{"id":"39113696-1bcc-4145-ad6a-925b7aa9820a","arxiv_id":"2506.04425","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper derives exact Euclidean distortion values for several orbit spaces, including cyclic quotients of C^n, seven wallpaper group quotients, and two-sided bounds for O(r), SO(r), E(r), and SE(r) actions.","lead":"This mathematics paper computes how much the geometry of orbit spaces, sets of points identified under rotations, permutations, or reflections, gets stretched when embedded into ordinary Euclidean space. It introduces new general tools and uses them to pin down exact distortion values for spaces relevant to invariant machine learning, from cyclic group actions to wallpaper patterns.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 16's embedding F is not well-defined on C^n/C_r as printed: φ(u)=u⊗u/∥u∥ is not C_r-invariant for r>2; the proof's own distance formula shows the intended map is u⊗\\bar u.","rationale":"The reader's weakest-assumption analysis focuses on the finite-determinacy principle and the geometry of Theorem 13, but a more immediate and more load-bearing defect sits in the statement and proof of the paper's central theorem. As printed, the map F in Theorem 16 is not well-defined on C^n/C_r: the first coordinate u⊗u/∥u∥ changes by ω² when u is replaced by ωu. The proof's own computation of ∥φ(u)−φ(v)∥² as 2−2|z|² cannot be correct for the printed φ, since for a complex bilinear tensor one would obtain 2−2Re(z²); it is correct for the phase-invariant map u⊗\\bar u/∥u∥. This is not a subtle inequality gap but a contradiction in the main construction, and it propagates to Lemma 8 and to the exact wallpaper distortion values that invoke Theorem 16. I do not regard this as a rejection of the mathematical program: the evident intended embedding u⊗\\bar u is phase-invariant and the displayed estimates appear designed for it. But the submitted theorem and its supporting notation must be corrected and rechecked, and the consequences for Lemma 8 and Theorem 26 should be re-verified. Hence the appropriate verdict is CONDITIONAL rather than unconditional ACCEPT.","tokens_in":43550,"tokens_out":20205,"duration_ms":202212,"concrete_test":"Take r=3, n=1, u=1, v=e^{2πi/3}; verify [u]=[v] while the printed F gives nonzero distance, confirming non-well-definedness. Then replace φ with u⊗\\bar u/∥u∥ and rerun the proof of Theorem 16: check that the displayed identities ∥φ(u)−φ(v)∥²=2−2|z|² and ∥ψ(u)−ψ(v)∥²=2−2Re z^r hold, and that the subsequent inequalities prove κ(F)=r sin(π/2r). If the corrected map satisfies the stated two-sided bound, Theorem 16 is restored after propagating u⊗\\bar u through the paper's notation.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Theorem 16, F([u]) is defined as (cos(π/2r)φ(u), sin(π/2r)ψ(u)) with φ(u)=u⊗u/∥u∥ and ψ(u)=u^{⊗r}/∥u∥^{r-1}. Since φ(ωu)=ω²φ(u) for ω=e^{2πi/r}, the first coordinate is not invariant under the C_r action when r>2, so F is not a function on the quotient C^n/C_r. Concretely, for r=3, n=1, u=1, v=ω, we have [u]=[v] but ∥F([u])−F([v])∥=cos(π/6)|1−ω²|>0, contradicting well-definedness. The proof's later identity ∥φ(u)−φ(v)∥²=2−2|z|² with z=u*v is inconsistent with φ=u⊗u (which would give 2−2Re z²) and instead matches φ(u)=u⊗\\bar u (equivalently uu*/∥u∥). Thus the headline theorem as stated is internally inconsistent. This is load-bearing because Theorem 16 is used for Lemma 8's Υ(C₃)=3/2 and for the exact wallpaper values in Theorem 26(c),(d). The surrounding argument appears repairable by replacing every u⊗u with u⊗\\bar u and adjusting the codomain accordingly, but the submitted text requires this correction before the central claim is valid.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops tools for estimating the Euclidean distortion c2(V/G) of orbit spaces under isometric group actions, introducing equivariant embedding lemmas, a new notion of Euclidean contortion of a finite group, quotient–orbit embeddings, and a local-to-global lower bound via tangent-space isotropy quotients. These tools are then applied to compute or bound c2 exactly for several families: scalar cyclic actions on C^n, orthogonal/special-orthogonal matrix actions, alternating subgroups of reflection groups, wallpaper groups, landmark quotients by Euclidean groups, and permutation actions on graphs and databases. The headline result is Theorem 16, claiming c2(C^n/C_r)=r sin(pi/(2r)), achieved by an explicit embedding mixing u⊗u/|u| and u^{⊗r}/|u|^{r-1}; Theorem 26 claims exact values for seven wallpaper group quotients. The paper is largely self-contained, with appendices supplying proofs of auxiliary results such as Proposition 21 and background on Bochner spaces.","tokens_in":43829,"tokens_out":5044,"duration_ms":51989,"significance":"If the main results are correct, the paper resolves the scalar cyclic case in all dimensions, supplies exact Euclidean distortions for several wallpaper-group quotients, introduces the transferable invariant Υ(G) with exact values for |G|≤3, and gives a new general local-to-global mechanism for lower bounds. The explicit nature of the embeddings and the inclusion of complete proofs for several delicate steps (e.g., Propositions 21, 34, 35, and the slice-theorem argument in Appendix B) are notable strengths. However, the central Theorem 16 as printed is not well-defined on the quotient, and since it feeds Lemma 8 and Theorem 26(c),(d), the significance is conditional on the straightforward repair described in the major comments.","major_comments":[{"comment":"The map F([u]) = (cos(π/2r)φ(u), sin(π/2r)ψ(u)) is not well-defined on C^n/C_r as printed, because φ(u)=u⊗u/‖u‖ is not C_r-invariant for r>2: for ω=e^{2πi/r}, φ(ωu)=ω²φ(u). Concretely, for r=3, n=1, u=1, v=ω, one has [u]=[v] but ‖F([u])−F([v])‖=cos(π/6)|1−ω²|>0, contradicting well-definedness. The proof itself uses the identity ‖φ(u)−φ(v)‖²=2−2|z|² with z=u^*v, which holds for φ(u)=u⊗`bar u` (equivalently the rank-one projection uu*/‖u‖), not for u⊗u. Since Theorem 16 is used to obtain Υ(C₃)=3/2 in Lemma 8 and the exact values in Theorem 26(c),(d), this is a load-bearing defect. The argument is repairable by replacing every occurrence of u⊗u with u⊗`bar u` in §3.1 (and correspondingly in §1.2.1 and §4.1) and adjusting the codomain to (C^n)⊗`overline{C^n}`, after which the displayed distance computations and bounds are consistent with the corrected map. The submitted text requires this correction before the central claim is valid.","section":"§3.1, Theorem 16"},{"comment":"The exact lower bounds for wallpaper groups of types 2*22 and 4*2 are derived from Lemma 25 (via Theorem 13) and the upper bounds are obtained by isometrically embedding R²/G into R²/±Id or R²/C₄, invoking Theorem 16 for the distortion of the latter quotients. Since Theorem 16 as stated is invalid without the correction to φ described above, the exact values √2 and 2√(2−√2) in Theorem 26(c) and (d) currently rest on an ill-defined embedding. After the repair, the same proof goes through verbatim, so this is a consequence of the first major comment rather than an independent defect.","section":"§3.4, Theorem 26(c),(d)"}],"minor_comments":[{"comment":"The text repeatedly writes the optimal U(1)-invariant map as z ↦ z⊗z/‖z‖; this should be z⊗`bar z`/‖z‖, in line with the correction required in Theorem 16, otherwise the notation is actively misleading.","section":"§1.2.1 and §4.1"},{"comment":"The proof cites 'Theorem 13 in [14]' to reduce to unit vectors; it would be helpful to state explicitly that this is a homogeneity argument, since the homogeneity of F is immediate but the cited theorem is not standard in the distortion literature.","section":"§3.1, proof of Theorem 16"},{"comment":"In the proof of (ii), the functions f₁ and f₂ are used in the text before being defined; they are introduced only later via the equation g' = f₁−f₂. Please define them at first use.","section":"§3.1, Figure 3 and surrounding text"},{"comment":"There are small typos that should be cleaned up: 'arbirary' in the proof of Corollary 12, 'mulivariate Bernoulli' in the proof of Lemma 19, and 'the the' in Section 1.2.1.","section":"§2.3.2 and §3.2"}],"recommendation":"major_revision","confidential_remarks":"The tensor-square issue is a genuine but local error: the proof of Theorem 16 already works with the corrected map u⊗`bar u`, so the fix is a matter of consistency rather than a new argument. I recommend asking the authors to audit every occurrence of `u⊗u` for invariance under the relevant group action, and to add a one-line verification of C_r-invariance for the corrected φ. I do not see other load-bearing problems; the finite-determinacy principle (Proposition 34) is nonconstructive but standard and is honestly flagged in Appendix A. The paper's self-containedness and explicit embeddings are strengths."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper builds a genuinely useful toolbox for bounding Euclidean distortion of orbit spaces: the contortion invariant, quotient–orbit embeddings, the local-to-global lower bound, and the translation-subspace reduction are all new and clearly explained. The applications are real: exact values for cyclic quotients, O(r)/U(r), several wallpaper groups, and unbounded distortion families. The writing is careful and the self-citations are upfront; Proposition 34's use of choice is flagged honestly.\n\nThe soft spots start with Theorem 16. The stress-test note is correct: the embedding F defined there is not invariant under C_r for r>2, because φ(u)=u⊗u/\\|u\\| picks up ω² under scaling by ω. The proof's own distance formula, with |u*v|², corresponds to u⊗\\bar u, not u⊗u. So the headline theorem as printed is internally inconsistent. This matters because Theorem 16 feeds Lemma 8 and the wallpaper bounds in Theorem 26. The good news is the intended map is obvious and the fix is local: replace u⊗u with u⊗\\bar u (or ⟨·,·⟩-type tensor) and adjust the codomain. The rest of the argument then appears sound; my spot-checks of the mean-value and Wirtinger steps found nothing else broken.\n\nOther concerns are minor. Proposition 17 leans on [7] for its key inequality, and the wallpaper geodesic-circle claims are compressed, but these are acceptable. The finite-determinacy principle (ultraproduct) is a real nonconstructivity, but the paper acknowledges it.\n\nBottom line: a serious paper with a load-bearing typo. The central machinery is good enough that a referee should see it, but the version I read should not be accepted as-is. I'd ask the authors to fix Theorem 16 and propagate the correction through Lemma 8 and Theorem 26, then re-review. If my colleague is deciding whether to engage: yes, engage, but read the corrected version.","headline":"Good tools and genuine applications, but Theorem 16 as stated is not well-defined; the fix is clear and the paper deserves review after correction.","tokens_in":44483,"tokens_out":1163,"would_cite":false,"duration_ms":15522,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46B85","57S15","54C25"],"pacs":[],"model":"deepseek-v4-flash","headline":"The quotient of $\\mathbb{C}^n$ by the $r$-th roots of unity embeds into Euclidean space with distortion exactly $r\\sin(\\pi/(2r))$.","keywords":["Euclidean distortion","orbit space","quotient metric","bilipschitz embedding","cyclic group action","wallpaper groups","invariant machine learning","contortion"],"falsifier":"Run the semidefinite-programming lower bound for a finite subset of $\\mathbb{C}^2/\\mathcal{C}_3$ that includes points with several relative phases on the unit sphere; if the computed distortion exceeds $3/2$, Theorem 16's exact value is false.","tokens_in":43292,"feed_emoji":"📐","tokens_out":8720,"duration_ms":78445,"temperature":0.7,"pith_summary":"Metric spaces that arise as quotients of Euclidean space by a group of symmetries, called orbit spaces, appear whenever data carries a symmetry ambiguity, and machine-learning pipelines often need to map such spaces into Euclidean space while distorting distances as little as possible. This paper develops a toolbox for computing the optimal distortion, and uses it to settle the exact value for several fundamental families. The headline result computes the Euclidean distortion of $\\mathbb{C}^n/\\mathcal{C}_r$, the quotient of complex $n$-space by the $r$-th roots of unity, showing it equals $r\\sin(\\pi/(2r))$; the same machinery gives exact distortions for seven wallpaper-group quotients of the plane and two-sided bounds for others. A recurring message is that low-distortion embeddings can be built from the classical polynomial invariants of the group action, suitably rescaled.","feed_headline":"Cyclic orbit spaces embed with distortion exactly r sin(pi/2r)","feed_subtitle":"The key case C^n/C_r is resolved in every dimension; wallpaper groups yield more exact values.","key_machinery":"The central mechanism is the quotient–orbit embedding (Theorem 11): if a map $\\phi$ embeds the quotient $X/G$, and a $G$-equivariant map $\\psi$ is orbit-expanding and alignment-preserving, then $x \\mapsto (\\phi([x]), c\\,\\alpha_\\phi\\, \\psi(x))$ is a bilipschitz embedding of $X$ with distortion controlled by $\\kappa(\\phi)$ and the Lipschitz constant of $\\psi$. For Theorem 16 the two ingredients are the degree-two tensor map $\\phi(u)=u\\otimes u/\\|u\\|$, which is optimal for the full circle action on $\\mathbb{C}^n$, and the degree-$r$ tensor map $\\psi(u)=u^{\\otimes r}/\\|u\\|^{r-1}$, which separates the $\\mathcal{C}_r$-orbits inside each circle-orbit; the scaled mixture $F=[\\cos(\\pi/(2r))\\,\\phi,\\; \\sin(\\pi/(2r))\\, \\psi]$ gives the exact distortion $r\\sin(\\pi/(2r))$. Supporting the general theory are the finite-determinacy principle ($c_2$ equals the supremum over finite subspaces), the contortion $\\Upsilon(G)$ (the largest distortion any quotient by $G$ can have), and the local-to-global inequality $c_2(T_pM/G_p) \\le c_2(M/G)$ for wandering isometric actions.","core_discovery":"On the paper's own terms, the central discovery is that the Euclidean distortion of an orbit space can often be computed exactly by combining the optimal embedding of a larger quotient with an invariant that separates the remaining orbits. Theorem 16 states that for the action of the cyclic group $\\mathcal{C}_r$ by scalar multiplication on $\\mathbb{C}^n$, the distortion satisfies $c_2(\\mathbb{C}^n/\\mathcal{C}_r) = r\\sin(\\pi/(2r))$, and the map $F([u]) = (\\cos(\\pi/(2r))\\, u\\otimes u/\\|u\\|,\\; \\sin(\\pi/(2r))\\, u^{\\otimes r}/\\|u\\|^{r-1})$ achieves this distortion. The proof reduces to the unit sphere and to verifying two trigonometric inequalities involving $|u^*v|$ and $\\operatorname{Re}((u^*v)^r)$. The same local-to-global and quotient-orbit tools yield exact distortions for seven wallpaper-group types, for instance $\\sqrt{2}$ for type $2{*}22$ and $2\\sqrt{2-\\sqrt{2}}$ for type $4{*}2$, and imply that permutation-symmetric quotients of graphs and databases have unbounded distortion as the size grows.","pith_inferences":["A natural generalization suggested by the quotient–orbit recipe is that for any subgroup $H$ of a compact group $G$, the optimal embedding of $V/H$ might be a scaled concatenation of the optimal $V/G$ embedding with a suitably normalized $H$-orbit-separating invariant; this could be tested computationally for small finite groups before seeking a proof.","The exact wallpaper values came from lower bounds at rotation centers and upper bounds by gluing; the same two ingredients could yield exact values for the remaining wallpaper types once the flat-torus distortion of their translation lattices is known.","The unboundedness results imply that any invariant feature map for graphs or point clouds that aims for bounded metric distortion must have target dimension growing with $n$, a concrete constraint for geometric deep learning architectures.","Since the proof of the finite-determinacy principle uses an ultraproduct, a constructive proof for the specific spaces treated here would be a worthwhile test of whether the exact values depend on the axiom of choice."],"forward_implications":["The exact value $c_2(\\mathbb{C}^n/\\mathcal{C}_r)=r\\sin(\\pi/(2r))$ supplies the scalar-cyclic case in all dimensions, including $c_2(\\mathbb{C}/\\mathcal{C}_3)=3/2$ used in the contortion computation for groups of order three.","Seven wallpaper-group types now have exact distortions: reflection-wall groups of types $*333$, $*442$, $*632$, and $*2222$ embed isometrically with distortion $1$, while types $2{*}22$ and $4{*}2$ embed optimally with $\\sqrt{2}$ and $2\\sqrt{2-\\sqrt{2}}$, respectively.","For $\\mathrm{SO}(r)$ acting on $\\mathbb{R}^{r\\times n}$, the distortion lies between $\\sqrt{2}$ and $2\\sqrt{2}$ whenever $n \\ge r \\ge 2$.","For landmark spaces $(\\mathbb{R}^r)^n$ modulo rotations and translations, embedding reduces to the centered configuration space: $c_2((\\mathbb{R}^r)^n/(K \\ltimes \\mathbb{R}^r)) = c_2((\\mathbb{R}^r)^{n-1}/K)$.","Quotients of weighted graphs and databases by row and column permutations have Euclidean distortion tending to infinity as the number of vertices or columns grows."],"supporting_citations":[{"why":"Supplies the finite-determinacy principle, the product-space distortion formula, the one-dimensional cyclic distortion, and the sphere reduction that Theorem 16 relies on.","marker":"[14]"},{"why":"Gives the Gram-matrix bilipschitz bounds for $\\mathbb{R}^{r\\times n}/\\mathrm{O}(r)$ that Proposition 17 adapts and proves tight.","marker":"[7]"},{"why":"Provides the closed-form maximum of $\\operatorname{Tr}(QM)$ over $Q\\in\\mathrm{SO}(r)$ used in the special-orthogonal quotient analysis.","marker":"[37]"},{"why":"Establishes distortion $1$ for reflection-group quotients, which the alternating-subgroup and wallpaper results build on.","marker":"[38]"},{"why":"Computes the Euclidean distortion of the circle and provides flat-torus bounds used for wallpaper groups of translational type.","marker":"[27]"},{"why":"Determines exact distortions of the hexagonal and $E_8$ lattices, used for the remaining translation-lattice wallpaper cases.","marker":"[43]"},{"why":"Gives the transportation-metric impossibility result whose proof idea underlies the unbounded-distortion results for permutation actions.","marker":"[5]"}],"fun_headline_variants":["Cyclic orbit spaces: distortion exactly r sin(pi/2r)","Exact Euclidean distortion for cyclic orbit spaces","Orbit space distortion: r sin(pi/2r) exactly for C^n/C_r","Exact distortion for cyclic orbit spaces and wallpaper groups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The general machinery relies on the finite-determinacy principle that a quotient's distortion equals the supremum of distortions of its finite subspaces, a nonconstructive ultraproduct assertion that the paper itself flags as using the axiom of choice.","fun_headline_variants_meta":{"raw":{"variants":["Cyclic orbit spaces: distortion exactly r sin(pi/2r)","Exact Euclidean distortion for cyclic orbit spaces","Orbit space distortion: r sin(pi/2r) exactly for C^n/C_r","Exact distortion for cyclic orbit spaces and wallpaper groups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000681,"raw_usage":{"total_tokens":3046,"prompt_tokens":849,"completion_tokens":2197,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":465,"completion_tokens_details":{"reasoning_tokens":2123}},"tokens_in":465,"tokens_out":2197,"duration_ms":14556,"temperature":1.0,"reasoning_tokens":2123,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T10:44:59.824866+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the semidefinite-programming lower bound for a finite subset of $\\mathbb{C}^2/\\mathcal{C}_3$ that includes points with several relative phases on the unit sphere; if the computed distortion exceeds $3/2$, Theorem 16's exact value is false.","supporting_citations":[{"cited_title":"Cahill, J","cited_arxiv_id":null,"evidence_quote":"Supplies the finite-determinacy principle, the product-space distortion formula, the one-dimensional cyclic distortion, and the sphere reduction that Theorem 16 relies on."},{"cited_title":"Balan, C","cited_arxiv_id":null,"evidence_quote":"Gives the Gram-matrix bilipschitz bounds for $\\mathbb{R}^{r\\times n}/\\mathrm{O}(r)$ that Proposition 17 adapts and proves tight."},{"cited_title":"Miranda, R","cited_arxiv_id":null,"evidence_quote":"Provides the closed-form maximum of $\\operatorname{Tr}(QM)$ over $Q\\in\\mathrm{SO}(r)$ used in the special-orthogonal quotient analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes distortion $1$ for reflection-group quotients, which the alternating-subgroup and wallpaper results build on."},{"cited_title":"Heimendahl, M","cited_arxiv_id":null,"evidence_quote":"Computes the Euclidean distortion of the circle and provides flat-torus bounds used for wallpaper groups of translational type."},{"cited_title":"Vallentin, P","cited_arxiv_id":null,"evidence_quote":"Determines exact distortions of the hexagonal and $E_8$ lattices, used for the remaining translation-lattice wallpaper cases."},{"cited_title":"Andoni, A","cited_arxiv_id":null,"evidence_quote":"Gives the transportation-metric impossibility result whose proof idea underlies the unbounded-distortion results for permutation actions."}],"review_version":1}