inferred.
- The exact tableau basis and fake-degree formula for the full diagonal coinvariant algebra are not needed to audit the status of 7.7.
Located concern
Native TeX lines 3570–3983, \label{thm:flag}, the basis line and the last sentence after the proof; \label{theorem:kw}, the line starting For any partition.
The coinvariant degree is conflated with cohomological degree, and the tableau formula omits the padded-label existence condition.
$x_j=c_1(E_j)$ has cohomological degree 2; the first polynomial piece is generally nonzero, whereas H^1 of the complete flag variety is zero. The bound d≤binom(n,2) does not by itself imply n≥|λ|+λ_1.
Use H^{2d}=R_d and H^{2d+1}=0, and impose n≥|λ|+λ_1 when writing V(λ)_n or tableaux of shape λ[n]. In the tail argument take max(∅)=0.
Local grading and domain corrections. Neither the ordinary nor the supplementary uniform reversed-map stability assertion fails.
Located concern
Native TeX lines 3570–3983, \label{thm:flagsp}, the lines together with $n$ if and only if and 2\sum j+|\lambda^-|, the subsequent tableau extension, and the final map sentence.
The printed flag major-index formula includes a spurious terminal contribution. The stated tableau extension is not descent-set preserving for the printed set. Also, the reversed flag maps and the section used for type C are only asserted by analogy.
For n≥2 the degree-one piece of the W_n coinvariant ring is the signed coordinate representation with double partition ((n−1),(1)), i.e. fixed padded label (∅,(1)). For degree one the printed formula would require an empty flag descent set. If the lone negative label is n, the printed set contains n. If it is j<n, j+1 is positive, hence below j, and the set contains j. The printed formula thus gives zero for a constituent occurring once at every such n. Moreover, when n is negative and n+1 is appended in the positive first row, the new ordinary descent at n replaces a terminal descent; when n is positive there is no new descent. The statistic needed for degrees must distinguish ordinary descents j<n from the terminal convention used for rank selections. The printed pigeonhole argument also discounts all negative boxes, even though earlier labels in the final string can occupy them.
Use the ordinary double-tableau descents j<n in the major-index sum, and check the convention against the actual Stembridge theorem; do not delete the terminal convention from a separate rank-selection statistic. For fixed descent set S, use n>max(S)+|λ^+|+|λ^-| before removing or appending the largest entry. Explicitly embed Lagrangian flags by appending a fixed isotropic line in the added symplectic plane; restriction of Chern classes sets x_{n+1}=0. In the signed coinvariant ring the monomial basis has 0≤a_k<2k: take the type-A basis in y_k=x_k² and multiply by the squarefree parity monomials. The section with a_{n+1}=0 generates by the same large-n support argument as in type A. An alternate argument, not validation of the printed tableau proof, uses the coinvariant presentations and cited Pieri rule: a homogeneous signed monomial orbit with r active coordinates is induced from a fixed W_r-module times the trivial W_{n-r}-module. After decomposing the W_r-module, Pieri makes the padded multiplicities constant: the interlacing inequalities on all rows below the first are fixed, and the growing first row eventually satisfies the remaining inequalities. Each homogeneous polynomial piece has finitely many such orbits. The decomposition of the polynomial ring into signed coinvariants times invariants follows from the type-A free decomposition in the y variables by multiplying by parity monomials: the sign subgroup acts diagonally on those monomials, and S_n permutes them. Invariants Q[e_1(y),…,e_n(y)] have stable dimensions in each degree. Comparing homogeneous pieces by induction on degree gives uniform multiplicity stability for the signed coinvariants.
The printed multiplicity calculation is false even in degree one, but the type-C reversed-map stability conclusion is repairable. The alternate argument is conditional on the indicated classical inputs, not on any later stability theorem.
Located concern
Native TeX lines 3570–3983, \label{section:schubert}, Let $T$ be a maximal torus in $G$; \label{theorem:schubert}, the shift display and the ordinary-cohomology aside.
An arbitrary maximal torus need not preserve a standard Schubert variety. The proof writes polynomial grading as cohomological grading.
For w=e, X_w is the standard flag point. In GL_2 a conjugate of the diagonal torus that does not fix the standard line does not preserve that point. Even for the correct torus and w=e, the displayed degree-i formula at i=1 would give linear polynomials, while H^1(BT;Q)=0. Finally, preserving Bruhat comparisons of elements already in S_m is not by itself enough to rule out new elements below w.
Fix the standard diagonal torus in B and its rank-compatible versions. State the formula in half-degree, H_T^{2d}=⊕{v≤w}P{d−ℓ(v)}, with odd part zero and negative polynomial degrees zero. A reduced word for fixed w in S_m uses only the simple generators of S_m; the subword criterion shows that its lower interval is unchanged. In the ordinary aside the nonzero terms in degree 2d are those with ℓ(v)=d. Geometrically the fixed-w Schubert varieties are identified by the append-the-standard-flag inclusions. In the equivariant theorem retain the stated multiplicity conclusion only.
The literal arbitrary-torus setup and regrading are wrong. Neither shows failure of the standard-torus multiplicity theorem.
Located concern
Native TeX lines 3570–3983, \label{section:lefschetz}, both endpoint-inclusive poset definitions, the $L'_n(S)$ display, and the Stanley [Sta] application in the cross-polytope proof.
The order complexes include the extrema, so the nontrivial descent formulas cannot compute their Lefschetz representations. The unreduced-to-reduced correction has the wrong sign. For cross-polytopes the manuscript selects by dimension, not by rank.
Coning at the fixed least element gives L_n(S)=L_n^C(S)=Q. For the Boolean case take S={1}, n≥2, and the trivial label: the printed signed L⊕Q has trivial multiplicity 2, while the one-row tableau has no descent and contributes zero. For nonempty proper-part complexes, reduced Lefschetz class is L−Q, not L⊕Q. In the cross-polytope lattice a nonempty face of dimension d has rank d+1, and facets have rank n; a type-B terminal descent n can matter when that rank is selected.
For the statements as written replace both proofs by equivariant contraction. To obtain the intended nontrivial statements, realize the selected proper parts, take their reduced Lefschetz classes, and use the sign associated with the number of selected proper ranks. Treat the empty selection with the reduced degree −1 convention. Translate selected cross-polytope dimensions into ranks, using the cross-polytope descent rule including terminal descent where the actual rank selection calls for it. For a fixed selection the stable tail is independent of these small-rank endpoint adjustments.
Both literal stability theorems are true; their nontrivial displayed character formulas are not. A corrected proper-part calculation is a change of object, not a reinterpretation of the given endpoint-inclusive definition.
G2
Checked statement: Sections 8.1–8.3: rational and modular congruence-group assertions, the finite-group labeling conventions, Definition 8.1, and the asserted degree-one evidence for Conjecture 8.2.
Returned assessment: claim_false
The first-order map I+pA↦A mod p is an equivariant homomorphism to the additive trace-zero matrices: multiplication gives additivity and the determinant gives trace zero. Its image contains the elementary off-diagonals and, by conjugation, spans the diagonal differences too. This calculation does not identify the entire abelianization. The unrestricted integral identification and the adjoint irreducibility assertion are false. For n≥3 the scalar line appears exactly when p divides n; the resulting additional trivial composition factor is consistent with, and explains the need for, period p in the conjecture. Over an algebraic closure the trace-zero module modulo its scalar center is simple: a nonzero submodule with a root weight contains its elementary matrix, whence permutation conjugates and elementary conjugation yield the root spaces and diagonal differences; a noncentral diagonal vector produces a root vector by conjugation. The resulting highest weight ω_1+ω_{n−1} is p-restricted for n≥3, so the finite-group simplicity step uses the restricted highest-weight theorem, not semisimplicity in characteristic p. For Sp, the map I+pA↦A mod p has target the Lie algebra with the conjugation action. The module underlying this Lie algebra is the symmetric tensors in H⊗H, isomorphic to the symmetric-square quotient only when 2 is invertible. The manuscript's parity list for the Lie algebra in characteristic two is compatible with the corrected symmetric-tensor filtration. Its stronger assertion about an extra H in field-coefficient congruence homology does not follow from an integral extension. Definition 8.1 counts composition factors, not direct summands, and also requires injectivity and orbit-span generation. Its printed uniform clause makes the period common, not explicitly the onset. The paper's statements about higher homology remain conjectures. Charney's cited stability result is conditional here; rational transfer yields invariants without any assertion over the localized integers. If rational stabilization is assumed, eventual triviality follows because a transvection on two handles or coordinates disjoint from the old image fixes the stabilized image pointwise, and its conjugates normally generate the finite special linear quotient. In the other direction, eventual triviality plus the compatible rational transfer and rational stability of the ambient SL_n(Z) gives rational stability of the kernels.
Premises used
- Charney [Ch], available_from: 1984; date_evidence: manuscript bibliography, Comm. Algebra 12, 2081–2123; date_basis: cited publication year; version: cited journal article, not inspected. Needed as asserted in 8.1: eventual stabilization for level-p kernels under standard inclusions with Q, and for the stronger sentence with Z[1/p]. The exact theorem number, coefficient and rank hypotheses are not supplied. The transfer identification used in 8.1 is over Q. The converse in the author's equivalence additionally uses ambient rational stability, stated in the introduction without a precise version or range; available_from and version for that particular invocation are unresolved.
- Lee–Szczarba [LSz], available_from: 1976; date_evidence: manuscript bibliography, Invent. Math. 33, 15–53; date_basis: cited publication year; version: cited journal article, not inspected. Needed: in the high-rank range, the integral abelianization of the level-p SL_n(Z) kernel, identified via the first-order reduction map; field coefficients then follow by tensoring. The exact prime/rank hypotheses are not verified. If one invokes the congruence subgroup theorem separately, [BMS], available_from: 1967; date_evidence: bibliography, Publ. Math. IHES 33, 59–137; date_basis: cited journal year; version: cited article, not inspected. Its title sets ambient rank conditions, but a generic citation to the congruence subgroup theorem is not the required commutator/abelianization calculation.
- Humphreys [Hu], available_from: 2006; date_evidence: manuscript bibliography, LMS Lecture Note Series 326; date_basis: cited edition year; version: cited book, not inspected. Needed: dominant highest weight classification for finite-dimensional rational algebraic representations over the algebraic closure, Weyl modules and simple heads, and irreducibility/distinct labeling on the finite group by p-restricted highest weights for the split simply connected groups over F_p. No unrestricted statement about abstract algebraic-group actions is used. For the displayed finite-group modules, the constructions are defined over F_p; absolute simplicity is checked via scalar extension.
- Hogeweij [Ho], available_from: 1982; date_evidence: manuscript bibliography, Indag. Math. 44, 441–452; date_basis: cited publication year; version: cited journal article, not inspected. Cor. 2.7 and Lemma 2.10 are used for the odd-characteristic symplectic adjoint and the characteristic-two primitive exterior-square factor calculation, subject to the rank and rational-group hypotheses of the cited results. Passage from algebraic to finite groups also needs restriction of the indicated restricted weights.
- Putman [Pu1] and work of Sato: available_from: Putman preprint record identified by arXiv:0908.0555, August 2009 at the work-record level only; date_evidence: manuscript bibliography gives that identifier and Duke Math. J. 161 (2012), 623–674; date_basis: preprint identifier for the candidate pre-cutoff record, cited journal year for the final article; version: the pre-cutoff version containing the needed congruence calculation is not supplied or verified. Sato's exact source, version, date_evidence, date_basis and available_from are unresolved. Needed: odd-p, g≥3 integral abelianization via first-order reduction, and the characteristic-two integral result with its actual coefficient, kernel and rank hypotheses. The 2012 version cannot substitute for that evidence.
Missing inputs
- The exact Charney and Lee–Szczarba theorem statements and range restrictions, and the exact ambient rational stabilization input for the asserted equivalence.
- A dated pre-cutoff Putman/Sato text containing the two symplectic congruence computations, with the integral kernel, coefficient reduction and compatible genus maps identified.
- An explicit convention for finite length, p-restricted labels, mixed labels, and whether the onset as well as period is uniform. Neither a finite-length hypothesis nor an infinite-length replacement for Jordan–Hölder multiplicity is written in Definition 8.1.
- Individual native line numbers are not displayed in the pasted body; the exact textual anchors within the packet's line bands are given above.
Scope changes
- The degree-one identities and evidence were checked; neither part of Conjecture 8.2 in general degree was treated as an established result.
- The motivating claim about nontrivial mod-p action is supported as a degree-one contrast, not as a universal assertion in every i; H^0 with trivial coefficients always has trivial action.
- Rational transfer uses invertibility of the finite quotient order. It is not an invariants argument over Z[1/p] merely because p is inverted, and is not such an argument over F_p.
Located concern
Native TeX lines 3984–4242, \label{eq:abelianize1} and the following unrestricted integral/field isomorphism.
The integral abelianization identification is false without a rank condition.
Take n=2, p=2. Modulo ±I, Γ_2(2) is free on the two upper and lower elementary unipotents with off-diagonal entry 2. Indeed even row operations reduce the first column to (±1,0): the top entry is odd and the lower even, so when both are nonzero the larger absolute value is strictly reduced by subtracting an even multiple of the smaller. The leftover matrix is an upper elementary unipotent times ±I. Freeness follows by the projective-line ping-pong sets |x|<1 and |x|>1: each nonzero upper-unipotent power sends the first into the second, and each nonzero lower-unipotent power sends the second into the first. Hence the integral abelianization has a Z² quotient, unlike the finite trace-zero module.
State the actual Lee–Szczarba range and separate first-order surjectivity from the computation of its kernel. Derive the field-coefficient formula only after specifying the integral result or an independent field computation.
The low-rank assertion is false. It does not refute eventual periodicity.
Located concern
Native TeX lines 3984–4242, the sentence \fsl_n(\F_p) is an irreducible and the display after \label{conjecture:modular1}.
The Weyl/adjoint module is incorrectly identified with its simple highest-weight quotient at all ranks.
For every n≥3 divisible by p, I_n is a nonzero trace-zero matrix fixed by all of SL_n(F_p), so the adjoint module has a proper trivial submodule. These ranks occur arbitrarily far out.
In the valid abelianization range use the factors V(1;1)_n and, precisely for p|n, V(0). Given the natural first-order identification, block inclusion is injective and its orbit generates: elementary off-diagonals at all positions are conjugates up to sign and their elementary conjugates give diagonal differences. Check trivial-weight multiplicity using period p, not ordinary mixed stability.
Refutes the asserted all-rank simplicity and the claimed unperiodic mixed stability explanation, but not the stated period-p conjecture in degree one.
Located concern
Native TeX lines 3984–4242, \label{section:periodicity}, lines identifying $\fsp_{2g}\F_2$ with $\Sym^2 H$ and the paragraph on its parity-dependent factors.
In characteristic two symmetric tensors and the symmetric-square quotient have different module structures.
Writing the symplectic Lie condition as symmetry after identifying H* with H gives (H⊗H)^{flip}. It has an invariant scalar. In contrast for g≥2 the symmetric-square quotient has no invariant: invariance under the independent SL_2(F_2) actions on the handles first forces a linear combination of a_i²+a_i b_i+b_i², and permutation of handles forces equal coefficients. The symplectic transformation a_1↦a_1+a_2, b_2↦b_2+b_1 (other indicated basis elements fixed) changes their sum by a_2²+b_1².
Write the Lie algebra as symmetric tensors, or the dual of Sym²H. Use 0→Λ²H→(H⊗H)^{flip}→H^{(1)}→0, where the injection sends x∧y to x⊗y+y⊗x and the quotient is given by the diagonal classes. The twist restricts to H over F_2. For contraction c:Λ²H→F_2, c(ω)=g mod 2; the invariant line is in ker c exactly for even g. The cited primitive-factor result can then be applied with the correct module.
False module isomorphism, not a refutation of the displayed parity list of Jordan–Hölder factors for the Lie algebra.
Located concern
Native TeX lines 3984–4242, \label{section:periodicity}, the line Putman also proved about $H_1(\Gamma_{2g}^{\rm Sp}(2),\F_2)$.
The extra characteristic-two kernel is treated as an extension in field-coefficient H_1. An integral extension does not imply that statement.
The explicit Z/4 characters in G3 detect independent doubles of upper and lower long-root transvections in the integral abelianization A. If the intended integral input is 0→H→A→sp→0, then 2A is contained in that kernel, and these 2g independent doubles force 2A to equal the whole kernel. Tensoring yields A/2A≅sp, not a further extension by H. This inference is conditional on the integral kernel computation, and does not presume that the cited version has been inspected.
Quote the exact integral result and compute the reduction modulo 2, including the image of multiplication by 2; alternatively quote a verified computation directly over F_2. With the natural first-order identification, block inclusion and orbit-span generation can be checked on symmetric tensors supported on at most two handles. These map conditions are not consequences of the parity calculation alone.
The manuscript has a coefficient-interface error. Period-two composition factors for field homology are compatible with the corrected calculation, but the external integral input is still conditional under the historical policy.
Located concern
Native TeX lines 3984–4242, highest-weight and Jordan–Hölder paragraphs; unlabeled Definition 8.1 [Stable periodicity]; \label{conjecture:modular1}.
The domain and quantifiers of the modular definition are incompletely specified.
Jordan–Hölder series need finite length; the locally finite convention referenced from §2.3 alone does not give it. A Frobenius twist changes an algebraic highest weight but not its restriction to F_p points. A single period for every label does not imply a single starting rank. I and II are written as hold, referring to a definition with a sufficiently-large-rank qualifier.
Require finite-dimensional modules, use dominant p-restricted weights as finite-group labels, and make the mixed-label rule explicit. State that I and II are eventual if that is intended, and say separately whether uniformity includes a common onset. In the finite-support degree-one cases the distinction between the two uniformity conventions does not change the conclusion.
Clarifies the scope of Definition 8.1 without turning Conjecture 8.2 into a theorem. A general failure of higher-degree modular stability has not been shown.
G3
Checked statement: Theorems 8.3 and 8.4: Torelli torsion and level-p mapping class group abelianizations, including the genus maps, rank and prime ranges, and their composition factors.
Returned assessment: claim_false
8.3's two composition-factor lists follow from the natural Boolean polynomial model if that model is granted. For the affine space of quadratic refinements over F_2, let B_d be the Boolean polynomials of degree at most d. The quotients B_0, B_1/B_0, and B_2/B_1 are respectively the trivial module, H, and Λ²H: in the leading degree the action is linear and the squarefree degree-two classes transform by exterior square. Contraction on Λ²H gives a trivial quotient, and the invariant bivector belongs to its kernel exactly in even genus. This gives the printed lists using the primitive-factor calculation. But a list alone does not verify the maps. If the Johnson torsion–B_2 identification is natural for genus extension, restriction of refinements induces an injective pullback on polynomials, and every squarefree degree-at-most-two basis monomial involves at most two handles. Permutations of handles give generation, including for the two-step maps on the parity subsequences. The theorem is therefore repairable at its map interface; no counterexample to it was found. 8.4 is false with its stated integral coefficient ring at p=2, independently of the external computations. For odd p the composition-factor proof has the congruence cases reversed. A corrected odd-p proof also needs the natural exact sequence, not just an abstract description at each genus. If the standard genus maps make the claimed odd-p exact sequence natural, both end terms are injective and orbit-generating: the standard tensor bases of Λ³H and the symplectic adjoint use at most three and two handles, respectively. Injectivity in the middle follows by chasing the exact rows; its orbit-span contains the kernel and surjects onto the quotient. Thus the odd-p conclusion can be repaired, subject to the exact historical extension theorem. An all-primes claim about F_p homology, or a claim about composition factors of a p-primary filtration instead of a vector space, would be a different theorem.
Premises used
- Johnson [Jo2], available_from: 1985; date_evidence: manuscript bibliography, Topology 24, 127–144; date_basis: cited publication year; version: cited journal article, not inspected. Needed: for the one-boundary Torelli group, g≥3 as stated, its integral abelianization has the torsion detected by the degree-at-most-two Birman–Craggs–Johnson quotient, equivariantly for Sp_{2g}(F_2). The group-level abstract splitting alone does not give that equivariance or the natural stabilization map.
- Shvartsman [Sh], available_from: 2009; date_evidence: manuscript bibliography, Mat. Zametki 86, 478–480; date_basis: cited Russian publication year; version: cited Russian article, not inspected. Needed: composition factors of B_2 for the modular symplectic action in the genus range used, not an integral splitting and not automatically a genus-map statement.
- Hogeweij [Ho] (1982 cited journal article) and Humphreys [Hu] (2006 cited book): available_from, date_evidence, date_basis and version as recorded in G2. Their relevant uses here are the primitive exterior-square factor, odd-characteristic symplectic adjoint, and p-restricted finite-group restriction. Neither states naturality of the mapping-class-group extensions. The exterior-cube test below is a direct calculation, not an imported later result.
- Hain [Ha3], available_from: 1995; date_evidence: manuscript bibliography, MSRI Publ. 28, 97–143; date_basis: cited volume year; version: cited publication, not inspected. Prop. 5.1 as used states vanishing of H^1 with integral trivial coefficients for g≥3 in the stated level setting. To deduce a torsion abelianization also use finite generation of the level group, from finite index in a finitely generated mapping class group. This does not imply exponent p and is not needed for the counterexample.
- Putman [Pu1], available_from: August 2009 for the candidate arXiv work record only, exact availability of the required result unresolved; date_evidence: arXiv:0908.0555 and Duke Math. J. 161 (2012) in manuscript bibliography; date_basis: preprint identifier and cited journal year; version: exact pre-cutoff version not verified, final journal version after cutoff. Needed for odd p: Thm. 7.8 as cited, for one boundary component, odd level specialized to prime p and g≥5 as stated, an integral extension with quotient sp_{2g}(F_p), kernel Λ³H, and an actual F_p-linear quotient action on the middle term; stabilization needs naturality. Exact versions for the even-level congruence calculation and the even-level mapping-class calculation are not supplied. Sato is mentioned without a precise reference: available_from, date_evidence, date_basis and version unresolved. Neither a separate even-level integral extension nor an assertion about mod-two homology is taken as verified.
- The standard reduction quotient and the surjection Mod_{g,1}→Sp_{2g}(Z) are stated in CF-v3-TeX. Surjectivity onto the level-p symplectic kernel follows by taking preimages in that displayed surjection. The p=2 obstruction uses only this interface and block matrix multiplication.
Missing inputs
- Johnson/BCJ naturality on torsion under the indicated surface inclusion, or a direct topological proof that the torsion map is pullback along restriction of quadratic refinements. The polynomial calculation checks the resulting map, not its identification with the group-induced map.
- Dated pre-cutoff Putman/Sato versions giving the exact odd- and even-level sequences, the rank bounds, the action, the coefficient reductions, and compatible maps. A journal theorem number in v3 does not identify the pre-cutoff statement.
- For a new p=2 field-coefficient stability theorem one must compute reduction of the actual integral extension, account for any killed layers, and check the degree-one maps. The printed display is not a computation of that new theorem.
- Individual native line numbers are not displayed in the pasted body; the exact labels and textual line anchors within the assigned band are given above.
Scope changes
- For Theorem 8.3 the modular period-one conclusions for the parity subsequences are understood with their two-step genus maps.
- The odd-prime and characteristic-two parts of Theorem 8.4 have different statuses: repairable conditional odd-prime argument, false integral F_2-module assertion.
- The adjacent higher-homology conjecture was examined only for its stated coefficient convention; no higher-homology stability claim was established or rejected on the basis of a later result.
Located concern
Native TeX lines 4243–4388, \label{thm:torabtor}, proof ending immediately after the two Shvartsman lists and the comparison with $\bwedge^2H$.
The proof checks the multiplicity condition but does not check I or II for the indicated integral-homology torsion maps.
Equal lists for ranks two apart do not determine even whether the map is nonzero. Conditional on the natural B_2 model, the structure map has a concrete description: any quadratic refinement on the smaller surface homology extends to the larger one, so pullback is injective; basis monomials use at most two handles, so permutations in the target Weyl-type family give the required symplectic group orbit generation. The same argument is valid for a jump of two in genus.
State the natural equivariant BCJ comparison on torsion, then supply the pullback and orbit-span argument. Describe each parity subsequence as modular composition-factor stability of period one, with its two-step maps, not as characteristic-zero complete-reducibility stability.
Proof gap for the full modular stability statement, not a false periodic list and not a counterexample to 8.3.
Located concern
Native TeX lines 4243–4388, \label{thm:ModgL}, the exterior-cube factor display at the start of the proof and the next odd-p display.
The p-congruence cases for the additional V(ω_1) factor of Λ³H are reversed.
Let c:Λ³H→H be symplectic contraction and L(v)=v∧ω. For g≥3, c is onto, L is injective, and cL=(g−1)id, as seen on a symplectic basis. If g≡1 mod p the image of L is inside ker c, so there are at least two standard factors, one inside the kernel and one in the quotient. This contradicts the printed list at arbitrarily large genera. In fact, over the algebraic closure, K=ker c modulo K∩L(H) is simple of highest weight ω_3. To see this directly, its weights are the three-distinct-handle weights and the single-handle weights. The first have one-dimensional spaces spanned by basis wedges; signed permutations take any one to all of them. In the a_i single-weight space the vectors in K are a_i∧∑_{j≠i}c_j a_j∧b_j with ∑c_j=0. The symplectic shear a_j↦a_j+a_k, b_k↦b_k−b_j applied to a wedge on three distinct handles produces the differences of these pair terms modulo three-handle wedges. Thus the three-handle weight spaces generate K. Conversely a single-weight vector not proportional to L(a_i) has two different coefficients; the same shear produces a nonzero three-handle component. Torus weight decomposition and signed permutations now show simplicity of the quotient. The argument also works when p=2; the two simple weights ω_3 and ω_1 are p-restricted.
Use V(ω_3) and two V(ω_1)'s when g≡1 mod p, and V(ω_3) and one V(ω_1) otherwise. For odd p in the cited g≥5 range add one V(2ω_1), conditional on the natural extension. Do not obtain an even-level integral F_2-module formula by merely switching the cases in the final display.
The stated multiplicities are false, but reversing the exceptional residue leaves period p intact for odd p. The argument in evidence supplies the exterior-power correction separately from the original proof.
Located concern
Native TeX lines 4243–4388, paragraph preceding \label{thm:ModgL}, the theorem's integral H_1 statement, and the proof paragraph For $p=2$.
For p=2 the integral abelianization is not a vector space over F_2, so 8.4 as stated is false.
There is a direct obstruction for every relevant genus. For M∈Γ_{2g}^{Sp}(2), write M=I+2X and use symplectic blocks X=(A B; C D) ordered by all a's then all b's. Modulo 2, B and C are symmetric and D=A^t. For any i set f_i(M)=B_ii+2∑_j A_ij B_ij mod 4. Under M,M' multiplication the new X is X+X'+2XX'. The defect in B_ii is 2∑j(A_ij B'ij+B_ij A'ij) mod 4, using the mod-two symplectic identities. This cancels the polarization of the quadratic summand, so f_i is a homomorphism. The level-two long-root transvection I+2E{i,g+i} maps to 1 in Z/4. Pull back f_i along Mod{g,1}(2)→Γ{2g}^{Sp}(2), which is onto. The integral H_1 therefore has a Z/4 quotient for arbitrarily large g, in fact for every g in the stated range. Applying the same construction with the two Lagrangians swapped detects the lower long-root transvections; together the characters detect independent doubles in the symplectic integral abelianization as used in G2.
Restrict the integral F_p-module theorem to odd p, correct its multiplicity lists, and provide the naturality needed for I and II. For p=2 formulate a separate claim either about H_1 with F_2 coefficients or about specified layers of the integral 2-primary abelianization. Determine those layers and their maps from a verified even-level input; reducing an integral extension by tensoring with F_2 is not left exact.
A counterexample to 8.4 itself, not just a gap in the printed proof. No improvement of the stable range removes it.
Located concern
Native TeX lines 4243–4388, Hain paragraph in Level $p$ mapping class groups and the unlabeled conjecture after \label{thm:ModgL}.
Vanishing of H^1 with integral coefficients does not on its own establish torsion without finite generation. The downstream conjecture retains the invalid integral vector-space coefficient convention.
H^1(G;Z)=Hom(G_ab,Z) can vanish for a torsion-free non-finitely-generated abelian group, for example Q. For the indicated mapping class group the needed finite generation follows by finite index. In the downstream conjecture H_0(-;Z)=Z is not an F_p-vector space for any p, and the counterexample to 8.4 obstructs the same convention already at i=1, p=2.
Insert finite generation in the Hain inference. Specify coefficient-field homology or an explicitly defined filtered integral stability notion in the conjecture.
The Hain inference is repairable and does not support exponent p. The conjecture is not counted as a theorem; its literal coefficient setup needs correction.
Outside the audit
- External texts, their exact version histories, and the publisher's typeset article; no journal oblatum comparison was performed.
- Absolute individual line numbers beyond the native TeX bands furnished in the packet; the finer locators use TeX labels or exact textual anchors.
- Sections outside the assigned bands except for definitions, standard inputs, bibliography, and proof interfaces actually used here.
- Higher-degree cases of Conjecture 8.2, the later FI-module answer to 7.7, and the higher-degree mapping-class-group conjecture beyond its coefficient convention.
- A complete corrected p=2 computation of level mapping-class-group H_1, including its integral extension structure or the stabilized field-coefficient composition factors.
Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.