{"id":"61e9e8b4-8912-40f2-aa9a-7363d51ac2f0","arxiv_id":"2507.13521","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Any graph automorphism group is realized as the automorphism group of a symmetric (3,2,2)-tensor space, and the paper constructs spaces where V is irreducible but V⊗2 has infinite length.","lead":"This paper builds vector spaces with extra multilinear structure whose symmetry groups can be any symmetry group of a graph, and uses this to produce examples where a space is irreducible but its tensor square has infinite length. The constructions give new intuition about when infinite-dimensional tensor spaces behave like finite-dimensional ones.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prop. 4.1(a) relies on a false character-uniqueness claim; Theorem 1.3's proof needs repair before acceptance.","rationale":"The paper's main automorphism-group constructions, especially Theorem 3.1 and Proposition 2.1, appear sound. The single genuine gap I found is the false uniqueness statement inside Proposition 4.1(a), which is load-bearing because Theorem 1.3 explicitly uses Proposition 4.1(a) to prove finite length of V⊗k for k<m. The defect is real: in any self-paired orbit, distinct tuples with the same multiset of coordinates give equal characters, so the irreducibility argument collapses. However, the proposition's conclusion may still be true; the finite group S_k permuting tensor factors commutes with G and should allow a finite decomposition of each orbit block. The minimal example Γ=S_X, k=2, m=3 exhibits the failure of the proof but appears to have finite length, suggesting the gap is repairable. I therefore recommend conditional acceptance rather than rejection or unchanged acceptance: the theorem may stand, but the proof as written needs correction. The reader's weakest_assumption identified Proposition 2.1, not Proposition 4.1(a), so I disagree with the reader on where the main risk lies.","tokens_in":5104,"tokens_out":27431,"duration_ms":363368,"concrete_test":"Reprove Proposition 4.1(a) using the S_k-action on V⊗k: S_k commutes with G=µ_m≀Γ and is finite, so V⊗k splits into finitely many S_k-isotypic components; verify that each component of a single Γ-orbit block has finite length by using distinctness of S-characters on S_k-orbits. As a minimal check, take Γ=S_X on a countably infinite X, k=2, m=3, and compute the length of V⊗2: confirm the off-diagonal block decomposes into exactly two irreducibles, the symmetric and alternating parts. If the corrected proof succeeds, Theorem 1.3 stands and only the proof needs revision; if any component has infinite length, Proposition 4.1(a) is false and Theorem 1.3 is unsupported.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Proposition 4.1(a), the proof asserts that 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 differ. This is false: α_x is the product of the λ-coordinates occurring in x, so it depends only on the multiset of coordinates of x. For k=2, the tuples (a,b) and (b,a) have the same character. The proof then uses this uniqueness to isolate a single basis vector from any nonzero vector in an orbit block W_j, concluding that each W_j is irreducible. That conclusion fails in self-paired orbits: for Γ=S_X acting on a countably infinite X with k=2 and m=3, the off-diagonal block contains both (a,b) and (b,a), and it splits into at least the symmetric part and the alternating part. Since Proposition 4.1(a) is invoked to obtain finite length of V⊗k for k<m in Theorem 1.3, the argument for Theorem 1.3 is incomplete as written. The proposition's conclusion may still be true, and the finite symmetric group S_k acting on tensor factors likely gives a finite decomposition that repairs the proof, but the stated proof does not establish it.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","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).","tokens_in":5355,"tokens_out":19211,"duration_ms":250059,"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":[{"comment":"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":"Section 4, Proposition 4.1(a)"},{"comment":"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.","section":"Section 3, Theorem 3.1"}],"minor_comments":[{"comment":"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":"Section 5(e)"},{"comment":"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":"Section 2, Proposition 2.1"},{"comment":"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.","section":"Section 5(d)"}],"recommendation":"major_revision","confidential_remarks":"The two major comments are both local and repairable, and the central ideas of the paper appear sound. The paper is short and well-written; I would be inclined to accept after a revision that fixes the proof of Proposition 4.1(a) and the misapplication of Proposition 2.1 in Theorem 3.1. The referee should check carefully that the repaired proof of Proposition 4.1(a) indeed yields finite length under the stated hypothesis, since the current argument is not merely incomplete but relies on a demonstrably false uniqueness claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a good, readable note with real new examples, but there is a hole in the proof of Proposition 4.1(a), and it matters for Theorem 1.3. The gap is repairable, but the paper as written overstates what it proves.\n\nWhat is genuinely new: Theorem 1.1 realizes any graph automorphism group as the full symmetry group of a symmetric (3,2,2)-space, with the natural basis action. That is a clean, useful observation. More importantly, the paper gives the first examples where finite length of V does not force finite length of tensor powers: Theorem 1.2 has V irreducible but V⊗2 of infinite length, and Theorem 1.3 gives an exact cutoff for all k < m. The constructions in Sections 2 and 3 are elementary and sound. Proposition 2.1 is correct, and the use of the diagonal form plus the structure relations is elegant.\n\nThe soft spot is Proposition 4.1(a). The proof claims that if x and y are distinct elements of X^k and k < m, then the characters of µ_X^m on the basis vectors e_x and e_y are distinct. That is false: the character depends only on the multiset of coordinates, so (a,b) and (b,a) have the same character. Consequently the proof's claim that each orbit block W_j is irreducible fails when the Γ-orbit identifies tuples that differ only by coordinate permutation. For a concrete example, take Γ = S_X on a countably infinite X with k = 2 and m = 3: the off-diagonal block splits into symmetric and alternating parts. So the proof of finite length in part (a) is incomplete. The conclusion may still be true — one could bring in the finite group S_k acting on tensor positions to decompose each orbit block into finitely many pieces — but that argument is not in the paper. This directly affects Theorem 1.3's claim about finite length for k < m. The rest of the paper, including Theorems 1.1 and 1.2, stands.\n\nThe reader's soundness score of 9 is too generous. This is a genuine, load-bearing gap, though likely patchable. I would send the paper to peer review; a competent referee can push the authors to fix Proposition 4.1. It is a worthwhile note for anyone working on tensor spaces or oligomorphic groups, and the examples deserve to be on record.","headline":"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.","tokens_in":5852,"tokens_out":3214,"would_cite":false,"duration_ms":40247,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20B27","03C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"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.","keywords":["tensor spaces","multilinear forms","automorphism groups","permutation groups","diagonal forms","wreath products","Fraïssé limits","oligomorphic groups"],"falsifier":"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.","tokens_in":4945,"feed_emoji":"🔀","tokens_out":8655,"duration_ms":92543,"temperature":0.7,"pith_summary":"Tensor spaces are vector spaces carrying finitely many multilinear forms, and this note asks which permutation groups can appear as their automorphism groups. The main theorem says that given any graph $X$, one can build a symmetric $(3,2,2)$-space $V$, with basis indexed by the vertices of $X$, such that $\\operatorname{Aut}(V)$ is exactly $\\operatorname{Aut}(X)$, acting by permuting basis vectors. The general version realizes the automorphism group of any structure over a finite relational language, and a variant realizes wreath products $\\mu_m \\wr \\Gamma$. The paper then constructs two families of examples: a symmetric $(6,3)$-space of countable dimension on which $\\operatorname{Aut}(V)$ acts irreducibly on $V$ but $V^{\\otimes 2}$ has infinite length, and, for every $m \\geq 3$, a space for which $V^{\\otimes k}$ has finite length exactly when $k < m$. If these theorems are right, finite length of a tensor space as a representation of its automorphism group does not control the lengths of its tensor powers, and any theory of tensor spaces with 'large' symmetry groups needs a stronger condition.","feed_headline":"Every graph symmetry group appears as a tensor space automorphism group","feed_subtitle":"New tensor spaces make V irreducible while its square has infinite length, showing finite length does not control tensor powers.","key_machinery":"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 $\\mu_m \\wr \\operatorname{Aut}(X)$.","core_discovery":"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.","pith_inferences":["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."],"forward_implications":["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."],"supporting_citations":[{"why":"Introduced the linearly oligomorphic condition and earlier tensor spaces whose tensor powers all have finite length, the context that made the pathological examples surprising.","marker":"[HS1]"},{"why":"Developed the geometric theory of tensor spaces with large automorphism groups, the program to which the authors supply new examples.","marker":"[HS2]"},{"why":"Classified length-two symmetric $(2,2)$-spaces, the prior result that reinforced the suspicion that finite length of $V$ controls tensor powers.","marker":"[DS1]"},{"why":"Supplies the Fraïssé limit construction and oligomorphic permutation group background used in Examples (c) and (e).","marker":"[Cam]"},{"why":"Provides the survey of homogeneous structures underlying the Fraïssé-limit facts quoted in the examples.","marker":"[Mac]"}],"fun_headline_variants":["Tensor spaces realize any relational structure's automorphism group","Automorphism groups of tensor spaces: any relational structure works","Tensor spaces: prescribed automorphism groups with pathological representations","Prescribed automorphism groups in tensor spaces, even pathological ones"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"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.","fun_headline_variants_meta":{"raw":{"variants":["Tensor spaces realize any relational structure's automorphism group","Automorphism groups of tensor spaces: any relational structure works","Tensor spaces: prescribed automorphism groups with pathological representations","Prescribed automorphism groups in tensor spaces, even pathological ones"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002183,"raw_usage":{"total_tokens":8402,"prompt_tokens":832,"completion_tokens":7570,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":448,"completion_tokens_details":{"reasoning_tokens":7504}},"tokens_in":448,"tokens_out":7570,"duration_ms":62752,"temperature":1.0,"reasoning_tokens":7504,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:22:47.671668+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"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.","supporting_citations":[],"review_version":1}