{"id":"828efe65-322f-4b5f-8161-cbd3f50f694d","arxiv_id":"2507.03195","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every approximately treeable countable group, the existentially closed measure-preserving actions form an axiomatizable class.","lead":"This paper proves that for every approximately treeable countable group, the class of existentially closed probability-preserving actions admits a model companion, meaning those generic actions can be described by explicit axioms. It also supplies an ergodic-theoretic open mapping criterion for when this happens and shows that limit groups have a known denseness property called MD.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5.26's 'if' direction is under-proved: the open-mapping condition does not, by itself, give the claimed continuous-logic first-order axioms, and Theorem 7.4 rests on exactly this step.","rationale":"The reader's weakest_assumption correctly focuses on Theorem 5.26. My independent reading confirms that the 'if' direction is load-bearing: Theorem 7.4 proves openness of the relevant projection maps and then cites Theorem 5.26 to conclude that T*_Gamma exists. The proof of that direction, however, is only a two-sentence sketch. The issue is not the mathematical truth of the implication; a direct ultraproduct argument using Lemma 5.24 and a basic open set whose closure lies inside U appears to work and would bypass the first-order translation. But the preprint does not give that argument, and the asserted first-order translation is not automatic because open sets in the space of invariant measures are not clopen. Thus the central existence theorem is currently conditional on a missing proof detail. I am not rejecting the paper; the concern is precise and testable, and if the test succeeds the verdict should revert to ACCEPT.","tokens_in":159,"tokens_out":30813,"duration_ms":881894,"concrete_test":"Formalize the (<=) direction of Theorem 5.26. Concretely, either (i) write, for a basic closed V contained in pi_*(U) and containing nu, a family of L_Gamma forall-exists sentences whose models satisfy 'if (beta^Gamma)_*mu is in V then there exists gamma with ((beta x gamma)^Gamma)_*mu in U'; or (ii) prove directly from the open-mapping condition that e.c. actions are closed under ultraproducts: represent beta in the ultraproduct by beta_i via Lemma 5.24, choose a basic open U0 with closure contained in U and lambda in U0, use openness to get nu_i in pi_*(U0) eventually, apply e.c. of the factors to get gamma_i, and show the ultraproduct gamma lands in U. If neither can be done, Theorem 7.4 lacks a valid route to T*_Gamma.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The main theorem is proved by invoking Theorem 5.26, whose 'if' direction asserts that the open-mapping condition on the projections pi_* makes the e.c. criterion of Lemma 5.25 first-order. The proof's entire justification is the sentence: 'by assumption, pi_*(U) is an open subset of Prob^Gamma(q^Gamma), and it then suffices to consider basic open subsets contained in pi_*(U).' This skips the central difficulty. In continuous logic, an open hypothesis on the distribution of beta is not a zero-set condition, so 'for every beta with (beta^Gamma)_*mu in V' is not directly expressible as an forall-exists sentence unless V is chosen closed, with a corresponding restricted quantifier, or a separate closure-under-ultraproducts argument is supplied. The e.c. criterion Lemma 5.25 is the only bridge from ergodic-theoretic openness to axiomatizability; if this step fails, the existence of T*_Gamma for approximately treeable groups is unsupported even if all measure constructions in Section 7 are correct.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies existentially closed (e.c.) probability-measure-preserving (p.m.p.) actions of a countable group Γ in the framework of continuous model theory, and the existence of the model companion T*_Γ of the class K_Γ of all p.m.p. actions. The main theorem (Theorem 7.4) asserts that if Γ is approximately treeable, then T*_Γ exists. The proof uses an open-mapping criterion (Theorem 5.26) that characterizes existence of T*_Γ in terms of openness of certain push-forward maps on spaces of invariant measures, and then proves this criterion for approximately treeable groups via an elaborate measure-theoretic construction (Section 7). The paper also contains several substantial auxiliary results: limit groups have Kechris property MD (Theorem 4.4); profinite completion actions are e.c. for groups with property EMD (Theorem 4.2); and for groups without property (T), the generic e.c. action is weakly mixing (Corollary 4.9). For treeable groups, a concrete ergodic-theoretic axiomatization is given (Theorems 6.20 and 6.21).","tokens_in":66599,"tokens_out":7029,"duration_ms":85270,"significance":"The main theorem significantly generalizes the earlier result of Berenstein–Henson–Ibarlucía for free groups, covering the broad class of approximately treeable groups, which includes all treeable and all universally free groups. The open-mapping characterization in Theorem 5.26 is a powerful new tool that yields ergodic-theoretic axioms whenever the model companion exists. The proof of the main theorem is highly nontrivial and involves a detailed constructive measure-theoretic argument. The auxiliary results on property MD for limit groups, on e.c. profinite completions, and on weakly mixing e.c. actions are of independent interest. The paper is generally careful and self-contained, with many proofs given from first principles; if the main theorem is correct, this is a major advance in the model theory of measure-preserving actions.","major_comments":[{"comment":"The (⇐) direction of the open mapping characterization is proved in a single compressed paragraph. The condition in Lemma 5.25 is of the form 'for every β with (β^Γ)_*μ ∈ π_*(U), there exists γ ...' where U is open in Prob^Γ(q^Γ × p^Γ). Since membership in an open set is not a zero-set condition in continuous logic, this is not directly a first-order axiom. The sentence 'it then suffices to consider basic open subsets of Prob^Γ(q^Γ) contained in π_*(U)' does not explain how the strict inequalities defining such basic open sets are converted into closed first-order conditions. A complete proof should, for each basic open set V contained in π_*(U), approximate V from inside by closed (zero-set) neighborhoods and use a countable family of ∀∃-axioms to handle the resulting disjunction. Because Theorem 7.4 relies exactly on this implication, this step needs to be expanded.","section":"§5.5, Theorem 5.26"},{"comment":"The proof of the continuity of the map θ(ω,F) is too compressed. In particular, the paragraph beginning 'Partition Δ into a collection V of clopen sets' asserts that for each piece of the partition the map (ω,F,E_F) ↦ θ(ω,F)(A) is continuous, but the argument depends on the choice of representatives γ_C for the E_F-classes, and it is not shown that the resulting measure is independent of that choice and varies continuously as E_F varies. Since Lemma 7.9 is the technical core of the measure construction used in Theorem 7.4, the continuity proof should be written out in full detail, or the relevant continuity statement should be isolated as a separate lemma with a complete proof.","section":"§7.4, Lemma 7.9"}],"minor_comments":[{"comment":"The name 'Rokhlin' is frequently misspelled as 'Rohklin' (see, e.g., Sections 2, 4, and 6).","section":"Throughout"},{"comment":"The word 'aforemtnioned' is a typo for 'aforementioned'.","section":"Introduction and §5.4"},{"comment":"In the proof, the expression H1H0H−1_1 should be clarified as the set {h1 h0 h1^{-1} : h1 ∈ H1, h0 ∈ H0}; as written it could be misread as a product of sets with a single inverse applied to the whole product.","section":"§7.1, Lemma 7.2"}],"recommendation":"major_revision","confidential_remarks":"The main result is likely correct and the paper is a strong contribution, but the two proof gaps I identify (the first-order conversion in Theorem 5.26 and the continuity details in Lemma 7.9) are load-bearing and must be fixed or addressed by a detailed argument. Both are likely repairable with standard techniques, but the current text is too terse for publication. No concerns about novelty or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Plainly: this is a strong paper, but the proof of Theorem 5.26's 'if' direction is too compressed for the weight it carries. The stress-test note is on target. The paper claims that the open mapping condition makes Lemma 5.25 first-order, but the two sentences given ('by assumption pi_*(U) is open... it then suffices to consider basic open subsets') skip the continuous-logic work: membership in an open set is not a zero-set condition, and turning 'for every beta with distribution in V' into a first-order axiom requires either a countable approximation by closed conditions or a separate ultraproduct argument. As written, the step does not close.\n\nThat said, I don't think the paper is broken. The intended fix is standard — approximate the open set from inside by closed definable sets, or prove closure under ultraproducts directly using Lemma 5.25 — and the rest of the paper is detailed enough that I expect the gap is repairable. The dense measure constructions in Section 7 (map-measure pairs, theta, the final continuity argument) look internally coherent; the reader's 'no fatal gap' verdict for those sections matches my reading, with the caveat that the density makes verification slow.\n\nThe genuinely new content is substantial: Theorem 4.4 (limit groups have MD), the concrete treeable axiomatization in Section 6, and the approximately treeable model companion in Theorem 7.4. The open mapping characterization itself, Theorem 5.26, is a nice tool even if its proof needs expansion. The paper also handles extension-MD, EMD profinite completions, and weakly mixing genericity — all real extensions of the BHI results.\n\nThe citation pattern is fine; self-citations support auxiliary lemmas, not the main theorem. No circularity red flag.\n\nWho should read this: anyone working on continuous model theory of group actions, weak containment, and treeability. It deserves a serious referee — the main result is important and likely correct, but the referee must demand a complete proof of Theorem 5.26 and a cleaner ultraproduct closure argument. I would send it to a top journal.","headline":"A serious, technically dense paper that generalizes BHI to approximately treeable groups; the main theorem is probably right, but the open mapping proof in Theorem 5.26 is under-written at exactly the load-bearing step.","tokens_in":67135,"tokens_out":5228,"would_cite":true,"duration_ms":62934,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["03C98","37A15","37A20","20F65"],"pacs":[],"model":"deepseek-v4-flash","headline":"For approximately treeable groups, the class of measure-preserving actions has a model companion: existentially closed actions are first-order axiomatizable and closed under ultraproducts.","keywords":["probability-measure-preserving actions","model companion","existentially closed actions","approximately treeable groups","treeable groups","continuous model theory","ergodic theory","property MD"],"falsifier":"Check whether the projection map $\\pi_* : \\operatorname{Prob}^{F_2 \\times \\mathbb{Z}}(q^{F_2 \\times \\mathbb{Z}} \\times p^{F_2 \\times \\mathbb{Z}}) \\to \\operatorname{Prob}^{F_2 \\times \\mathbb{Z}}(q^{F_2 \\times \\mathbb{Z}})$ is open for the approximately treeable group $F_2 \\times \\mathbb{Z}$: if some pair $p,q$ gives a non-open image, then Theorem 7.4 is false. Equivalently, build a nonprincipal ultraproduct of existentially closed actions of an approximately treeable group and test whether it remains existentially closed; any failure would refute the main claim.","tokens_in":66197,"feed_emoji":"🎲","tokens_out":8189,"duration_ms":88944,"temperature":0.7,"pith_summary":"Given a countable group $\\Gamma$, this paper asks when the class $\\mathcal{K}_\\Gamma$ of probability-measure-preserving (p.m.p.) actions of $\\Gamma$ has a model companion: a first-order theory whose models are exactly the existentially closed actions. The central result is that if $\\Gamma$ is approximately treeable, then the model companion of $\\mathcal{K}_\\Gamma$ exists. The class of approximately treeable groups contains all treeable groups, all universally free groups, and examples such as $F_2 \\times \\mathbb{Z}$, so the result substantially enlarges the known cases. The proof goes through an open mapping characterization of the existence of the model companion, which also yields ergodic-theoretic axioms; for treeable groups these axioms become especially simple. Along the way, the paper shows that limit groups have property MD, that profinite completion actions of EMD groups are existentially closed, and that groups without property (T) admit weakly mixing existentially closed actions.","feed_headline":"Approximately treeable groups get a model companion","feed_subtitle":"Generic measure-preserving actions of these groups form an axiomatizable first-order class.","key_machinery":"The load-bearing mechanism is the open mapping characterization of Theorem 5.26: $T^*_\\Gamma$ exists if and only if every projection map $\\pi_* : \\operatorname{Prob}^\\Gamma(q^\\Gamma \\times p^\\Gamma) \\to \\operatorname{Prob}^\\Gamma(q^\\Gamma)$ is open. Approximate treeability enters through invariant Borel probability measures on the space $\\mathcal{F}(\\Gamma)$ of directed forests on $\\Gamma$, whose component equivalence relations can approximate the indiscrete relation. The measure construction re-randomizes fiber measures independently over the connected components of such forests, producing invariant joint measures with prescribed first marginal. For the explicit axioms, the paper introduces the extension-MD property, meaning finite-to-one extensions are dense among all extensions of a given action, and the definable cocycle property, meaning almost-cocycles to finite groups are uniformly near actual cocycles; both are verified for strongly treeable groups, and suitably adapted for treeable groups, using retractions $r_T$ from the space of cochains to the space of cocycles attached to trees $T$.","core_discovery":"The paper proves Theorem 7.4: if $\\Gamma$ is an approximately treeable group, then the model companion $T^*_\\Gamma$ of the class of p.m.p. actions of $\\Gamma$ exists. Equivalently, the existentially closed actions form an axiomatizable class and are closed under ultraproducts. The proof establishes an open mapping criterion: $T^*_\\Gamma$ exists exactly when, for every $p, q \\in \\mathbb{N}$, the push-forward map $\\pi_* : \\operatorname{Prob}^\\Gamma(q^\\Gamma \\times p^\\Gamma) \\to \\operatorname{Prob}^\\Gamma(q^\\Gamma)$ is open in the weak-$*$ topology. For approximately treeable groups, the openness is verified by constructing invariant measures on $q^\\Gamma \\times p^\\Gamma$ with prescribed first marginal, using invariant probability measures on the space of directed forests of $\\Gamma$. For treeable groups, the paper gives a concrete characterization: an action is existentially closed exactly when it weakly contains a free treeable action, its trivial extension with atomless fibers is an existentially closed extension, and coboundaries are dense among cocycles valued in every finite symmetric group.","pith_inferences":["The open mapping criterion reduces the existence of the model companion to checking openness of a single family of projection maps, so candidate counterexamples to the open question of whether every countable group has a model companion could be sought by testing non-openness for specific groups.","The treeable-group characterization suggests that being existentially closed is largely a cohomological saturation condition: for such groups, failure of density of coboundaries among $\\operatorname{Sym}(k)$-valued cocycles would be a concrete, model-theory-free obstruction to being existentially closed.","The forest re-randomization construction resembles an independence argument over approximate equivalence relations; it may extend to other groups with approximate ergodic dimension at most one, or to measure-equivalence invariants, though those extensions are not claimed in the paper.","One could attempt to write down explicit axioms for $F_2 \\times \\mathbb{Z}$ using the general open mapping axioms and check whether the resulting conditions detect the failure of treeability."],"forward_implications":["For every approximately treeable group $\\Gamma$, the existentially closed p.m.p. actions are exactly the models of a single first-order theory $T^*_\\Gamma$, so they are closed under ultraproducts.","Whenever the model companion exists, the open mapping criterion provides ergodic-theoretic axioms, so the existence question becomes a concrete topological property of equivariant measure spaces.","All treeable groups, all universally free groups, and groups such as $F_2 \\times \\mathbb{Z}$ now have a model companion; existence also passes to subgroups and to extensions by coamenable normal subgroups with a model companion.","For treeable groups, an action is existentially closed if and only if it weakly contains a free treeable action, its trivial extension with atomless fibers is an existentially closed extension, and $B^1(a, \\operatorname{Sym}(k))$ is dense in $Z^1(a, \\operatorname{Sym}(k))$ for every $k$.","Groups with property EMD have existentially closed profinite completion actions, and groups without property (T) have weakly mixing existentially closed actions in the enforceable sense."],"supporting_citations":[{"why":"Supplies the prior model-companion result for free groups that this paper generalizes and whose explicit axioms are replaced by ergodic-theoretic ones.","marker":"[5]"},{"why":"Introduces approximately treeable groups under the guise of approximate ergodic dimension at most one, the class featured in the main theorem.","marker":"[20]"},{"why":"Provides the measure-equivalence description of treeability used to position the class of treeable groups.","marker":"[26]"},{"why":"Defines properties MD and EMD and supplies the profinite approximation facts used to construct concrete existentially closed actions.","marker":"[30]"},{"why":"Gives the free coamenable normal subgroup in limit groups, used to show universally free groups are approximately treeable.","marker":"[34]"},{"why":"Provides subgroup separability for limit groups, used in the proof that limit groups have property MD.","marker":"[43]"},{"why":"Shows Bernoulli actions are weakly contained in every free action, used to transfer treeing measures to all free actions of strongly treeable groups.","marker":"[1]"},{"why":"Provides hyperfiniteness of amenable orbit equivalence relations, used in the coamenable subgroup and approximate treeability arguments.","marker":"[36]"}],"fun_headline_variants":["Approximately treeable groups have model companions","Existentially closed actions for approximately treeable groups","Model companion exists for approximately treeable groups","Generalizing free groups: model companions for pmp actions","Axiomatizable actions for approximately treeable groups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result depends on the open mapping characterization of when the model companion exists: if that equivalence has a hidden gap, especially in the step converting the existentially closed criterion into first-order axioms, the main theorem would not follow even if the measure constructions are correct.","fun_headline_variants_meta":{"raw":{"variants":["Approximately treeable groups have model companions","Existentially closed actions for approximately treeable groups","Model companion exists for approximately treeable groups","Generalizing free groups: model companions for pmp actions","Axiomatizable actions for approximately treeable groups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000214,"raw_usage":{"total_tokens":1514,"prompt_tokens":1123,"completion_tokens":391,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":739,"completion_tokens_details":{"reasoning_tokens":319}},"tokens_in":739,"tokens_out":391,"duration_ms":5312,"temperature":1.0,"reasoning_tokens":319,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:15:57.493534+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check whether the projection map $\\pi_* : \\operatorname{Prob}^{F_2 \\times \\mathbb{Z}}(q^{F_2 \\times \\mathbb{Z}} \\times p^{F_2 \\times \\mathbb{Z}}) \\to \\operatorname{Prob}^{F_2 \\times \\mathbb{Z}}(q^{F_2 \\times \\mathbb{Z}})$ is open for the approximately treeable group $F_2 \\times \\mathbb{Z}$: if some pair $p,q$ gives a non-open image, then Theorem 7.4 is false. Equivalently, build a nonprincipal ultraproduct of existentially closed actions of an approximately treeable group and test whether it remains existentially closed; any failure would refute the main claim.","supporting_citations":[{"cited_title":"Berenstein, C","cited_arxiv_id":null,"evidence_quote":"Supplies the prior model-companion result for free groups that this paper generalizes and whose explicit axioms are replaced by ergodic-theoretic ones."},{"cited_title":"Gaboriau,Invariants ℓ2 de relations d’équivalence et de groupes, Publications mathéma- tiques de l’IHÉS95 (2002), 93-150","cited_arxiv_id":null,"evidence_quote":"Introduces approximately treeable groups under the guise of approximate ergodic dimension at most one, the class featured in the main theorem."},{"cited_title":"Hjorth,A lemma for cost attained, Ann","cited_arxiv_id":null,"evidence_quote":"Provides the measure-equivalence description of treeability used to position the class of treeable groups."},{"cited_title":"Kechris,Weak containment in the space of actions of a free group, Israel J","cited_arxiv_id":null,"evidence_quote":"Defines properties MD and EMD and supplies the profinite approximation facts used to construct concrete existentially closed actions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the free coamenable normal subgroup in limit groups, used to show universally free groups are approximately treeable."},{"cited_title":"Wilton,Hall’s theorem for limit groups, Geom","cited_arxiv_id":null,"evidence_quote":"Provides subgroup separability for limit groups, used in the proof that limit groups have property MD."},{"cited_title":"2, 323-333","cited_arxiv_id":null,"evidence_quote":"Shows Bernoulli actions are weakly contained in every free action, used to transfer treeing measures to all free actions of strongly treeable groups."},{"cited_title":"Ornstein and B","cited_arxiv_id":null,"evidence_quote":"Provides hyperfiniteness of amenable orbit equivalence relations, used in the coamenable subgroup and approximate treeability arguments."}],"review_version":1}