gpt-6.2-sol high: cf-foundations
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 foundations contain several independent false assertions, not a single gap propagated by citation. The largest theorem-level failures are nonuniform SL closure in 3.1 and nonuniform cancellation in 3.5. The printed proof of finite uniform cancellation also fails, but the finite uniform result is recoverable by a separate common-rank argument. Theorem 3.2 and the finite-dimensional uniform closure results in 3.1 have sound qualitative routes conditional on their identified classical inputs; the errors in some ranks, displays, definitions, and examples should not be promoted to counterexamples to those narrower conclusions. Locations use the supplied native line spans plus exact TeX labels and unique proof anchors; separately certified single-line offsets for the unlabeled passages were not available. This is a source-only audit of the final author TeX v3 dated 2013-06-10. The historical cutoff used is 2010-08-07, the recorded first-arXiv-submission fallback, not a verified journal received date. External chronology is based on the supplied bibliographic evidence at its stated precision, not inspection of the citations. Later knowledge cannot be removed from model weights; no later result is used to close an original obligation. The optional 2011 publication and its earlier preprint identifier are treated separately. No significance is attached to an automatic compilation date.
F1
Checked statement: Section 2: the labeling conventions, definitions of consistency and ordinary, uniform, multiplicity, reversed-map, strong, and mixed tensor stability, and the examples and implications attached to those definitions.
Returned assessment: claim_false
The main distinctions are sound once the algebraic representation category and naming conventions are made explicit. Consistency means equivariance for the fixed rank inclusion, not injectivity. Ordinary stability requires eventual injectivity and orbit generation, and allows a separate threshold for each fixed label. Uniform stability makes the multiplicity threshold simultaneous, with absent labels counted as zero. Multiplicity stability drops the map requirements altogether. Strong stability adds the condition on the orbit span of every irreducible copy, not just on the image of each isotypic component. In characteristic zero, multiplicity-one branching gives the stronger statement in Remark 2.13: the span of k copies is at most k copies, and restriction plus injectivity gives at least k. Thus IV implies I, and IV plus II gives stabilization of multiplicities in the range in question. The converse from III′ and IV to II really needs a finiteness condition on the multiplicities: a proper shift on a countable sum of trivial modules exhibits the obstruction. For finite groups, the induced-module criterion for II is the explicit extension g⊗v ↦ gφ(v). Reversed stability instead requires a surjective map and a linear section whose image generates by the larger group; it is not the forward criterion on the transposed maps. Mixed stability uses fixed pairs of partitions and the same forward conditions I and II. The highest-weight formula for an individual mixed tensor dual is correct, but it does not justify the sequence-level assertions discussed below.
Premises used
- Highest-weight classification, complete reducibility for finite-dimensional rational algebraic modules, and the Schur realizations: version = Fulton–Harris, Representation Theory, GTM 129, 1991, especially §§6, 15.3, 17 and the relevant classical-group chapters; available_from = 1991 at year precision; date_evidence = manuscript bibliography FH; date_basis = bibliographic publication year. The argument uses algebraic representations in characteristic zero, not all abstract actions of rational points.
- Multiplicity-one branching: version = Fulton–Harris 1991, Eq. 4.42, Ex. 6.12 and Eq. 25.36, and Geck–Pfeiffer 2000, Lemma 6.1.3; available_from = 1991 and 2000 respectively; date_evidence = manuscript bibliography FH and GP; date_basis = bibliographic publication years. Needed only for the same padded label in a restriction by one rank. For SL the normalized partitions in the applicable range have zero last part, which excludes extra determinant translates; for GL the retained tail is constant; for Sp the interlacing sequence for the unchanged label is forced. For S and W delete the box at the end of the added first row.
- The transfer interface in Remark 2.4 can be checked directly: restrict a rational free group resolution for the extension group to the normal subgroup, take normal-subgroup coinvariants, then take quotient-group coinvariants. Finite-group averaging makes the latter functor exact and identifies coinvariants with invariants. Averaging need not preserve injectivity when the acting group changes.
Missing inputs
- The cited books were not supplied. The branching and classification references are identified from the manuscript; their external text has not been independently verified.
- A convention is needed for equality of infinite multiplicities: actual cardinals, or a restriction such as countable dimension under which the single symbol ∞ is unambiguous.
Scope changes
- No collapse of ordinary, uniform, multiplicity-only, reversed, strong, or mixed stability is made. In particular, the numerical test of mixed stability in Definition 2.16 is pointwise in a pair of partitions.
- The repairs to Proposition 2.14 and to the exterior-power example are new arguments or choices of maps, not validations of the literal text.
Located concern
Lines 583–1350; §2 opening definition of 'representation' and \label{section:repstab:def}, paragraph ending 'decomposes as a direct sum of finite-dimensional irreducibles'.
Local finite dimensionality does not imply semisimplicity for abstract actions of the listed groups of rational points.
For GL_n(Q), let g act on Q² by the matrix with diagonal entries 1, lower-left entry 0, and upper-right entry v_2(det g). This is a finite-dimensional action, consistent in every rank with identity transitions. It is a nonsplit extension of the trivial representation by itself: diag(2,1,…,1) acts by a nonidentity unipotent Jordan block. The manuscript's algebraic highest-weight discussion does not classify this action.
For SL, GL, and Sp stipulate locally finite rational algebraic representations, with each vector contained in a finite-dimensional rational algebraic submodule. A sum of such completely reducible submodules is semisimple. Retain the usual characteristic-zero category for S and W.
The decomposition required by the definitions is not available for the literal class of actions admitted by the introductory wording. This is repaired by making the intended category explicit.
Located concern
Lines 583–1350; §2.1, W_n padding display; §2.2, GL pseudo-partition displays and Remark 2.1; \label{rem:Spweights}.
Some naming and weight-coordinate assertions are not literally consistent.
For GL, the displayed Schur–determinant formula extends (λ_1,…,λ_k) by repeating λ_k, whereas the asserted weight ∑_{i≤k} λ_i L_i omits that tail. In rank >1 the label (1) in that formula gives det, not the standard module; (1,0) gives the latter. Uniqueness is only modulo repetition of the final entry, as the following remark itself illustrates. For W allow the first part of the empty positive partition to be 0 and omit a padded zero row when the positive side is empty. In the Sp comparison, a_i^−b_i^ restricts to twice the functional given by a_i^*, not to that functional.
Use constant-tail equivalence classes for GL, with a declared least usable rank or a canonical representative, and write the full weight including the tail. Write fixed polynomial labels with zero tail when using this naming convention; at the smallest rank use the polynomial Schur functor itself. Normalize Sp coordinates by L_i(diag(t,−t)) = t_i.
The intended labels and stable branching survive. The literal GL highest-weight and uniqueness assertions and the Sp weight normalization need correction.
Located concern
Lines 583–1350; the paragraph following \label{definition:repstab1} on λ-stability and unlabeled Remark 2.4 beginning 'Fix either'; \label{definition:repstabrev}.
The transfer remark conflates multiplicities and isomorphisms on specified maps, and the reversed-map definition has a missing subspace designation.
Exact sequences at individual ranks specify no transition maps. Even with compatible maps and I–II, infinite stable trivial multiplicities need not give homological isomorphisms. In degree one take A_n=E⊕Std_n as an additive rational group, E=⊕{r≥0}Qe_r trivial, and Γ_n=A_n⋊S_n. Send e_0 to a nonzero S_n-fixed vector in Std{n+1}, e_{r+1} to e_r, and Std_n by its standard inclusion. The map is injective and its S_{n+1}-span is all of A_{n+1}, but the induced coinvariant map kills e_0. Conversely, stable quotient-group homology alone imposes no I–II requirements on the nontrivial types. In Definition 2.8 the section image is a subspace of V_{n+1}, not necessarily all of V_{n+1}.
Specify compatible maps of extensions. With II on the original representations the maps on coinvariants are surjective; finite, eventually constant trivial multiplicities then imply eventual isomorphisms. State separately the converse about trivial multiplicities, without inferring I–II for all types. In Definition 2.8 name the subspace U⊂V_{n+1}.
Transfer identifies the relevant spaces and their dimensions, but does not prove the claimed unrestricted equivalence. The reversed-map typographical repair preserves the distinct intended definition.
Located concern
Lines 583–1350; \label{example:tensor}, the following unlabeled exterior-power Non-example 2.10, and \label{nonexample:regular}.
Three examples overstate what is obtained from their multiplicity decompositions.
The standard tensor-square example is uniformly stable with the usual maps from rank 2, but its extension to arbitrary uniformly representation stable inputs is false: the second construction in §3.1 Non-example 3.3 has the same validity for GL with upper-left inclusions and misses the exterior-square projection. For the usual maps on ∧*Q^n, the image and orbit span have no component in degree n+1. Thus the example is pointwise multiplicity stable, not ordinarily representation stable with those maps. For regular representations, the standard group-algebra inclusions are injective and their orbit spans contain the entire target, but V(0) has multiplicity exactly one for all n, for both S and W. The moving irreducible dimensions are not all unbounded; it suffices to use the growing dimension of V(1), inflated for W, to disprove full stability.
Limit the general tensor claim to uniform multiplicity stability, or require strong stability for the map-sensitive conclusion. For the exterior example either claim only multiplicity stability or specify different maps: keep the usual maps below degree n, and send the old top exterior line diagonally and nontrivially into the S L_n-fixed line ∧^nQ^n in degree n and the new top exterior line in degree n+1. The two target types are distinct for n≥2, so the orbit span meets both. Except the trivial label from the regular-module claim.
The original standard tensor-square and regular nonstability examples remain valid. The general tensor assertion, the exterior example with the usual maps, and the assertion about all fixed labels for the regular modules are false as stated.
Located concern
Lines 583–1350; \label{prop:SLstrong}, proof of IV′ ⇒ IV, especially 'By possibly rechoosing v' and 'the only weight λ′ occurring'.
The proof confuses ordinary weights with highest weights and does not justify rechoosing a vector when IV is required for every irreducible copy.
Even in a constant-support polynomial sequence there can be several weights above one old weight. In Q⊕Q^{n+1}, the weight of the last standard basis vector and the trivial weight both restrict to zero. Also, an image of an old highest-weight vector need not in general be a single large-torus eigenvector.
For IV′ ⇒ IV, decompose the image of each old highest-weight vector into large-torus components. They are all killed by the larger positive root algebra, and all have the same restricted weight, so only occurring highest weights need be considered. Uniform support and sufficiently large rank make restriction injective on those highest-weight labels (using normalized weights for SL and constant tails for GL). The image, nonzero by hypothesis, then generates a single diagonal copy of the required simple module. For IV ⇒ IV′, same-label branching puts the image on the old-group span of the large highest-weight line. The last-column unipotent fixes the line and is normalized by the old group, so fixes that span.
Repairable proof error, not a counterexample to Proposition 2.14 in the intended category. Its injectivity assumption is substantive when invoking the proposition as stated.
Located concern
Lines 583–1350; §2.4, paragraphs beginning 'Note that mixed tensor stability implies' and 'In general, the dual'.
For GL the mixed-label test misses fixed nonzero determinant tails, so mixed stability need not imply ordinary stability. Ordinary stability does not imply mixed stability of the duals, even at the multiplicity level; transposition moreover reverses arrows.
Write X_n=Q^n. Use Q⊕Sym²X_n at odd ranks and Q⊕Sym²X_n⊕X_n at even ranks. On odd-to-even transitions send the trivial generator to the sum of the new trivial generator and the last basis vector in X_{n+1}, and use the ordinary Sym² inclusion. On even-to-odd transitions keep the trivial and Sym² inclusions and send the extra vector v to v e_{n+1} in Sym²X_{n+1}. These maps are equivariant and injective; their orbit spans generate all target types. Tensor this sequence by the determinant with compatible identifications of its lines. All fixed mixed labels are eventually absent, but the fixed ordinary label det⊗X_n has alternating multiplicity. Separately, carry out the same construction with dual standard spaces, extending a dual basis by zero and using the last dual basis vector. This is ordinarily stable in the manuscript's GL labeling, since the nontrivial labels move. The dual modules have alternating multiplicity of V(1;0)_n regardless of what transition maps one tries to choose.
For SL, fixed ordinary labels are included among the mixed labels. For GL, either restrict the implication to an appropriate untwisted category, or include a fixed determinant exponent in the mixed labeling and test it too. Restrict the duality assertion to the identity V(λ;μ)_n^*≅V(μ;λ)_n and to mixed multiplicity stability if the original modules are already mixed multiplicity stable. Address transition maps separately.
The individual mixed tensor dual formula and the definition are usable. Both general sequence-level assertions are false.
F2
Checked statement: Theorem 3.1, all parts and the strong-stability clause, and Theorem 3.2, both parts: closure for the stated input groups, hypotheses, ranks, and structure maps.
Returned assessment: claim_false
Theorem 3.1(1)–(5) are false under the stated nonuniform SL hypothesis. Fixed-label stability does not keep the polynomial degree or number of rows of the moving summands bounded. The printed proof uses a polynomial GL size equality after restriction to SL, where full columns can disappear. The finite-dimensional uniform assertions have a different outcome: in a stable rank there are only finitely many labels and finite multiplicities. For (1), use the stable tensor laws for fixed input labels. For (2), use the fixed-label plethysm laws and apply the finite direct-sum formula and (1) inductively. For (3), that same formula has only finitely many fixed-degree terms. For (4), apply (1) and (2) successively to the tensor product, or use the given finite Kronecker expansion with corrected variables. For (5), apply (2) twice. For (6), restrict polynomial modules using the Schur direct-sum formula for Q^{n−k}⊕Q^k; determinant twists restrict correctly, and normalized SL labels are valid once the remaining rank exceeds the row bound. In Sp, eventually the interlacing pattern does not change; iterate the one-step result k times. For (7) use removal of the GL determinant exponent for GL→SL, and stable Littlewood restriction for GL/SL in dimension 2n to Sp_{2n}. Nothing in (6) or (7) asserts forward orbit generation. For the finite-dimensional strong clause, tensor products and Schur functors preserve injectivity and the last-column fixed-subspace property. Apply the corrected Proposition 2.14 to the uniformly multiplicity stable outputs, then Remark 2.13 to get generation. No strong closure for Sp is asserted. In Theorem 3.2 the number of labels in the common support is finite even when some multiplicities are infinite, because S_N is finite. The fixed-pair uniform tensor decompositions and one-step branching thus apply simultaneously to that support; tensor products distribute over arbitrary algebraic direct sums, using cardinal multiplicities. This checks both claims of 3.2 conditionally on the cited symmetric-group laws. No result with transition-map stability as conclusion is needed for 3.2.
Premises used
- Polynomial LR tensor and direct-sum formulas, fixed-degree plethysm, and Schur functors of tensor products: version = Fulton–Harris 1991, Eq. 6.7 and Exs. 6.11, 6.11(b), 6.12, 6.17(a); available_from = 1991; date_evidence = bibliography FH; date_basis = publication year. Used for fixed partitions and characteristic-zero vector spaces; a polynomial GL partition of full height cannot be treated as a same-size normalized SL label.
- Symplectic tensor product, interlacing branching, and polynomial GL-to-Sp restriction: version = Fulton–Harris 1991, Eqs. 25.27, 25.36, 25.39; available_from = 1991; date_evidence = bibliography FH; date_basis = publication year. The tensor formula is used only in its stable rank for fixed input partitions, not at every rank where the inputs exist. Branching is used for restriction by one symplectic pair. The manuscript invokes Littlewood restriction for a polynomial input with at most n rows in dimension 2n.
- Symplectic plethysm: version = Kabanov, J. Algebra 195 (1997), Theorem 7; available_from = 1997; date_evidence = bibliography Kab; date_basis = publication year. Used for a fixed outer Schur partition λ and a fixed Sp label μ, with the cited range n≥ℓ(μ)|λ| and output length bound ℓ(μ)|λ|; only eventual finite-support stability is required.
- Symmetric-group tensor stability: version = Murnaghan, Amer. J. Math. 60 (1938), §1 as cited; available_from = 1938; date_evidence = bibliography Mu; date_basis = publication year. Used in the form that the full padded decomposition for a fixed pair stabilizes, not merely pointwise convergence without a bound on the possible output labels.
- Symmetric-group restriction: version = Fulton–Harris 1991, Eq. 4.42; available_from = 1991; date_evidence = bibliography FH; date_basis = publication year. The single-box rule suffices for all fixed shifts by iteration; Ex. 4.44 is not needed.
- Optional sharper Kronecker range: version = Briand–Orellana–Rosas, J. Algebra 331 (2011), Theorem 1.2, n≥|λ|+|μ|+λ_1+μ_1; available_from = 2011 for the cited journal version and unresolved for the needed version by the cutoff; date_evidence = bibliography BOR gives arXiv:0907.4652 as well as the 2011 publication; date_basis = publication year and earlier preprint identifier only. The sharp bound is not needed for Theorem 3.2.
Missing inputs
- The cited texts were not supplied. In particular, exact stable-range qualifications of the symplectic formulas and the cited Kabanov statement have not been checked against the external versions.
- A dated BOR preprint version containing the quoted bound is needed to admit that quantitative statement under the first-submission cutoff; an early identifier does not identify the version containing the result.
Scope changes
- The verified uniform clauses are uniform multiplicity statements, except for the separately justified strong clause. No claim of orbit generation for restrictions or for the outputs of Theorem 3.2 has been inferred.
- Evaluation of the BOR bound is separated from evaluation of Theorem 3.2; chronology of that optional refinement does not block the qualitative result.
Located concern
Lines 1351–1849; \label{thm:classicalstability}, first paragraph and Parts 1–5; proof Part 1b at 'For SL_n Q the assumption of uniform stability is not necessary', and the nonuniform argument after \label{eq:Schursum}.
All five nonuniform SL closure claims fail.
For n≥3 take V_n=X_n=Q^n and U_n=X_n^* at even ranks, 0 at odd ranks. The normalized label of X_n^* is (1^{n−1}), so the inputs are finite-dimensional and pointwise multiplicity stable, as required; the theorem does not require structure maps for this clause. The trivial multiplicity of V_n⊗U_n alternates 1 and 0. For (2) use Sym² of the pointwise stable sequence X_n⊕U_n, and for (3) use Sym²(V_n⊕U_n); the cross term provides exactly the alternating invariant, since Sym²X_n and Sym²X_n^* have no diagonal-torus invariant for n≥3. For (4) use the one-box Schur functor on V_n⊗U_n; for (5) apply the one-box inner and symmetric-square outer Schur functors to X_n⊕U_n. In Part 1b the deleted full column is seen explicitly in X_n⊗∧^{n−1}X_n.
Delete the pointwise-only SL clause or use the uniform finite-dimensional clause. Another sufficient restriction for this proof is bounded polynomial size of all occurring normalized input labels in the eventual range: this gives finite possible label support and promotes pointwise to uniform multiplicity stability.
False theorem at the advertised nonuniform generality, not merely a missing proof. The counterexamples do not affect the uniform hypotheses or the strong clause.
Located concern
Lines 1351–1849; Theorem 3.1(6)–(7), proof at \label{eq:Spres} and \label{eq:SLSpres}, and concluding display of Part 4.
Several interfaces or ranges in the otherwise eventual assertions are misstated.
For Sp one-step restriction the printed 'once n≥ℓ(λ)' includes a rank where an output with ℓ(λ) rows cannot exist at rank n−1. For example V(1,1) in rank 2 restricts to rank 1, where (1,1) is unnameable, whereas it survives on restricting rank 3 to rank 2. Part 7's final display uses V_n where the input is V_{2n} and identifies unrelated thresholds. The stated cross-group restrictions are not operations on arbitrary Sp inputs. In Part 4 the concluding display says V_n⊕U_n and uses W_n in a tensor factor, though the preceding formula and the claim concern V_n⊗U_n.
For a fixed Sp label in (6), allow n≥ℓ(λ)+1 for the one-step argument, then increase the threshold for a k-step shift. In (7) name the source groups and use c_{λ,2n} for the symplectic restriction; require 2n in the input stability range and n at least the row bound on the fixed polynomial parts. In Part 4 use Schur_λ(V_n⊗U_n) and Schur_μ(V_n)⊗Schur_ν(U_n). In finite-rank GL tensor formulas keep track of determinant factors if an output polynomial has full height; in the stable range finite support allows that height to be avoided.
The one-step rank statement is too early and the displays need correction; none disproves eventual uniform multiplicity stability in the meaningful restriction directions.
Located concern
Lines 1351–1849; Theorem 3.1's last sentence and proof 'Strong stability', final application of \label{remark:strong}.
If the last sentence is read as a separate clause not inheriting finite dimensionality, the proof omits a step: III′ and IV do not imply II for arbitrary infinite multiplicities.
The proper shift on infinitely many trivial copies in F1 shows that this invocation needs a hypothesis. It is not a counterexample with strong inputs.
Specify finite-dimensional inputs if that is the intended clause. Alternatively the following is a new extension argument. For strongly stable algebraic inputs, same-label multiplicity-one branching identifies every stable transition, on each type, with the fixed simple inclusion tensored with a linear map of multiplicity spaces. I–II force that linear map to be an isomorphism. Transport a basis to write each input as a direct sum of copies of simple sequences with those fixed maps. Tensor products distribute over these copies. For a Schur functor of fixed degree d, decompose by the multidegrees in the summand copies: a piece uses at most d copies, and is a natural direct summand of the Schur functor on that finite sum. In such a split system I and II pass to each multidegree via the commuting projections, IV passes by taking the same orbit span, and uniform multiplicities follow from I, II, and same-label branching. Up to relabeling the copies, only finitely many multidegree profiles exist because the set of simple input types is finite. Hence there is a common range. Direct summing the pieces preserves I–II and III′; IV′ on the sum and Proposition 2.14 then give IV. Composing the operations handles Parts 3–5.
No false strong closure claim found for algebraic representations. The finite-dimensional reading is covered by the corrected printed argument; the broader reading needs an additional argument.
Located concern
Lines 1351–1849; §3.1 Non-example 3.3, first construction immediately following \label{thm:classicalSn}.
The stated orbit span in the first construction has one extra copy of the standard representation.
The image of V_n is the line spanned by (1,e_{n+1}). Its tensor square lies in the symmetric part under exchange of the two tensor factors. In the two mixed standard copies its projection therefore spans the diagonal copy, not both copies. Its projections to the trivial and symmetric-square factors are nonzero.
Use V(0)⊕V(1)⊕V(2) for the span at large rank. Keep the second, injective construction: its extra standard input is sent to a mixed monomial in the symmetric-square factor, and the exterior-square projection of the image tensor is zero.
The calculation is wrong, but the failure of tensor orbit generation without strong stability is correctly exhibited by the second construction.
F3
Checked statement: Theorem 3.5 (cancellation of W_n) and Remark 3.6 (recovery from tensor squares), including the representation-ring projections for SL, GL, and Sp, the uniform and pointwise versions, and the omitted finiteness hypothesis.
Returned assessment: claim_false
The ordinary finite-dimensional algebraic representation ring at a fixed rank is a domain: the highest-weight term makes the character map injective, and the character lattice is free abelian, whose integral group ring is a Laurent polynomial domain. This proves uniqueness of a finite solution at one rank when the multiplier is nonzero. It does not compare solutions at different ranks. The particular comparison maps used in the manuscript are not ring homomorphisms for any of the three families. The uniform finite-dimensional conclusions nevertheless have a separate repair. In contrast, pointwise cancellation has finite-dimensional counterexamples for each family, and tensor-square recovery already fails for finite-dimensional pointwise stable squares in SL and GL. On the locally finite interpretation, both uniform statements have counterexamples with infinite multiplicities. The proof's use of the finite representation ring does not itself insert a finite-dimensional hypothesis into the statements.
Premises used
- Fixed-rank highest weights and their characters: version = Fulton–Harris 1991, §23, including the manuscript's Thm. 23.24 and Ex. 23.36(d); available_from = 1991; date_evidence = bibliography FH; date_basis = publication year. Only injectivity of the character map for finite-dimensional algebraic modules is needed; the full theorem identifying its image with the Weyl-invariant subring is unnecessary. The commented-out polynomial-ring description is not an assertion of the rendered proof.
- For the uniform finite-dimensional repair below, use only Theorem 3.1(1)'s fixed-label stable tensor law from LR and the stable symplectic tensor law; version = Fulton–Harris 1991, Eq. 6.7 and Eq. 25.27; available_from = 1991; date_evidence = bibliography FH; date_basis = publication year. No use is made of the false pointwise extension. The Cartan constituent used to bound labels is obtained directly by tensoring two highest-weight vectors.
Missing inputs
- No external text of FH or journal version was supplied. The repaired uniform statement is conditional only on the standard fixed-label tensor laws, not on the manuscript's projections.
- No conclusion is reached here on whether the separate finite-dimensional, nonuniform tensor-square assertion admits a special result for Sp alone. The general Remark 3.6 is already false, and its uniform version as literally stated admits infinite-multiplicity counterexamples.
Scope changes
- New repair of finite uniform cancellation: choose one fixed label of the eventually nonzero W_n. For any label of V_n, its Cartan sum with that fixed label occurs in V_n⊗W_n. In SL and Sp this is ordinary addition of the normalized partitions. In GL it adds the constant tails and adds the separately normalized polynomial partitions. The stable finite support of the product therefore forces the labels of V_n into one finite set. For n and m beyond the finitely many fixed-label tensor thresholds, interpret their decompositions at a common rank t at least as large as both and past all those thresholds. By the fixed-label tensor laws the two product decompositions at t are equal, and their W factors are the same nonzero element. The character-domain argument cancels that element. This proves equality of the decompositions of V_n and V_m. For squares, the Cartan constituent with label twice each input label gives the same finite bound on support. At the common rank use x²=y² and the domain property; x+y≠0 for nonzero actual modules, so x=y, with the zero case immediate. This bypasses, rather than validates, the printed projections.
- This repair gives eventual uniform multiplicity stability only; it gives neither map-sensitive conditions nor simple stability at the first rank at which each label exists.
Located concern
Lines 1850–1891; \label{thm:LRinvert}, statement and opening representation-ring proof; Remark 3.6.
The statements omit a finiteness condition, and cancellation needs an eventually nonzero multiplier.
In each family take W_n countably many copies of Q, and V_n one trivial copy at even ranks and two at odd ranks. W_n and V_n⊗W_n have the same uniform countably infinite trivial multiplicity at every rank, but V_n is not pointwise stable. For the square variant take X_n standard and let E_n be countably many copies of Q⊕(X_n⊗X_n). Add to E_n one copy of X_n at even ranks and two at odd ranks. In large ranks X_n is absent from E_n: use scalar weights for GL, the algebraic center for SL, and −I for Sp. When squaring, the cross terms with E_n already have countably infinite multiplicities on their support, regardless of adding one or two copies, and E_n⊗E_n already contains infinitely many copies of X_n⊗X_n. The square is uniformly multiplicity stable by fixed-label tensor stability, but the original sequence is not. If 'nonzero' were interpreted merely as not identically zero, an eventually zero W_n would also make cancellation impossible.
For the uniform results require finite-dimensional algebraic modules, or equivalently finite multiplicities once uniform finite support has been established for the multiplier and product. Require W_n≠0 throughout an eventual range. Do not regard ∞ as a coefficient in the ordinary representation ring.
The unrestricted uniform statements are false under the manuscript's convention allowing infinite multiplicities.
Located concern
Lines 1850–1891; \label{thm:LRinvert}, paragraph defining π for SL and paragraph beginning 'For G_n=Sp_{2n} Q'.
The rank projections are additive label maps, not the multiplicative maps needed for cancellation.
For SL_{n+1}, π kills ∧^nX_{n+1}; but its tensor with X_{n+1} contains the trivial module because ∧^nX_{n+1}≅X_{n+1}^*. The trivial summand survives π. For Sp_{2n+2}, π kills the simple with label (1^{n+1}), whose tensor square contains the trivial module by self-duality. For GL_{n+1}, π kills ∧^nX_{n+1}, whose final two coordinates differ, but X_{n+1}⊗∧^nX_{n+1} contains ∧^{n+1}X_{n+1}=det, which projects to det in rank n. All surviving contributions are positive, so none cancels the indicated term.
Do not project products at the rank boundary. Use the bounded-support, common-large-rank proof below for the finite uniform statement.
The central inference of the supplied uniform proof is erroneous even with finite dimensions. The highest-weight sum does verify the distinct claim that a disappearing high-length label produces a high-length constituent in its product with any nonzero module; that does not make π multiplicative.
Located concern
Lines 1850–1891; \label{thm:LRinvert}, final paragraph beginning 'We sketch a compactness argument'.
The pointwise cancellation theorem is false for finite-dimensional modules as well. The proposed degree truncation is not valid.
Let V_n alternate between Q and Q² with trivial action. Choose W_n=X_n^* in SL and GL, and W_n=Sym^n(X_n) for the standard 2n-dimensional X_n in Sp. These W_n are irreducible, nonzero in every rank, and their labels escape every fixed label: (1^{n−1}), (0^{n−1},−1), and (n), respectively. Both W_n and V_n⊗W_n are pointwise multiplicity stable with every fixed multiplicity eventually zero. The trivial multiplicity of V_n alternates. Also, large partitions do not span the claimed ideal: in SL a high-degree dual pair has trivial tensor constituent, and in Sp every high-degree simple is self-dual. Even a polynomial degree quotient that is an algebra quotient can have zero divisors and can kill the multiplier.
Remove the general pointwise clause. Use the finite uniform result only where that stronger input is actually supplied.
The failure is not repaired by adding finite dimensionality or by clarifying 'nonzero'. It occurs for every listed family.
Located concern
Lines 1850–1891; unlabeled Remark 3.6, first sentence and the largest/next-largest algorithm.
The nonuniform tensor-square assertion is false, and the highest-label algorithm does not justify passage to the pointwise limit.
For SL or GL take V_n=Q for even n and the trace-zero adjoint module for odd n, in ranks at least 3. Each square has exactly one trivial constituent, by irreducibility and self-duality of the adjoint module. In SL only weights with trivial algebraic central character occur in the square, so any nonempty fixed partition λ with n>|λ| is absent. This central-character test can be made after extending scalars to an algebraic closure. For GL all scalar matrices act trivially, while a fixed full weight with constant tail c has total sum nc plus a fixed integer. A nonzero tail eventually gives a nonzero sum, and with zero tail a nonempty polynomial partition also gives a nonzero sum. Thus the square is pointwise multiplicity stable and V_n is not. At a fixed rank an algorithm must first subtract the full contribution of the highest summand squared before reading the next residual highest term; the next constituent of the raw square need not be the sum of the two largest input labels.
Retain the finite uniform square result with the common-rank argument below. For a fixed-rank constructive formulation, recover the top multiplicity from its square, subtract products of known terms, and read the next largest residual Cartan term with its cross multiplicity. Do not claim a pointwise-limit argument from that procedure.
The nonuniform assertion in the remark fails already for finite-dimensional SL and GL representations. A fixed-rank algorithm or a finite uniform theorem does not give that assertion.
Outside the audit
- Targets outside F1–F3 were not independently refereed. The Lie-algebra results are cited only to distinguish which foundational interface is used; their complete chain-map and range arguments were not checked.
- The external papers and books, exact manuscript-version histories of imported theorems, publication comparison, and journal oblatum/received date were not inspected. The cited BOR preprint version containing the sharp bound remains to be identified.
- The finite-dimensional pointwise square-recovery problem for Sp as an isolated special case was not settled here. The general assertion in Remark 3.6 is refuted without settling that special case.
- No later conjectural resolution, modern reformulation, or journal correction was used as historical evidence.
Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.