REVIEW 2 major objections 3 minor 10 references
A remark on automorphisms of tensor spaces
T0 review · 2 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read Every graph's automorphism group is realized as the full symmetry group of a symmetric tensor space, and the same construction yields tensor spaces whose square has infinite length as a module over their own symmetries.
desk verdict Nice note with a genuine gap in Proposition 4.1(a): the character-uniqueness claim is false, so Theorem 1.3's proof needs repair, though the fix looks straightforward. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the diagonal form $f = \sum_{i \in I} x_i^d$ on the basis $\{e_i\}_{i \in I}$, with $d \geq 3$. Proposition 2.1 identifies its automorphism group with the wreath product $\mu_d \wr S_I$; the proof is the observation that the directional derivative $\partial_v f = d \sum a_i x_i^{d-1}$ is a power of a linear form exactly when $v$ is a scalar multiple of some $e_i$, so every symmetry must permute the basis lines. Adding the degree-two and degree-three diagonal forms $g_2$ and $g_3$ kills the scalar factors and yields exactly $S_I$, after which the relation forms $f_i$ encode the structure being represented. The variant construction repeats each coordinate $m$ times, which changes the scalar group from $\mu_d$ to $\mu_m$ and then uses the same derivative argument to obtain $\mu_m \wr \operatorname{Aut}(X)$.
What would settle it
Take the simplest graph $X$ with two vertices and one edge and write down the three forms from Theorem 3.1: the relation form $f_1$ equals zero (no arity-3 relation), and $g_2, g_3$ are the standard diagonal forms on $\mathbb{C}^2$. Compute the full subgroup of $\mathrm{GL}_2(\mathbb{C})$ fixing $g_2$ and $g_3$; if any transformation other than the identity and the transposition of basis vectors fixes both forms, then the claimed equality $\operatorname{Aut}(V) \cong \operatorname{Aut}(X)$ is false. The same check can be run for any small graph by solving the polynomial equations imposed on a $2 \times 2$ or $3 \times 3$ matrix.
Extended reading notes
Core claim
The paper's central claim is a flexible realization result: the full symmetry group of a finite collection of multilinear forms can be prescribed, up to the natural wreath structure, as the automorphism group of an arbitrary relational structure. Theorem 3.1 shows that the forms $(f_1, \ldots, f_r, g_2, g_3)$ attached to a $d$-structure $X$ have symmetry group exactly $\operatorname{Aut}(X)$; Theorem 3.2 shows that the variant forms $(f'_1, \ldots, f'_r, g_m)$ have symmetry group $\mu_m \wr \operatorname{Aut}(X)$. The representation-theoretic payoff is that one can build spaces with prescribed orbit behavior: for the $\mathbb{Z}$-indexed path graph, $\mu_3 \wr \mathbb{Z}$ acts irreducibly on $V$ but $V^{\otimes 2}$ splits into infinitely many irreducibles, and the hypergraph construction in Example (e) tunes the threshold $m$. The paper is explicit that this answers a question suggested by earlier examples in which finite length of $V$ seemed to force finite length of all tensor powers.
Load-bearing premise
The whole construction hinges on the fact that for $d \geq 3$ the only directions in which the derivative of the diagonal form $\sum x_i^d$ is a pure power of a linear form are the coordinate directions; if this fails (as it does for $d = 2$), the wreath-product description of $\operatorname{Aut}(f)$ collapses and the realization arguments no longer go through.
Editorial extensions
If this is right
- Automorphism groups of graphs, ordered sets, vector spaces over finite fields, Rado graphs, and Fraïssé limits all occur as $\operatorname{Aut}(V)$ of a symmetric tensor space, so the class of tensor-space symmetries is as wide as the class of permutation groups from model theory.
- Wreath products $\mu_m \wr \Gamma$ are also realized, so one can arrange for scalar roots-of-unity factors to coexist with any permutation group in the symmetry group.
- Irreducibility of $V$ does not imply finite length for $V^{\otimes 2}$; the $\mu_3 \wr \mathbb{Z}$ example gives a compact counterexample.
- For every $m \geq 3$ there is a tensor space whose tensor-power lengths break exactly at the $m$-th power, showing the phenomenon can be tuned rather than being a single pathology.
- Any definition of 'large' automorphism group for tensor spaces that is meant to yield finite-length representations in tensor powers must require more than finite length of $V$ itself; the paper points toward linear oligomorphy as the natural candidate.
Reading between the lines
- The diagonal-form lemma is likely characteristic-sensitive: over a field of characteristic dividing $d$, powers of linear forms are less distinguishable, so the realization theorem may need forms of higher degree or adjusted coefficients; a natural next test is the same construction over finite fields of small characteristic.
- Because the construction converts any finite-relational structure into a tensor space with the same automorphism group, it suggests a translation dictionary between model-theoretic notions (Fraïssé limits, oligomorphy, homogeneous structures) and tensor-space representation theory; under that dictionary, model-theoretic constructions could yield new tensor-space phenomena automatically.
- One could test whether the pathological examples still have well-behaved higher structure, for instance whether the category generated by tensor powers of $V$ is locally finite or has a Krull–Schmidt property even when some $V^{\otimes k}$ has infinite length; the paper does not address this.
- A concrete extension would be to compute, for the $\mu_3 \wr \mathbb{Z}$ example, the explicit decomposition of $V^{\otimes 2}$ into irreducible $G$-modules; this would make the transition from irreducible $V$ to infinite-length square visible in coordinates, and might suggest invariants that predict the break.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies tensor spaces, i.e., complex vector spaces equipped with a finite collection of multilinear forms, and realizes a large class of permutation groups as automorphism groups of such spaces. Theorem 1.1 states that for any graph X, there is a symmetric (3,2,2)-space V whose automorphism group is isomorphic to Aut(X) acting naturally on a basis indexed by X. Theorem 3.1 generalizes this to arbitrary finite-relational structures, and Theorem 3.2 gives a variant whose automorphism group is a wreath product μm ≀ Γ. Using these constructions, the paper produces two pathological examples: an irreducible V whose tensor square has infinite length (Theorem 1.2), and a family of spaces where V⊗k has finite length precisely for k < m (Theorem 1.3). The proofs are short and self-contained, relying on the determination of the automorphism group of the diagonal form (Proposition 2.1) and on a representation-theoretic lemma about tensor powers of permutation modules (Proposition 4.1).
Significance. If the results are correct, they provide a substantial enlargement of the known classes of automorphism groups of tensor spaces and give counterexamples to the natural intuition that finite length of V as a representation of its automorphism group should force finite length of its tensor powers. The constructions are elegant and concise, and the paper is written in a clear style. The main theorems are precisely stated and, apart from the issues described below, the arguments are sound. The paper is self-contained and does not rely on unstated prior work; the cited literature is used for motivation and comparison only. These strengths make the paper a useful contribution to the developing theory of infinite-dimensional tensor spaces, provided the gaps in the proofs of Propositions 4.1 and Theorem 3.1 are repaired.
major comments (2)
- [Section 4, Proposition 4.1(a)] The proof contains a false assertion: it claims that for k < m, if x and y are distinct elements of X^k then the characters α_x and α_y of μ_X^m on the basis vector e_x are different. This is not true, since α_x depends only on the multiset of coordinates of x. For example, when k=2, the tuples (a,b) and (b,a) have the same character. Consequently, V⊗k is not multiplicity-free as a μ_X^m-module, and the subsequent argument that each orbit block W_j is irreducible fails. This is not a purely cosmetic issue: for Γ = S_X acting on a countably infinite set X, with m=3 and k=2, the off-diagonal orbit block contains both (a,b) and (b,a) and splits into at least the symmetric and alternating parts. Since Proposition 4.1(a) is used to prove the finite-length statement for k < m in Theorem 1.3, the proof of Theorem 1.3 is incomplete as written. The proposition may still be true and a repair could likely be obtained by using the finite symmetric group S_k acting on tensor factors, but that argument is not present and needs to be supplied.
- [Section 3, Theorem 3.1] The proof states that "By Proposition 2.1, the subgroup of GL(V) preserving g_k is μ_k ≀ S_X" for k = 2 as well. This is incorrect: Proposition 2.1 requires d ≥ 3, and for k=2 the stabilizer of the symmetric bilinear form g2 is the full orthogonal group O(V), which is much larger than μ_2 ≀ S_X. The conclusion that the common stabilizer of g2 and g3 is S_X is nevertheless true: by Proposition 2.1 the stabilizer of g3 is μ_3 ≀ S_X, and requiring this subgroup to preserve g2 forces λ_x^2 = λ_x^3 = 1, hence λ_x = 1 for all x. However, the proof as written is not correct and should be amended. The same issue affects Remark 3.3, where the m=2 variant uses g2.
minor comments (3)
- [Section 5(e)] The sentence "choosing an isomorphism τ : X → σ∗(X)" appears to contain a typo: it should refer to an isomorphism τ : Y → σ∗(Y), since σ∗(X) has not been defined and the argument concerns the structure Y. The following sentence should also say that (τ, σ) is an automorphism of the structure X, not of the structure Y.
- [Section 2, Proposition 2.1] The proof asserts without justification that if v has at least two nonzero coefficients with respect to the basis {e_i}, then ∂_v f is not a power of a linear form. This is true for d ≥ 3 but not completely immediate; a short explanation would improve the exposition.
- [Section 5(d)] The claim that V is an irreducible representation of G = μ_3 ≀ Z is stated without proof. It follows because the μ_3^Z-weights of the basis vectors are all distinct, but this step is not spelled out and would be helpful for the reader.
Circularity Check
No significant circularity; the constructions are self-contained, and prior self-citations are motivational only.
full rationale
The central results are proved by explicit constructions that do not reduce to their inputs. Proposition 2.1, the key lemma, is proved from first principles: it computes Aut(f) for the diagonal form f = sum x_i^d by observing that a partial derivative is a power of a linear form only along basis directions. No prior result is invoked there. The main theorems (1.1, 1.2, 1.3) then encode arbitrary relational structures into multilinear forms and use Proposition 2.1 to identify the full automorphism group of the resulting tensor space. The group is not defined as the automorphism group of the structure, and the equality is not assumed; it is derived. Self-citations to [DS1, HS1, HS2, BDDE, BDS] appear only in the introduction as motivation and context, e.g., 'In [HS1, HS2], Harman and the second author constructed tensor spaces V with very large automorphism groups G', and as comparison for the new phenomena. These citations are not load-bearing for the proofs. No parameter is fitted and no prediction is renamed from a fit. The skeptical concern about Proposition 4.1(a) — that the character alpha_x depends only on the multiset of coordinates, so distinct tuples can share a character — identifies a possible correctness gap in the proof of finite length for k < m, but that is not a circularity: the statement is not assumed, and the gap does not involve reducing a prediction to an input. Thus the paper's derivation chain is not circular.
Assumptions & free parameters
assumptions (3)
- standard math Axiom of Choice, or an equivalent, to ensure existence of bases for arbitrary vector spaces.
- standard math Fraïssé limit theorem for countable relational structures.
- standard math Standard facts about wreath products and multilinear algebra.
Cite this review
Pith. "Pith review of A remark on automorphisms of tensor spaces." pith.science (2026). https://pith.science/paper/T3M67SEW
@misc{pith2026250713521,
author = {Pith},
title = {Pith review of: A remark on automorphisms of tensor spaces},
year = {2026},
howpublished = {\url{https://pith.science/paper/T3M67SEW}},
note = {Machine review of arXiv:2507.13521}
}
read the original abstract
A tensor space is a vector space equipped with a finite collection of multi-linear forms. In recent years, a rich theory of infinite dimensional tensor spaces has emerged. In this note, we show that a large class of permutation groups can occur as the automorphism groups of such tensor spaces. Using this, we show that a tensor space can behave somewhat pathologically as a representation of its automorphism group.
Reference graph
Works this paper leans on
- [1]
-
[2]
Arthur Bik, Jan Draisma, Rob H. Eggermont, Andrew Snowden. The geometry of polynomial representations. Int.\ Math.\ Res.\ Not.\ IMRN 2023, no. 16, pp. 14131--14195. doi:10.1093/imrn/rnac220 arXiv:2105.12621
arXiv 2023
-
[3]
Isogeny classes of cubic spaces
Arthur Bik, Alessandro Danelon, Andrew Snowden. Isogeny classes of cubic spaces. arXiv:2207.13951
-
[4]
Oligomorphic permutation groups
Peter J.\ Cameron. Oligomorphic permutation groups. London Mathematical Society Lecture Note Series, vol. 152, Cambridge University Press, Cambridge, 1990. doi:10.1017/CBO9780511549809
-
[5]
Biquadratic spaces of length two
Alessandro Danelon, Andrew Snowden. Biquadratic spaces of length two. arXiv:2412.20681
-
[6]
Isogeny classes of tensor spaces
Alessandro Danelon, Andrew Snowden. Isogeny classes of tensor spaces. In preparation
-
[7]
Ultrahomogeneous tensor spaces
Nate Harman, Andrew Snowden. Ultrahomogeneous tensor spaces. Adv.\ Math. 443 (2024). doi:10.1016/j.aim.2024.109599 arXiv:2207.09626
arXiv 2024
-
[8]
Tensor spaces and the geometry of polynomial representations
Nate Harman, Andrew Snowden. Tensor spaces and the geometry of polynomial representations. arXiv:2407.19132
Show all 10 references
-
[9]
Properties of high rank subvarieties of affine spaces
David Kazhdan, Tamar Ziegler. Properties of high rank subvarieties of affine spaces. Geom.\ Funct.\ Anal. 30 (2020), pp. 1063--1096. doi:10.1007/s00039-020-00542-4 arXiv:1902.00767
2020 arXiv
-
[10]
A survey of homogeneous structures
Dugald Macpherson. A survey of homogeneous structures. Discrete Math. 311 (2011), no. 15, pp. 1599--1634. doi:10.1016/j.disc.2011.01.024
2011 doi
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.