{"id":"78e276ab-d6d1-4fcb-a305-42aafe082cb0","arxiv_id":"2509.22821","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Under non-collapsed equivariant Gromov-Hausdorff convergence with Ricci bounds, dim(G) ≥ limsup dim(G_i), with applications to RCD isometry groups.","lead":"This pure-math paper shows that in non-collapsing families of curved spaces with Ricci bounds, the dimension of the symmetry group cannot drop in the limit: the limit space has at least as many dimensions of symmetry as the approximating spaces. It settles an open question by Harvey and extends symmetry-dimension bounds to RCD spaces.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem A hinges on Theorem 7.1, an unproved passage from finite BGT approximate groups to open precompact subsets with Haar measure; unless this extension is supplied, Lemmas 8.6–8.7 and Theorem 7.3 lack support.","rationale":"The reader and I converge on the same load-bearing assumption: Theorem 7.1. I do not see a fatal flaw in the rest of the architecture; given Theorem 7.1, the Borsuk-Ulam contradiction in Theorem 1.10 is coherent, and the reduction from eGH convergence to good approximations (Theorem C) is carefully done. The issue is not that the BGT extension is false—it is likely true—but that it is a non-obvious deep transfer from finite combinatorics to non-discrete locally compact groups, and the manuscript's sketch does not exhibit the required uniform covering constants or the Haar-measure argument. This is exactly the sort of gap that makes a conditional verdict appropriate: the central claim is plausible and probably correct, but the main technical engine is not fully demonstrated. I also note secondary presentation problems (the abstract attribution mismatch mentioned by the reader, the very terse Section 9), but those are not the deciding issue.","tokens_in":22823,"tokens_out":18168,"duration_ms":191771,"concrete_test":"Write out a full proof of Theorem 7.1 from BGT: fix a nonprincipal ultrafilter, form the ultralimit (G̃, B̃, φ̃) of the good approximations, and verify (i) B̃ is an open precompact strong K-approximate subgroup with K independent of i, and (ii) the induced map to G satisfies the BGT good-model axioms (Definition 3.5). Then rerun the proof of [BGT12, Theorem 8.1] line by line with left-invariant Haar measure, checking every step that used counting measure and showing the resulting constants depend only on K. If either the finite-covering bound or the Haar-measure step fails, Theorem 7.1 (and hence Lemmas 8.6–8.7 and Theorem A) is unsupported; if it succeeds, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 7's Theorem 7.1 is the technical engine of the paper. It asserts a uniform escape-norm inequality for all g_1,...,g_m∈G_i with respect to the open precompact symmetric neighborhoods B_i, citing [BGT12, Theorem 8.1]. The proof given is a sketch: the sets B_i are claimed to be strong K-approximate groups for large i, the ultralimit of φ_i:B_i^8→G is said to be a good model with 'finite' replaced by 'open pre-compact', and the only comment on the change of measure is that one uses left-invariant Haar measure instead of counting measure. This is load-bearing: Lemma 8.6 uses Theorem 7.1 to prove subadditivity of |·|_i, Lemma 8.7 converts this to a genuine norm, Theorem 7.3 uses the same theorem to prove H_i is a subgroup and maximal, and Theorem 1.10's contradiction rests on the resulting escape-norm estimates. The finite proof of BGT Theorem 8.1 is not formal: it relies on counting/covering estimates for powers A,A^2,A^3; in the open-precompact case these become uniform finite-covering-number bounds, and it must be shown that conditions (I)-(V) imply such bounds with constants independent of i. Replacing counting measure by Haar measure does not automatically give those bounds (Haar measure of an open set can be infinite or meaningless without normalization), and the good-model ultralimit needs to be a genuine local group model for non-discrete B_i. Since the authors explicitly say condition (I) is used only here and give no proof of the extension, this is the weakest load-bearing point of the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that the dimension of the isometry group of a non-collapsed Ricci-limit space is upper-semicontinuous under equivariant Gromov–Hausdorff convergence. Concretely, Theorem A states that for closed subgroups G_i ≤ Iso(X_i) of pointed complete n-manifolds with Ric ≥ -(n-1) and vol(B_1(p_i)) ≥ v > 0, if (X_i,G_i,p_i) eGH-converges to (X,G,p), then dim(G) ≥ limsup_i dim(G_i). The same is claimed for RCD spaces of essential dimension n. The proof passes through a new axiomatization of eGH convergence in terms of 'good approximations' between the isometry groups (Theorem C), a reduction to good approximations with target R^k (Lemma 6.1), and an approximate Borsuk–Ulam argument (Theorem 4.21). The main technical engine is a claimed extension of the Breuillard–Green–Tao Gleason lemmas to open pre-compact symmetric subsets of locally compact groups (Theorem 7.1), which is used to construct an escape norm and maximal small subgroups. Applications to dimension bounds for isometry groups of RCD spaces are given in Section 9.","tokens_in":23278,"tokens_out":9126,"duration_ms":72447,"significance":"If the proof is completed, the paper solves a natural open question left by Harvey: the dimension inequality for isometry groups under non-collapsed convergence holds without the compactness or uniform boundedness assumptions on the orbits. The good-approximation framework is a clean and potentially reusable abstraction, and the applications to RCD spaces go beyond previous results by removing compactness, co-Lipschitz, and measure-preservation hypotheses. The paper is generally well organized and the overall strategy is attractive. However, the current version contains a load-bearing unproved extension of BGT's approximate-group Gleason lemmas; the dimension proof depends on that extension at several points. The significance is high, but the manuscript is not yet in publishable form.","major_comments":[{"comment":"This theorem is the technical heart of the paper, but its proof is only a sketch. It asserts that [BGT12, Theorem 8.1] extends verbatim from finite approximate groups to open pre-compact symmetric subsets of locally compact Hausdorff groups satisfying conditions (I)–(V). The extension is not a formality: the BGT proof uses counting measures and finite covering estimates for powers A,A^2,A^3, while the present setting requires uniform finite covering numbers or a normalization of Haar measure, none of which follows immediately from (I)–(V). The claimed ultralimit 'good model' with 'finite' replaced by 'open pre-compact' must be shown to satisfy the BGT axioms, and the descent of the escape-norm estimate to the original groups must be proved. This is load-bearing: Lemma 8.6, Lemma 8.7, and Theorem 7.3 all rely on Theorem 7.1, and Theorem 1.10 (hence Theorem A) depends on them. The authors","section":"Section 7, Theorem 7.1"},{"comment":"The proof of Lemma 8.6 invokes Lemma 4.22, whose hypothesis is that the open sets B_i satisfy h ∈ B_i^3 \\ B_i ⇒ h^2 ∉ B_i. This property is plausible for the small exponential balls B(r) in a Lie group, but the sets B_i are only given by Proposition 4.14 as preimages under φ_i of compact/open neighborhoods, and it is not shown that they inherit this property uniformly in i. Since Lemma 8.6 is the step that converts the escape norm into a subadditive quantity and hence into a genuine norm in Lemma 8.7, this missing verification is not merely cosmetic. It should either be proved directly or absorbed into a fully stated version of Theorem 7.1.","section":"Section 8, Lemma 8.6"},{"comment":"The equality/rigidity part of Theorem 2.3 is not adequately justified. The proof asserts that geodesics in X/G can be lifted to geodesics in X ('Via a selection argument...'), that a sequence of atomic measures can be chosen with W_2 convergence, and that the non-branching property of X/G forces it to be a one-dimensional manifold. These are nontrivial steps, especially because the group G is not assumed compact and the quotient X/G is not known to be an RCD space. The claim that X/G is non-branching is also quoted without proof or reference. This argument needs to be expanded substantially or replaced before Theorem 2.3 can be accepted.","section":"Section 9, Proof of Theorem 2.3"}],"minor_comments":[{"comment":"The abstract block supplied for the paper attributes the prior result to 'Mazur–Rong–Wang', while the introduction and full-text abstract attribute it to Harvey. Please align these attributions.","section":"Abstract / Introduction"},{"comment":"Typo: 'pais' should be 'pairs' in the last paragraph of the proof.","section":"Section 4.4, Proposition 4.14"},{"comment":"Typo: 'outisde' should be 'outside' in the paragraph after the escape norm definition.","section":"Section 4.6"},{"comment":"The statement says 'normal neighborhoods' but Definition 1.7 calls them 'regular neighborhoods'. Please use consistent terminology.","section":"Section 7, Theorem 7.3"},{"comment":"The notation Γ, \\tildeΓ, e_0, and the measures ν_n, η_n is introduced very quickly. In particular, the claim that W_2(ν_n,η_n) = W_2(p_#ν_n,p_#η_n) and that the transport is along geodesics projecting to quotient geodesics needs a precise definition and proof.","section":"Section 9, Proof of Theorem 2.3"}],"recommendation":"major_revision","confidential_remarks":"The main gap is localized to Theorem 7.1, and the paper's central theorem would be convincing if the authors supply a complete proof of the BGT extension or cite a published statement that covers exactly this setting. The Section 2 applications are also currently less rigorous than the main theorem; they should be revisited. I do not see evidence of a fatal flaw in the overall strategy, but the manuscript is not acceptable in its present form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis is a serious paper with a real result, but the main technical engine has a gap that needs closing before I'd trust Theorem A.\n\nWhat's genuinely new: Harvey's dimension inequality is extended to the non-compact, Ricci-bounded, non-collapsed setting, without bounded orbits, and to RCD spaces. The good approximations framework (Definition 1.7) is useful, and the reduction from eGH convergence to good approximations (Theorem C) is clean. The approximate Borsuk–Ulam argument in Section 8 is elegant, and Lemma 6.1 is a nice piece of work.\n\nThe soft spot: everything rests on Theorem 7.1, the claimed extension of BGT's Gleason lemmas to open precompact symmetric sets. The proof is a sketch: take ultralimits, replace counting measure with Haar measure, and assert the BGT machinery works verbatim. That is not a routine change. The BGT proof uses counting/covering estimates on powers A, A^2, A^3; with Haar measure you need uniform finite covering bounds with constants independent of i, and the good-model ultralimit needs to be a genuine local group. The authors themselves note condition (I) is only used here, which tells you how isolated this step is. Lemma 8.6 and the normality/maximality of H_i in Theorem 7.3 depend on it. If this gap can be filled, the proof goes through. If not, the main theorem is unsupported.\n\nMinor issues: the abstract says 'generalizing a result of Mazur–Rong–Wang' but the body says Harvey. Check that. The RCD applications in Section 9 are terse; the measure-transport argument in Theorem 2.3 needs more detail.\n\nI agree with the reviewer's conditional verdict. The paper is worth a serious referee: the central claim is important, the framework is fresh, and the gap is specific and likely fixable. I'd recommend sending it to review with a targeted request to the referee to check Section 7 carefully.","headline":"A genuinely new result with an elegant framework, but the proof of the load-bearing BGT extension in Theorem 7.1 is only sketched; that gap needs to be closed before the main theorem is fully supported.","tokens_in":23722,"tokens_out":2208,"would_cite":true,"duration_ms":17271,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C23","53C24","22E15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that under non-collapsed equivariant Gromov–Hausdorff convergence, the limit isometry group's dimension is at least the limit superior of the approximating isometry groups' dimensions.","keywords":["equivariant Gromov–Hausdorff convergence","dimension of isometry groups","upper semicontinuity","Ricci curvature lower bound","RCD spaces","good approximations","escape norm","approximate Borsuk–Ulam"],"falsifier":"Exhibit a sequence of Lie groups G_i with the no-small-subgroup property and good approximations to R^k (satisfying conditions I–V) with dim(G_i) > k for infinitely many i; that would refute Theorem 1.10 and hence Theorem A. A concrete starting point is to check whether the escape-norm subadditivity inequality (Lemma 8.6) holds for G_i = R² with A_i = B_{1/i}(0) and a natural projection to R; failure there would show the Gleason-extension step is false.","tokens_in":22733,"feed_emoji":"📐","tokens_out":10252,"duration_ms":66998,"temperature":0.7,"pith_summary":"This paper proves a semicontinuity result for symmetry: if a sequence of manifolds with Ricci curvature bounded below and volume non-collapsing converges, equivariantly, to a limit space, then the dimension of the limit's isometry group is at least the limiting dimension of the sequence's isometry groups. This answers a question left open by earlier work that required uniformly bounded orbits. The same statement holds in the synthetic setting of RCD spaces of a fixed essential dimension. The proof reduces the geometric convergence to a group-theoretic statement about 'good approximations' between locally compact groups, then uses an approximate antipodal-map (Borsuk–Ulam) argument. If correct, this makes symmetry dimension an upper semicontinuous invariant of non-collapsed convergence, with rigidity consequences for RCD spaces with large isometry groups.","feed_headline":"Limit isometry group dimension exceeds approximating ones","feed_subtitle":"Under Ricci lower bounds and volume non-collapsing, symmetries of the limit control the sequence; extends to RCD spaces.","key_machinery":"The load-bearing object is the 'good approximation': a sequence of maps φ_i from the isometry groups G_i to the limit group G, together with open pre-compact 'regular neighborhoods' A_i ⊂ G_i and A ⊂ G satisfying five conditions (I–V) that force φ_i to be an almost-homomorphism on bounded sets. The proof shows eGH convergence yields such maps. The dimension argument then hinges on the escape norm ∥·∥_A on each G_i (how many powers of an element stay inside A), which is turned into a genuine norm on the Lie algebra of G_i via a convex-hull construction. The key inequality that makes the norm work is the approximate-subadditivity of the escape norm, which the paper obtains by extending Gleason","core_discovery":"The central claim (Theorem A) is that for any non-collapsing sequence of complete n-manifolds with Ricci ≥ -(n-1) and volume of unit balls ≥ v > 0, any eGH-convergent sequence of closed isometry groups G_i yields dim(G) ≥ limsup dim(G_i), with no boundedness assumption on orbits. The same holds for RCD spaces of essential dimension n. The proof passes through a purely group-theoretic theorem (Theorem 1.10): any sequence of Lie groups G_i with the no-small-subgroup property that admits 'good approximations' to a Lie group G must have dim(G_i) ≤ dim(G) for large i. Good approximations are maps from G_i to G that are almost homomorphisms on precompact neighborhoods, satisfying five regularity c","pith_inferences":["The group-theoretic core (Theorem 1.10) is likely transferable to any convergence theory that yields good approximations and the no-small-subgroup property, beyond the Ricci/RCD settings considered here.","The equality case in dimension bounds for RCD isometry groups suggests a rigidity classification: spaces achieving the bound should be infinitesimally homogeneous, which may lead to a local-to-global rigidity statement not written out in the paper.","The proof's only unproven ingredient is the claimed extension of the finite-approximate-group Gleason lemmas to open pre-compact sets with Haar measure; if that extension fails, the escape-norm subadditivity (Lemma 8.6) would need a different proof, though the theorem might still hold.","The maximal small subgroup H_i can be viewed as the 'ineffective kernel' of the action at scale zero; its existence suggests that non-collapsed eGH limits forget exactly the infinitesimal symmetries, which could be formalized as a categorical adjunction between convergence and Lie group quotient."],"forward_implications":["Symmetry degree is upper semicontinuous under pointed Gromov–Hausdorff convergence of non-collapsed Riemannian manifolds with Ricci ≥ -(n-1) (Corollary 1.3).","For RCD spaces of essential dimension n, the isometry group of a closed subgroup satisfies dim(G) ≤ n(n+1)/2, with equality only for the model spaces Sⁿ, RPⁿ, Rⁿ, Hⁿ, without compactness or action-regularity hypotheses (Theorem 2.2).","A non-transitive closed isometry group on an RCD space of essential dimension n has dim(G) ≤ n(n-1)/2, with equality only when the quotient is a real interval or a circle (Theorem 2.3).","For a compact G on an RCD space, dim(G) ≤ (n-m)(n-m+1)/2, where m is the essential dimension of the quotient (Theorem 2.5).","Given good approximations to a Lie group G, the approximating groups contain a maximal small subgroup whose quotient is a Lie group of dimension ≤ dim(G), provided the groups are generated by elements of bounded displacement (Theorem B)."],"fun_headline_variants":["Limit isometry group size can't shrink in Ricci-noncollapsed limits","Symmetry dimension monotone in equivariant GH limits","Noncollapsed Ricci limits preserve or grow isometry dimension","Isometry group dimension: limit dominates sequence","RCD limit isometry dimension at least limsup"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof relies on the unproven claim that the Gleason lemmas, originally proved for finite approximate groups, extend verbatim to open pre-compact symmetric subsets of locally compact groups after replacing counting measure with left-invariant Haar measure; this extension is the only use of condition (I).","fun_headline_variants_meta":{"raw":{"variants":["Limit isometry group size can't shrink in Ricci-noncollapsed limits","Symmetry dimension monotone in equivariant GH limits","Noncollapsed Ricci limits preserve or grow isometry dimension","Isometry group dimension: limit dominates sequence","RCD limit isometry dimension at least limsup"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000686,"raw_usage":{"total_tokens":2939,"prompt_tokens":730,"completion_tokens":2209,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":474,"completion_tokens_details":{"reasoning_tokens":2129}},"tokens_in":474,"tokens_out":2209,"duration_ms":16074,"temperature":1.0,"reasoning_tokens":2129,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T14:47:12.817336+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a sequence of Lie groups G_i with the no-small-subgroup property and good approximations to R^k (satisfying conditions I–V) with dim(G_i) > k for infinitely many i; that would refute Theorem 1.10 and hence Theorem A. A concrete starting point is to check whether the escape-norm subadditivity inequality (Lemma 8.6) holds for G_i = R² with A_i = B_{1/i}(0) and a natural projection to R; failure there would show the Gleason-extension step is false.","supporting_citations":[],"review_version":1}