{"id":"649d09e7-d957-43c5-bc85-f8226507f42f","arxiv_id":"2608.07649","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A transitive permutation group whose point stabilizer has relative Property (T) yields a Bauer or Poulsen simplex of invariant measures, with the Bauer case characterized by Property (T) of the group.","lead":"For many natural infinite permutation groups, the collection of all invariant probability measures is proved to be one of only two possible shapes: a well-behaved Bauer simplex or a maximally large Poulsen simplex. The result extends a classical theorem of Glasner and Weiss and settles open questions in exchangeability theory.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The dichotomy rests on the strong ε–δ form of relative Property (T) in Definition 4.1, whose equivalence with the standard weak form is open for Polish groups; the proof of Proposition 4.5 also omits convergence on mixed cylinder sets.","rationale":"The reader's weakest assumption identifies the same load-bearing issue: the strong ε–δ definition is essential to the proof and its equivalence to the standard weak relative Property (T) is explicitly open in Remark 4.3. I agree this makes the advertised scope of Theorem 13.11 conditional. I found no fatal internal contradiction: the model-theoretic framework, the face-preserving Robinson theory PMP_{G/H}, and the application of Theorem 10.6 are coherent, and the countable-density reduction in Proposition 13.7 is plausible because every term depends on finitely many coordinates. The only additional omission I noted is the missing mixed-cylinder convergence check in Proposition 4.5; it is a real gap in the written proof, but it is local and fixable with the same estimates, so it does not by itself change the verdict. Because the central claim is well supported but depends on a definitional choice the author flags as open, the appropriate verdict remains CONDITIONAL.","tokens_in":54834,"tokens_out":32219,"duration_ms":323014,"concrete_test":"Recompute the convergence claim in Proposition 4.5 for a mixed pattern, e.g., z=(1,0), verifying that lim_k μ_k(a_k ∩ g_1 a_k^c)=0 using μ_k(a_k△g_i a_k)→0 and μ_k(a_k)→λ. Then independently re-derive Lemma 4.4 from the weak relative Property (T) assumption; if the δ-closeness bound cannot be produced, search for a closed pair (G,H) in Sym(N) with weak but not strong relative Property (T), which would disprove Remark 4.2(i) and remove the examples.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central Bauer–Poulsen dichotomy is driven by Definition 4.1, a strong relative Property (T) notion: Lemma 4.4 and Lemma 4.6 require approximate invariance to be upgraded to closeness to an H-invariant vector, and Proposition 4.5 uses exactly that upgrade to separate the Bauer case from the Poulsen case. Remark 4.3 concedes that the equivalence with the usual weak definition is open for general Polish groups. The motivating examples are verified by citing ordinary Property (T) of Sym(N) via Tsankov, not the strong form, so unless the two forms coincide for closed permutation groups, Theorem 13.11 may not apply where advertised. A second, more local gap is that Proposition 4.5 asserts convergence of ν_k to ν after checking only constant cylinder sets U_{1^n} and U_{0^n}; these do not form a basis, so the mixed-cylinder limits must be supplied. The missing estimates are likely recoverable from μ_k(a_k△g_i a_k)→0, but they are not in the text.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a Bauer–Poulsen dichotomy for the Choquet simplex M_inv(K^S) of invariant measures for the induced action of a transitive permutation group G on a countable set S, under the hypothesis that the closure of the point stabilizer H has relative Property (T) in the closure of G. The main theorem (Theorem 13.11) states that, for every compact metrizable K with |K|>1, M_inv(K^S) is Bauer exactly when the closure of G has Property (T), and is the Poulsen simplex otherwise. The proof is model-theoretic: the paper develops affine Robinson theories, face-preserving and qf-simplicial theories, qf-measurable fields and direct integrals, and obtains a general dichotomy theorem (Theorem 10.6) for such theories under a decomposability hypothesis (D). This is then applied to the theory PMP_{G/H} constructed in Section 12, whose quantifier-free type spaces are identified with the simplices M_inv((2^x)^{G/H}). The paper also shows that the cube-exchangeability contexts of Aldous/Austin fall on the Poulsen side, confirming Austin's conjecture for those examples.","tokens_in":55027,"tokens_out":19499,"duration_ms":170446,"significance":"If the technical gaps identified below are repaired, this is a substantial contribution. It extends the Glasner–Weiss dichotomy from countable groups to a class of transitive permutation groups with a relative Property (T) assumption, and it provides the first general Bauer–Poulsen dichotomy subsuming both [BIT, Thm. 20.8] and the permutation-group examples. The model-theoretic machinery — qf-measurable fields, face-preserving Robinson theories, the qf-convex realization property, and the decomposability criterion of Theorem 8.4 — is genuinely new and likely to be reusable. The paper is also admirably transparent about the nonstandard strong definition of relative Property (T) and about the open equivalence question, and the applications to Austin's examples are correctly derived from ordinary Property (T) of the relevant closed subgroups.","major_comments":[{"comment":"Lemma 10.1 is stated for extreme types p ∈ E^qf_x(T), but in the proof of Theorem 10.6 it is applied to p0, which is chosen to be non-extreme. The printed expression 'p0 ∈ E^qf_x(T) \\overline{E^qf_x(T)}' is also self-contradictory; the intended choice is evidently p0 ∈ \\overline{E^qf_x(T)} \\setminus E^qf_x(T). The proof of the lemma, through Lemma 5.15, actually establishes the needed statement for every p in the closure of the types realized in M, so the issue is repairable. Nevertheless, the lemma as stated does not cover the use in the Claim, and the proof of the dichotomy must either state and prove the generalized version or be reorganized so that Lemma 10.1 is invoked only for types satisfying its hypothesis.","section":"§10, Lemma 10.1 and proof of Theorem 10.6"},{"comment":"The proof verifies convergence of ν_k to ν only on the constant cylinder sets U_{1^n} and U_{0^n}. These sets do not form a neighborhood basis of the product topology on 2^{G/H}; to conclude weak-* convergence to ν = λδ_1 + (1−λ)δ_0 one must prove the limits on mixed cylinders U_z. The missing limits (which are zero for mixed z) are recoverable from μ_k(a_k △ g_i a_k) → 0 by elementary estimates, but they are not present in the text. As written, the sentence 'This proves our claim' is not justified by the displayed computations.","section":"§4, Proposition 4.5"},{"comment":"The proof begins with the assertion that, by Remark 13.1, 'up to passing to a dense subgroup ... we may assume that G is countable.' This reduction is load-bearing, because the subsequent use of [BIT, Thm. 28.3] and the ergodic decomposition is explicitly said to require countability. The text does not explain why the decomposition obtained for the countable dense subgroup yields a decomposition of M as a model of the original theory PMP_{G/H}; one must check that the qf-measurable field and direct integral respect the full language, including predicates for elements of the larger closure, and that H-invariance of the fibers transfers from the dense subgroup to the whole stabilizer. Please provide a complete justification for this step, or restrict the statement of Theorem 13.11 accordingly if the reduction cannot be made to work.","section":"§13, Proposition 13.7"},{"comment":"Definition 4.1 is a strong ε-δ version of relative Property (T), and Remark 4.3 states that its equivalence with the standard weak definition is open for general Polish groups. Lemma 4.4 and Proposition 4.5 rely on exactly the strong form. The abstract and Theorem 13.11 state the hypothesis as 'relative Property (T)' without qualification, which could mislead readers who use the standard weak notion. I recommend adding an explicit qualification (e.g., 'in the sense of Definition 4.1') in the abstract and theorem statement, and noting in the introduction that the motivating examples are not affected because they satisfy the strong form through ordinary Property (T) of the relevant subgroup.","section":"Definition 4.1 and abstract/Theorem 13.11"}],"minor_comments":[{"comment":"The sentence 'The inequality (4.5) then follows directly from (i)' skips the case where max_{g∈F} μ(A△gA) ≥ ε²/4; in that case (4.5) is trivially true since the left-hand side is at most 1/2. Adding one sentence would make the argument complete.","section":"§4, Lemma 4.6(ii)"},{"comment":"In the construction of M^G, the assertion that d_0 is a pseudometric 'is indeed encoded in the theory' would be easier for the reader to verify if the relevant axioms of PMP_{G/H} were made explicit at that point or if a precise reference to the axioms were given.","section":"§12, Lemma 12.2"},{"comment":"The relation M ⪯^{ec*} N ('full extension') is defined inside the proof of Theorem 10.6 and used repeatedly in the Claim and the transfinite chain argument; it would improve readability to define it before the statement of the theorem or in a short preliminary paragraph.","section":"§10, proof of Theorem 10.6"},{"comment":"The reuse of G and H for the original permutation group and for its closure (with bars introduced in the text) is sometimes confusing, particularly in the statements of Proposition 13.7 and Lemma 13.6; a consistent notation with ar G and ar H throughout would help.","section":"Notation, Section 13"}],"recommendation":"major_revision","confidential_remarks":"The paper leans heavily on the preprint [BIT] by the same research group for core machinery (direct integrals, extremal decomposition, convex realization property). This dependency is disclosed, and the paper does adapt the needed tools to the quantifier-free setting, but the editor may wish to confirm that [BIT] is available and of sufficient standing, since Proposition 13.7 and Theorem 10.6 cite it at crucial points. The strong definition of relative Property (T) and the open equivalence question should be prominently flagged to avoid any appearance of overclaiming the standard relative Property (T) hypothesis."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper proves a Bauer–Poulsen dichotomy for invariant measure simplices of transitive permutation groups under relative Property (T), and it does so by building a genuinely new model-theoretic framework (qf-simplicial face-preserving Robinson theories). The main theorem is real, the finer-grained cube examples are genuinely new, and the proof is a long but carefully organized chain. Two caveats to know before relying on it: Definition 4.1 uses the strong epsilon–delta form of relative Property (T), whose equivalence with the standard weaker form is open for Polish groups, so the advertised scope for arbitrary permutation groups is conditional on that; and the proof of Proposition 4.5 checks convergence only on constant cylinders, not mixed ones, though the missing estimates are immediately recoverable from μ_k(a_k △ g_i a_k) → 0.\n\nWhat is new: Theorem 13.11 extends Glasner–Weiss from countable groups to transitive permutation groups without countability, under relative (T), and Corollary 13.12 settles Austin's open cases for cube-exchangeability. The model-theoretic core, Theorem 10.6, is a reusable Bauer–Poulsen dichotomy for quantifier-free type spaces of Robinson theories, going beyond [BIT] by working with existentially closed models rather than full affine type spaces. Appendix A also gives a clean treatment of open and face-preserving maps.\n\nThe soft spots are mostly exposure. Lemma 10.1 is stated for extreme types but used for a non-extreme type in the closure; the proof goes through for the broader class, so it is a harmless statement/use mismatch. The reliance on the preprint [BIT] is real—several key results are imported—but the author re-proves or adapts what he needs, and the new arguments (Lemma 4.4, Prop 4.5, Prop 13.7) are developed in detail. The definitional caveat is the one I would want a referee to press: if strong and weak relative (T) diverge for a closed permutation group, the theorem as stated only covers the strong version. The author is upfront about this in Remark 4.3, which counts for something.\n\nWho it is for: model theorists using affine logic, ergodic theorists working on invariant measure simplices, and people in exchangeability. It deserves a serious referee; the main theorem is significant, the proofs are detailed, and the caveats are openly stated. I would send it out.","headline":"Strong paper: a real Bauer–Poulsen dichotomy for permutation groups under a strong relative (T) hypothesis, with a reusable affine-logic framework; watch the definitional caveat and a small gap in Prop 4.5.","tokens_in":55555,"tokens_out":5129,"would_cite":true,"duration_ms":45564,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["03C98","22D10","37A15"],"pacs":[],"model":"deepseek-v4-flash","headline":"A transitive permutation group whose point stabilizer has relative Property (T) in its Polish closure yields a Bauer–Poulsen dichotomy for the simplex of invariant measures on $K^S$, with Bauer exactly when the closure has Property (T).","keywords":["Bauer–Poulsen dichotomy","relative Property (T)","invariant measures","affine logic","existentially closed models","Robinson theories","Choquet simplex","permutation groups"],"falsifier":"If one can produce a transitive permutation group $G$ on a countable set $S$ for which the closure of the point stabilizer has relative Property (T) only in the weak sense while the strong sense of Definition 4.1 fails, and for which $\\mathcal{M}_\\mathrm{inv}(2^S)$ is neither Bauer nor Poulsen, then the dichotomy's scope collapses at the boundary of the definition. A cheaper check is the finer-grained cube context $S=(\\mathbb{Z}/m\\mathbb{Z})^{\\oplus\\mathbb{N}}$ with $m>1$: the theorem predicts the Poulsen simplex for every such context, so exhibiting any one of these simplices that is not Poulsen would falsify the theorem.","tokens_in":54607,"feed_emoji":"📐","tokens_out":10262,"duration_ms":86722,"temperature":0.7,"pith_summary":"This paper proves a two-sided geometry theorem for the set $\\mathcal{M}_\\mathrm{inv}(K^S)$ of invariant probability measures on the product space $K^S$, where a countable transitive permutation group $G$ acts on $S$ and hence on coordinates. The theorem says: if the pointwise closure $\\overline{H}$ of the stabilizer of a point has relative Property (T) inside the Polish group $\\overline{G}$ (in the strong epsilon-delta sense of Definition 4.1), then $\\mathcal{M}_\\mathrm{inv}(K^S)$ is always either a Bauer simplex or the Poulsen simplex, and for $|K|>1$ it is Bauer exactly when $\\overline{G}$ itself has Property (T) and Poulsen otherwise. This subsumes the classical dichotomy for countable groups, settles the Bauer-vs-Poulsen question for the cube-exchangeability contexts left open by earlier work, and is obtained from a new model-theoretic dichotomy theorem about existentially closed models in affine logic.","feed_headline":"Relative property (T) forces a Bauer–Poulsen dichotomy","feed_subtitle":"Transitive groups: invariant-measure simplices are Bauer or Poulsen, settling the exchangeability question.","key_machinery":"The load-bearing objects are (i) the strong form of relative Property (T) for the pair $(\\overline{G},\\overline{H})$ of Polish groups, which upgrades approximate invariance to near exact invariance (Definition 4.1); (ii) the affine Robinson theory $\\mathrm{PMP}_{G/H}$, whose models are probability-measure-preserving $G$-systems generated by a distinguished $H$-invariant subalgebra, and whose quantifier-free type spaces are homeomorphic to $\\mathcal{M}_\\mathrm{inv}((2^x)^{G/H})$; and (iii) the abstract dichotomy Theorem 10.6, which says that a qf-simplicial, face-preserving, irreducible Robinson theory—meaning its quantifier-free type spaces are Choquet simplices, restriction maps send faces to faces, and models have the joint embedding property—is either qf-Bauer or qf-Poulsen provided every model of the common $\\forall\\exists$-theory of its affinely existentially closed, qf-extremal models is decomposable as a direct integral of qf-extremal models. Relative Property (T) enters through Lemma 13.4 and Proposition 13.7, which establish exactly this decomposability for $\\mathrm{PMP}_{G/H}$.","core_discovery":"The central claim (Theorem 13.11) is that for a countable transitive permutation group $G\\curvearrowright S$ with $H$ the stabilizer of a point, if the closure $\\overline{H}$ of $H$ has relative Property (T) in the Polish group $\\overline{G}$ according to Definition 4.1, then for every compact metrizable $K$ with $|K|>1$ the Choquet simplex $\\mathcal{M}_\\mathrm{inv}(K^S)$ is Bauer precisely when $\\overline{G}$ has Property (T), and otherwise it is the Poulsen simplex. The theorem removes all intermediate geometries: under this hypothesis the simplex is never neither Bauer nor Poulsen, and the boundary between the two cases is exactly Kazhdan's property for the full closure. The proof proceeds by encoding the invariant-measure problem in affine logic: the quantifier-free type spaces of the theory $\\mathrm{PMP}_{G/H}$ are affinely homeomorphic to the simplices of invariant measures, and a general Bauer–Poulsen dichotomy for qf-simplicial, face-preserving, irreducible Robinson theories (Theorem 10.6) applies once relative Property (T) is used to show that every model of the common $\\forall\\exists$-theory of full models is decomposable as a direct integral of qf-extremal models.","pith_inferences":["If the strong and weak forms of relative Property (T) are ever shown to differ for closed permutation groups, the scope of Theorem 13.11 is tied to Definition 4.1; a natural check is whether a weak-relative-(T) pair can produce a non-Bauer, non-Poulsen simplex.","The decomposability hypothesis (D) of Theorem 10.6 may hold for a wider class of affine Robinson theories than those arising from relative Property (T); in particular, proving openness and face-preservation of the variable-restriction maps of $\\mathrm{PMP}_{G/H}$ would extend the dichotomy to groups without relative Property (T).","The theorem suggests that relative Property (T) of a point stabilizer is the right topological replacement for oligomorphicity in exchangeability contexts: it yields the Poulsen side exactly when the full group lacks Kazhdan's property, whereas oligomorphicity would force the Bauer side."],"forward_implications":["The classical Bauer–Poulsen dichotomy for countable groups follows as the special case where the stabilizer is trivial, since the pair $(G,1)$ trivially has relative Property (T).","The cube-exchangeability example and its finer-grained variants $S=(\\mathbb{Z}/m\\mathbb{Z})^{\\oplus\\mathbb{N}}$ with the natural semidirect product action fall on the Poulsen side, confirming that no Aldous–Hoover–Kallenberg representation theorem is possible for them.","For any transitive group meeting the hypothesis there is no third geometry: the simplex of invariant measures is never neither Bauer nor Poulsen, answering that part of the open exchangeability question within this class.","When the closure of the full group has Property (T), the Bauer conclusion holds for every compact metrizable space $K$, not only for zero-dimensional ones, by the general Glasner–Weiss-type argument recalled in Proposition 3.3.","The theory $\\mathrm{PMP}_{G/H}$ is shown to be qf-Bauer or qf-Poulsen, so the model-theoretic dichotomy is available as a template for further ergodic-theoretic applications beyond permutation groups."],"supporting_citations":[{"why":"Provides the original Bauer–Poulsen dichotomy for countable groups and the proof strategy that Proposition 4.5 adapts to Polish groups.","marker":"[GW97]"},{"why":"Raises the open questions on exchangeability that the theorem answers and supplies the cube-exchangeable Poulsen example the result subsumes.","marker":"[Aus08]"},{"why":"Supplies the affine-logic framework, simplicial type spaces, direct integrals, and extremal decomposition results that Theorem 10.6 extends.","marker":"[BIT]"},{"why":"Shows that the relevant closure of the stabilizer in the cube-exchangeability example is the full symmetric group on a countable set, which has Property (T), placing the example in the theorem's scope.","marker":"[Tsa12]"},{"why":"Proves that oligomorphic Polish groups have Property (T), which explains the Bauer side in oligomorphic exchangeability contexts.","marker":"[ET16]"},{"why":"Supplies the construction of approximately invariant, $H$-fixed measurable sets from a representation without invariant vectors, adapted in Lemma 4.4.","marker":"[CW80]"},{"why":"Records the equivalence of weak and strong relative Property (T) for locally compact sigma-compact groups, against which the paper's Definition 4.1 is contrasted.","marker":"[Jol05]"},{"why":"Establishes uniqueness of the Poulsen simplex, which the Poulsen side of the dichotomy relies upon.","marker":"[LOS78]"}],"fun_headline_variants":["Bauer-Poulsen dichotomy from relative property (T)","Relative (T) decides Bauer vs Poulsen simplex","Invariant simplices: Bauer if (T), Poulsen else","Relative (T) yields Bauer or Poulsen invariant simplices"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument stands on the strong epsilon-delta version of relative Property (T) in Definition 4.1—approximate invariance inside a unitary representation must be upgraded to a genuinely close exactly invariant vector—and the paper notes that whether this strong form is equivalent to the standard weaker definition for general Polish groups is open, so the advertised scope of the main theorem depends on that definitional choice.","fun_headline_variants_meta":{"raw":{"variants":["Bauer-Poulsen dichotomy from relative property (T)","Relative (T) decides Bauer vs Poulsen simplex","Invariant simplices: Bauer if (T), Poulsen else","Relative (T) yields Bauer or Poulsen invariant simplices"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001276,"raw_usage":{"total_tokens":5243,"prompt_tokens":997,"completion_tokens":4246,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":613,"completion_tokens_details":{"reasoning_tokens":4177}},"tokens_in":613,"tokens_out":4246,"duration_ms":29771,"temperature":1.0,"reasoning_tokens":4177,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T00:29:04.145018+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If one can produce a transitive permutation group $G$ on a countable set $S$ for which the closure of the point stabilizer has relative Property (T) only in the weak sense while the strong sense of Definition 4.1 fails, and for which $\\mathcal{M}_\\mathrm{inv}(2^S)$ is neither Bauer nor Poulsen, then the dichotomy's scope collapses at the boundary of the definition. A cheaper check is the finer-grained cube context $S=(\\mathbb{Z}/m\\mathbb{Z})^{\\oplus\\mathbb{N}}$ with $m>1$: the theorem predicts the Poulsen simplex for every such context, so exhibiting any one of these simplices that is not Poulsen would falsify the theorem.","supporting_citations":[],"review_version":1}