{"id":"04136e35-bf20-4bdb-992d-66edcec7eec0","arxiv_id":"2411.17434","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"One generic complex orbit determines a finite linear symmetry group up to isomorphism; two generic real orbits suffice, and concrete recovery needs an orbit count governed by representation multiplicities.","lead":"The paper proves that one generic orbit of a finite group acting on a complex vector space reveals the group up to isomorphism, and two generic orbits do the same for real spaces. It also gives representation-theoretic formulas for how many orbits are needed to recover the group as explicit matrices.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 2.12's proof is incomplete: a real isometry of the union of two real orbits need not preserve the complexified orbit unless the two orbit-actions are synchronized, and the proof asserts this without proving it.","rationale":"The reader's weakest-assumption analysis correctly identifies the central gap: the proof of Theorem 2.12 defines an injection into Aut(G(v+iw)) without proving that the complexified isometry preserves the complex orbit. This is directly load-bearing because Theorem 2.12 is one of the paper's main abstract-recovery results over the reals. The concern is not merely pedantic: if the two real orbits are not synchronized, α(σ) may send a point g(v+iw) = gv+igw to σ(gv)+iσ(gw), which need not be of the form h(v+iw). Lemma 3.1 is the natural repair, but it is introduced later and, in its current form, only proves that σ is determined by its action on Gv, not that this action is induced by an element of G. I considered the other candidate concern, the non-sharpness of Corollary 3.5 exposed by Example 3.8; that is explicitly acknowledged in the text and affects the framing of 'sharp bounds' rather than the validity of a central theorem. The Theorem 2.12 gap, by contrast, affects a central theorem and is not acknowledged. The paper's results are plausible and the gap is probably repairable, so a conditional verdict is appropriate; no change to the reader's decision is needed.","tokens_in":14090,"tokens_out":24568,"duration_ms":248869,"concrete_test":"Independently re-derive Theorem 2.12 using Lemma 3.1: for generic (v,w), let β:Gv→Gw be the nearest-point bijection and write β(gv)=g c w. Check whether the equivariance condition β(σ(gv)) = σ(β(gv)), together with norm separation and genericity, implies that if σ(gv) = h_g gv then σ(gw) = h_g gw. If yes, insert this argument before defining α and the proof is repaired; if no, α cannot be shown to map G(v+iw) to itself and the theorem is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Theorem 2.12, the proof assumes G ≅ Aut(G(v+iw)) for generic (v,w) and defines α(σ)(x+iy) = σ(x)+iσ(y) for x+iy ∈ G(v+iw). The line 'Since σ ∈ Aut(Gv∪Gw), it extends to a linear isometry M, hence α(σ) extends to the linear isometry x+iy ↦ Mx+iMy' only shows that α(σ) is a linear isometry of the complexification; it does not prove that α(σ) maps the finite set G(v+iw) to itself. For this, writing σ(gv) = h_g gv would require σ(gw) = h_g gw for the same h_g, i.e., the induced permutations on the two real orbits must be synchronized. The proof does not establish this. Lemma 3.1, introduced later, provides an equivariant nearest-point bijection that would supply the needed synchronization, but Theorem 2.12 neither cites nor proves it. Moreover, Lemma 3.1 only guarantees that an automorphism of the union is determined by its action on the first orbit; it does not directly prove that this action lies in the image of the canonical map from G. Thus the two-orbit theorem, a central claim, is not proven as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the inverse problem of recovering an unknown finite group G of automorphisms (linear isometries) of a finite-dimensional real or complex Hilbert space V from one or more generic G-orbits. In the complex case, it proves (Theorem 2.6) that a single generic orbit determines G up to isomorphism, and that the canonical map G to Aut(Gv) is an isomorphism. In the real case, it proves (Theorem 2.12) that two generic orbits determine G up to isomorphism via the automorphism group of their union. For concrete recovery, Theorem 3.4 and Corollary 3.5 give a representation-theoretic bound on the number of generic orbits needed to determine G as a subset of Aut(V), together with a converse. The paper also provides examples, open problems, and a table of bounds.","tokens_in":14330,"tokens_out":20822,"duration_ms":180348,"significance":"The problem is natural and well motivated by symmetry learning in data science. The complex one-orbit theorem and the representation-theoretic concrete recovery bound are elegant and appear correct; the derivations use generic polynomial arguments and Schur's lemma with no free parameters or fitting. The paper also offers instructive examples showing real-orbit behavior (e.g., Example 2.11) and a counterexample to possible over-strong sharpness (Example 3.8). The main caveat is that the proof of the two-orbit theorem is incomplete as written; however, the gap is local and can be repaired using the paper's own orbit-pairing lemma. If that repair is made, the results constitute a solid contribution.","major_comments":[{"comment":"The proof defines alpha(sigma)(x+iy)=sigma(x)+i sigma(y) and argues that because sigma extends to a linear isometry M, alpha(sigma) extends to the linear isometry x+iy maps to Mx+iMy and hence belongs to Aut(G(v+iw)). This only shows alpha(sigma) is a linear isometry of the complexification; it does not show that alpha(sigma) maps the finite set G(v+iw) into itself, as required by the definition of Aut. One needs synchronization: for each g in G, the same group element h_g must satisfy sigma(gv)=h_g gv and sigma(gw)=h_g gw. This synchronization follows from Lemma 3.1 because beta(gv)=gw and sigma(gw)=beta sigma(gv), but Lemma 3.1 is neither stated nor cited before Theorem 2.12. The proof should be reorganized to use Lemma 3.1 (or to prove the synchronization directly) before constructing alpha; with that addition the remainder of the argument is valid.","section":"Section 2, Theorem 2.12 (proof)"},{"comment":"The proof is only a sketch. The sentence 'we cannot determine whether G acts trivially on the orthogonal complement' does not verify the claim that every combination of k orbits can be realized as orbits of another subgroup. A construction is needed: for arbitrary v_1,...,v_k put S=span(union_i Gv_i); since k fails (1), Theorem 3.4 gives codim S >= r >= 1. Choose a nonzero u in S^perp and let R be the orthogonal reflection in the line through u (fixing u^perp). Then H={g direct sum r : g in G, r in <R>} is a proper subgroup of Aut(V) different from G, and H v_i = G v_i for all i. Inserting this construction would make the lower bound rigorous.","section":"Section 3, Corollary 3.5(a)"}],"minor_comments":[{"comment":"The displayed equation in the injectivity argument contains a typo: the equality should involve alpha(sigma)(g(v+iw)) = sigma(gv) + i sigma(gw), not alpha(sigma)(v+iw) on the right-hand side.","section":"Section 2, Theorem 2.12"},{"comment":"The chain proving beta sigma = sigma beta for arbitrary sigma implicitly assumes sigma^{-1}(Gw)=Gw, which is established only later in the proof via the norm inequality ||v|| != ||w||. The proof should establish the norm inequality first, or explicitly note that for the G-equivariance step the chain is applied only to sigma in G.","section":"Section 3, Lemma 3.1"},{"comment":"The phrase 'sharp bounds' overstates the real abstract-recovery case: the paper proves an upper bound of 2 and leaves Conjecture 2.7 (one orbit suffices) open. Consider saying 'we give sharp bounds in the complex case and bounds in the real case.'","section":"Abstract and Table 1"},{"comment":"The assertion that the label-preserving automorphism group of the complete Cayley graph is isomorphic to G is stated without proof; a one-line verification (phi(h)=phi(1)h follows from label preservation on edges from 1 to h) would make the proof self-contained.","section":"Theorem 2.6(a)"},{"comment":"In the definition of Aut(S), the permutation is first called pi and then sigma in 'M|S=sigma'; the notation should be unified.","section":"Section 1.2"},{"comment":"The wording 'Corollary 3.5 reports that k orbits determine G as a concrete group only if k>=d' is slightly imprecise because Corollary 3.5(b) is a sufficient condition and part (a) concerns arbitrary combinations; the genericity caveat in the following sentence is important and should be integrated into the phrasing.","section":"Example 3.8"}],"recommendation":"major_revision","confidential_remarks":"For the editor: the main weakness is the proof of Theorem 2.12, which is fixable by moving Lemma 3.1 before the theorem or citing it. I see no signs of circularity or unsupported empirical claims. The paper fits the journal's scope and should be acceptable after a moderate revision that fills the two proof gaps identified in the major comments."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this paper proves a new one-orbit theorem over C and a plausible two-orbit theorem over R, but the proof of the two-orbit theorem has a genuine gap that the authors can fix with their own later lemma. The paper deserves a serious referee, not a desk reject.\n\nWhat's genuinely new: Theorem 2.6 is a clean result — for a finite group acting unitarily on a complex Hilbert space, one generic orbit determines the group up to isomorphism. The argument via the Gram graph and its automorphism group is simple and effective. The concrete recovery bound in Corollary 3.5 is also new, and the use of the regular representation to characterize when orbits span a large subspace (Theorem 3.4) is a nice piece of representation theory. The examples, like the Q8 16-cell orbit, are well chosen.\n\nThe soft spots are real but manageable. The proof of Theorem 2.12 asserts that an automorphism σ of the union of two real orbits complexifies to an automorphism of the complexified orbit G(v+iw). As written, that only shows α(σ) is a linear isometry of the complexification, not that it preserves the finite set. You need the two orbit-actions to be synchronized — exactly what Lemma 3.1 later provides with the equivariant nearest-point bijection. The theorem doesn't cite or prove that lemma. That's a fixable gap: move Lemma 3.1 before Theorem 2.12 and use it. Also, the abstract and Table 1 call the bounds \"sharp,\" but Example 3.8 explicitly shows a single generic orbit can determine the concrete group when Corollary 3.5(b) demands k≥d. So the bounds are sufficient, not necessary for generic orbits; \"sharp\" is overstated. And Corollary 3.5(a) is sketched in a way that needs more care — the argument that any other subgroup could realize the orbits is hand-wavy. None of this undermines the main claims; it just needs a revision.\n\nThe paper is for someone in symmetry learning or representation-theoretic identifiability. The complex one-orbit theorem is a nice, citable result. I'd send it to review with a request to fix the two-orbit proof and soften the sharpness claims. A serious referee would catch these and the paper would come back stronger.","headline":"Genuinely new one-orbit theorem over C, a plausible two-orbit theorem over R with a fixable proof gap, and an overstated sharpness claim; worth reviewing after revision.","tokens_in":14845,"tokens_out":2938,"would_cite":true,"duration_ms":27449,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"One generic orbit of an unknown finite unitary group on a complex Hilbert space determines the group up to isomorphism, and two suffice on real Hilbert spaces.","keywords":["finite group recovery","generic orbits","Hilbert space automorphisms","Gram graph","Cayley graph","representation theory","orbit recovery","symmetry learning"],"falsifier":"Sample a random pair of orbits of the quaternion group $Q_8$ acting by left multiplication on $\\mathbb{R}^4$ and count the automorphisms of their union; Theorem 2.12 predicts exactly 8 for every pair in its generic set, so finding a pair with more symmetries, particularly one that preserves distances but is not a single complexified isometry, would disprove the two-orbit theorem.","tokens_in":13874,"feed_emoji":"🧩","tokens_out":8363,"duration_ms":75979,"temperature":0.7,"pith_summary":"This paper asks what an unknown finite group of symmetries of a finite-dimensional Hilbert space can be recovered from a small number of its orbits, with 'generic' meaning open and dense. On a complex Hilbert space one generic orbit is enough: the inner products among points of the orbit form an edge-labeled graph isomorphic to the group's complete Cayley graph, so its label-preserving automorphism group is isomorphic to the original group. On a real Hilbert space two generic orbits suffice to identify the isomorphism class, and one suffices in special cases, such as prime order or dimension two. If the goal is not just the isomorphism class but the concrete set of unitary or orthogonal transformations, the paper gives a sharp representation-theoretic bound on the number of generic orbits needed, with matching lower and constructive upper bounds.","feed_headline":"One generic orbit reveals a finite group's isomorphism class","feed_subtitle":"Complex spaces need just one orbit; real spaces need two. A sharp bound covers recovering the exact set of transformations.","key_machinery":"The load-bearing object is the Gram graph of an orbit: the edge-labeled directed graph whose vertices are the orbit's points and whose edge $s \\to t$ carries label $\\langle s, t \\rangle$. For a generic complex orbit this graph is isomorphic to the complete Cayley graph of $G$ (edge $h \\to k$ labeled by $h^{-1}k$), which follows from the fact that a generic $x$ makes the level sets of $g \\mapsto \\langle x, gx \\rangle$ coincide with those of $g \\mapsto g$ (or $g + g^{-1}$ in the real case). Since every label-preserving automorphism of a complete Cayley graph is left multiplication by a group element, $\\operatorname{Aut}(Gv) \\cong G$ follows. In real spaces the same graph is too coarse, so the paper pairs two generic orbits and uses the orbit pairing lemma: the map sending each point of the first orbit to its nearest point in the second is a well-defined bijection that is equivariant with respect to every symmetry of the union, transferring the action between orbits and yielding $\\operatorname{Aut}(Gv \\cup Gw) \\cong G$. For concrete recovery the mechanism is representation-theoretic: orbits are viewed as images of equivariant linear maps from $k$ copies of the regular representation, and Schur's lemma translates 'the spans of the orbits have codimension smaller than $r$' into the multiplicity bound $k \\ge \\max_\\pi (n_\\pi(V)-(r-1)[\\pi=1])/n_\\pi(R)$.","core_discovery":"The paper establishes that, for a finite group $G$ acting by unitary automorphisms on a finite-dimensional complex Hilbert space $V$, a single generic orbit determines $G$ up to isomorphism: the canonical map $G \\to \\operatorname{Aut}(Gv)$ is an isomorphism (Theorem 2.6). On a real Hilbert space, two generic orbits suffice, with $G \\to \\operatorname{Aut}(Gv \\cup Gw)$ an isomorphism (Theorem 2.12). For concrete recovery, Corollary 3.5(b) states that if $k \\ge [\\mathbb{C}:F]$ and $k$ meets the multiplicity bound $k \\ge \\max_\\pi (n_\\pi(V)-(r-1)[\\pi=1])/n_\\pi(R)$, where $r$ is the smallest dimension of a nontrivial representation, then $k$ generic orbits determine $G$ as a subset of $\\operatorname{Aut}(V)$; conversely, if $k$ fails that bound, every collection of $k$ orbits can be realized by another subgroup of $\\operatorname{Aut}(V)$.","pith_inferences":["The Gram graph plus automorphism computation gives a concrete algorithmic route to symmetry discovery from unlabeled point clouds: collect one generic complex orbit, build inner-product labels, and compute the label-preserving automorphism group; the two-orbit theorem supplies the synchronization rule that makes the same pipeline work for real data.","The orbit pairing lemma's nearest-point bijection is a natural target for numerical experiments: it suggests that approximate orbits can be matched by nearest neighbours before estimating the group, and its equivariance could be checked statistically.","If the real one-orbit conjecture is true, the distinction between complex and real cases in abstract recovery would vanish entirely, and the second orbit in Theorem 2.12 would be an artifact of the proof rather than an information-theoretic necessity.","The concrete-recovery bound is a worst-case generic threshold, but Example 3.8 shows genericity itself can carry extra information (such as nonzero centroid), so the number of orbits actually needed in structured families may be far smaller."],"forward_implications":["In the complex case, an observer who receives one generic orbit can compute the Gram graph and read off the isomorphism class of $G$ without prior knowledge of the dimension or order.","In the real case, two generic orbits recover the abstract group as $\\operatorname{Aut}(Gv \\cup Gw)$; one orbit already suffices when $G$ has prime order or $V$ has dimension two.","The concrete recovery threshold is sharp: if $k$ fails the multiplicity bound, any $k$ orbits can be reinterpreted as orbits of a different subgroup of $\\operatorname{Aut}(V)$, while if $k$ meets both the bound and $k \\ge [\\mathbb{C}:F]$, $k$ generic orbits determine $G$ exactly.","For $V$ equal to the regular representation over $\\mathbb{C}$, a single generic orbit determines the concrete group action; for $G = \\{\\pm I\\}$, genericity lets a single orbit determine the group even though the general bound asks for $d$ orbits."],"supporting_citations":[{"why":"Supplies the generic-identifiability method: a nonzero polynomial's complement is open and dense, which underlies every 'for generic x' statement in the proofs.","marker":"[2]"},{"why":"Introduces the symmetry group of a finite frame, the object whose generic complex version the paper proves is isomorphic to G and uses for orbit automorphisms.","marker":"[50]"},{"why":"Provides the graph- and representation-based toolkit for describing highly symmetric frames that shapes the orbit-pairing and regular-representation arguments.","marker":"[54]"},{"why":"Gives the classification of finite subgroups of O(2) used in the motivating examples and the two-dimensional real case.","marker":"[56]"}],"fun_headline_variants":["One generic orbit reveals a complex finite group","Two generic orbits recover real finite groups exactly","Sharp orbit count: one for complex, two for real","Minimal generic orbits determine finite groups","Few orbits identify finite groups on Hilbert spaces"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The two-orbit theorem assumes that the two real orbits are permuted in lockstep—that every symmetry of their union also gives a symmetry of the complexified orbit $G(v+iw)$—an assumption supplied by the later orbit pairing lemma but not cited there.","fun_headline_variants_meta":{"raw":{"variants":["One generic orbit reveals a complex finite group","Two generic orbits recover real finite groups exactly","Sharp orbit count: one for complex, two for real","Minimal generic orbits determine finite groups","Few orbits identify finite groups on Hilbert spaces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000299,"raw_usage":{"total_tokens":1656,"prompt_tokens":802,"completion_tokens":854,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":418,"completion_tokens_details":{"reasoning_tokens":786}},"tokens_in":418,"tokens_out":854,"duration_ms":7992,"temperature":1.0,"reasoning_tokens":786,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:10:33.577872+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Sample a random pair of orbits of the quaternion group $Q_8$ acting by left multiplication on $\\mathbb{R}^4$ and count the automorphisms of their union; Theorem 2.12 predicts exactly 8 for every pair in its generic set, so finding a pair with more symmetries, particularly one that preserves distances but is not a single complexified isometry, would disprove the two-orbit theorem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the symmetry group of a finite frame, the object whose generic complex version the paper proves is isomorphic to G and uses for orbit automorphisms."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the graph- and representation-based toolkit for describing highly symmetric frames that shapes the orbit-pairing and regular-representation arguments."},{"cited_title":"Weyl, Symmetry, Princeton U","cited_arxiv_id":null,"evidence_quote":"Gives the classification of finite subgroups of O(2) used in the motivating examples and the two-dimensional real case."}],"review_version":1}