{"id":"67b824ed-ba19-4cab-ae1c-bb4c3ce0ce7c","arxiv_id":"2509.04613","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every virtually special action of a countable group on a CAT(0) cube complex yields a hyperfinite orbit equivalence relation on its Roller boundary.","lead":"This paper proves that any countable group acting virtually specially on a CAT(0) cube complex generates a hyperfinite orbit equivalence relation on the Roller boundary. This extends prior hyperfiniteness results from hyperbolic groups to a much wider class, including right-angled Artin groups.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section 4's reduction hinges on an unproved but true boundary-extension lemma; the rest of the proof appears coherent, so the paper needs a small addition rather than a correction.","rationale":"I read the paper in good faith. The main theorem is plausible and the proof strategy is coherent: reduce a virtually special action to a right-angled Artin group via Haglund-Wise, prove hyperfiniteness for RAAG boundary actions by a smoothness argument, then transfer back. The RAAG proof in Section 3 is detailed and, on inspection, the key steps — Lemma 3.9's complete invariant, the countable-to-one Borel map in Proposition 3.12, and the subrelation arguments — appear sound. The weakest point is exactly what the reader identified: the boundary transfer in Section 4. The assertion that a convex embedding induces an injection on Roller boundaries is not proved or cited, and equation (13) is load-bearing. I believe the missing lemma is true and standard, so this is a presentation gap rather than a mathematical error. I found no independent fatal flaw; the dependence on the author's preprint [Oya25] is confined to a remark and to Section 5, not to the main reduction. The appropriate disposition is therefore unchanged from the reader's verdict: conditional acceptance pending a written proof of the transfer lemma and the Borelness of pψ(BRX).","tokens_in":20189,"tokens_out":40850,"duration_ms":412471,"concrete_test":"Write out and verify the missing transfer lemma in the special case C = rψ(X), D = X(Γ): prove that a convex embedding of CAT(0) cube complexes extends to a continuous, injective, equivariant map on Roller boundaries, that pψ(BRX) is Borel, and that (13) holds as an equality of equivalence relations. If this lemma cannot be proved, the reduction in Theorem 4.1 to Proposition 3.12 fails; if it is proved, the central argument goes through.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section 4, after Lemma 2.26, the paper asserts: 'Hence, rψ induces the injection pψ : BRX → BRX(Γ)' and states equation (13), which identifies E_H on BRX with the pulled-back orbit relation of ψ_*(π1(Y,y)) on pψ(BRX). No proof or citation is given for the boundary extension. This is genuinely load-bearing: the reduction to the RAAG case (Proposition 3.12) works only if pψ is injective, equivariant, and has Borel image. The statement is very likely true: for a convex subcomplex C of a CAT(0) cube complex D, each ambient hyperplane either meets C (in which case it gives a hyperplane of C) or lies entirely on one side of C, so a Roller boundary point of C extends canonically to one of D; injectivity follows because distinct boundary points of C are separated by a hyperplane of C, which is also an ambient hyperplane; equivariance follows from Remark 2.27 on vertices and continuity; and Borelness of pψ(BRX) follows from Lusin-Souslin since pψ is continuous and injective on a Polish space. But none of this is written in the paper, and equation (13) is asserted without derivation. A careful reader cannot verify the central reduction without supplying this missing lemma.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that if a countable group G acts virtually specially on a CAT(0) cube complex X, then the orbit equivalence relation induced by G on the Roller boundary BRX is hyperfinite (Theorem 1.2). This generalizes the hyperfiniteness theorem of Huang–Sabok–Shinko for boundary actions of cubulated hyperbolic groups. The proof proceeds in three parts. Section 3 treats the case of right-angled Artin groups from scratch: Lemma 3.9 constructs an explicit invariant f for the action of A(Γ) on the space H^N of hyperplane sequences and shows that f separates orbits exactly, so the orbit relation is smooth; Corollary 3.10 and Proposition 3.12 then derive hyperfiniteness of the action on the Roller boundary BRX(Γ) using tail equivalence relations. Section 4 reduces the general case to the RAAG case: a finite-index subgroup H of G acts freely with special quotient Y, the Haglund–Wise theorem gives a local isometry ψ : Y → R(Γ), its lift rψ : X → X(Γ) is a convex embedding by Lemma 2.26, and the paper asserts that rψ induces an injection pψ : BRX → BRX(Γ) satisfying equation (13), which identifies the H-orbit relation on BRX with the ψ_*(π1(Y,y))-orbit relation on pψ(BRX). Section 5 shows that the main theorem indeed generalizes [HSS20], via a finite-to-one surjective equivariant Borel map from the Roller boundary to the Gromov boundary of a hyperbolic CAT(0) cube complex (Lemmas 5.1–5.4 and Proposition 5.5).","tokens_in":20391,"tokens_out":37306,"duration_ms":335195,"significance":"If the proof is completed as indicated, this is a substantial and natural generalization of [HSS20], extending hyperfiniteness of boundary actions from cubulated hyperbolic groups to all countable groups admitting a virtually special cubical action, and it bears directly on Question 1.1 (measure-hyperfinite versus hyperfinite). The RAAG case is the core contribution and is genuinely from scratch: Lemma 3.9 gives an explicit, parameter-free smoothness witness for the action on the hyperplane sequences, with the disk-diagram lemmas 3.2–3.5 supplying the geometry, and there is no fitting, hidden assumption, or use of the target theorem as an input. The applications in Corollaries 4.2–4.4 (hyperbolic groups, contact graphs, extension graphs) are attractive and clearly derived. The descriptive set theory is standard and carefully cited (Dougherty–Jackson–Kechris, Jackson–Kechris–Louveau), and the paper is honest about which steps are routine.","major_comments":[{"comment":"The reduction to the RAAG case hinges on the assertion 'rψ induces the injection pψ : BRX → BRX(Γ)', followed by equation (13). Lemma 2.26, as cited, establishes only that the lift rψ is an embedding of X as a convex subcomplex of X(Γ); it says nothing about Roller boundaries. To make the reduction work, the paper needs to prove: (a) existence of a canonical extension of Roller boundary points from the convex subcomplex rψ(X) to X(Γ) (an ambient hyperplane either crosses rψ(X), in which case it restricts to a hyperplane of X, or lies entirely on one side of rψ(X)); (b) injectivity (distinct points of BRX are separated by a hyperplane of X, whose ambient counterpart separates their images); (c) equivariance with respect to the π1(Y,y)-action and the ψ_*(π1(Y,y))-action, using Remark 2.27; and (d) continuity, hence Borelness, of pψ and Borelness of the image pψ(BRX), for example via Lusin–Souslin. None of this appears in the manuscript, and equation (13), which identifies the orbit relation of H on BRX with the pulled-back orbit relation of ψ_*(π1(Y,y)) on pψ(BRX), is asserted without derivation. The statement is very likely correct, but as written a careful reader cannot verify the central reduction; please add this as a lemma with a full proof.","section":"Section 4 (after Lemma 2.26; Eq. (13))"},{"comment":"Smoothness of E^{H^N}_{A(Γ)} requires the invariant map to be Borel, but the Borelness of the map f constructed in Lemma 3.9 is never stated or proved; the construction is an inductive procedure (the sets A_n, B_n, C_n, D_n, the stabilization C⃗_h, and the words s_n, t_n) whose measurability is not self-evident and should be justified. Likewise, in Proposition 3.12 the Borelness of the partition sets BRX(Γ)_I and of the map f : BRX(Γ)_I → (H^N)^I is only asserted with 'it is not difficult to see'. These measurability facts are load-bearing: they are what allow Lemma 3.1 and Corollary 3.10 to conclude hyperfiniteness. The arguments are likely short (each step is definable from countable data: the A(Γ)-orbit of a hyperplane and the ordering of H_i(o,ξ)), but they should be written down explicitly.","section":"Section 3 (Lemma 3.9 and Proposition 3.12)"}],"minor_comments":[{"comment":"The sentence 'any finite dimensional CAT(0) cube complex with countably many hyperplanes is countable' is asserted without proof. A short argument suffices: fixing a base vertex o, every vertex v of the 1-skeleton is connected to o by a finite path, so v is separated from o by only finitely many hyperplanes; hence v is determined by a finite subset of the countable set H(X), and X(0) is countable. Please include this argument.","section":"Theorem 4.1 (preamble)"},{"comment":"The step 'by applying Theorem 2.15 and Theorem 2.16 to the set {(x,f(x)) : x ∈ X}, there exists a Borel subset A ⊂ X such that f|A : A → Y is Borel isomorphic' is too compressed. The natural argument is to Borel-uniformize the fibers f^{-1}(y) with Lusin–Novikov, obtaining a Borel section s : Y → X, and then note that A = s(Y) is Borel by Lusin–Souslin since s is injective and Borel. Please spell this out and clarify the role of Arsenin–Kunugui.","section":"Lemma 5.1 (proof)"},{"comment":"The step 'By f(x) = f(y), we have d_X(p_n,q_n) ≤ δ' is asserted without justification. It follows from Proposition 2.30(1) by applying condition (1) to the geodesic triangle with vertices o, p_N, q_M for N, M large enough that n ≤ (p_N,q_M)^S_o; please include this line, since the displayed inequality is otherwise unmotivated.","section":"Lemma 5.3(2), (iii) ⇒ (i)"},{"comment":"The displayed equivalence preceding 'we can see that BRX(Γ)_I is Borel' appears garbled in the manuscript (the second half of the display does not parse). Please rewrite this line cleanly, and also justify explicitly that every ξ ∈ BRX(Γ) belongs to some BRX(Γ)_I with I nonempty (i.e., that H(o,ξ) is infinite for boundary points, so some H_i(o,ξ) is infinite).","section":"Proposition 3.12"},{"comment":"In the proof of Lemma 5.2(4), the inequality |F| + 2 ≤ |Lk_G(p_N)| assumes that the points supplied by Lemma 5.2(3) for the elements of F are distinct from p_{N-1} and p_{N+1} and from each other. This should be said explicitly; distinctness follows because the auxiliary edges are dual to distinct hyperplanes and hyperplanes in a CAT(0) cube complex do not self-osculate.","section":"Lemma 5.2(4)"},{"comment":"Remark 3.13 sketches an alternative route through the quotient BRX(Γ)/G and cites [Oya25, Theorem 1.1]. Since the remark is not used in the proof of Proposition 3.12, either delete it or state explicitly that it is optional; the main text should be self-contained.","section":"Remark 3.13"}],"recommendation":"major_revision","confidential_remarks":"The two major issues are both, in my assessment, short fixes rather than signs of a deeper problem: the Roller-boundary injection under a convex embedding is true (each ambient hyperplane either crosses the convex subcomplex or lies entirely on one side of it), and the measurability claims in Section 3 follow from the explicit, countable-data nature of the constructions. I would therefore expect a focused revision to suffice. The self-citations ([Oya24], [Oya25]) occur only in supporting remarks and lemmas and do not carry the main argument; I see no citation-pattern problem. The paper is a good fit for this journal's readership, sitting between geometric group theory and descriptive set theory."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead Oyakawa's paper. The main theorem is genuinely new: any countable group acting virtually specially on a CAT(0) cube complex has a hyperfinite orbit equivalence relation on the Roller boundary. This extends Huang-Sabok-Shinko and Marquis-Sabok beyond hyperbolic groups to RAAGs and other virtually special groups.\n\nThe RAAG case is the real meat. Lemma 3.9, the smoothness of the action on the space of hyperplane sequences, is a solid argument and deserves to be written up carefully. The reductions around it—splitting the boundary by the infinite hyperplane sets, using the well-order on hyperplanes, and then assembling the hyperfinite relations—are coherent. I checked the logical flow and I don't see circularity or a hidden use of a target result. The use of Haglund-Wise to reduce to RAAGs is standard.\n\nThe one genuine soft spot is in Section 4, right after Lemma 2.26. The paper says \"Hence, rψ induces the injection pψ : BRX → BRX(Γ)\" and then asserts (13), the pullback identification of the orbit relations. No proof is given. This is load-bearing: without an injective, equivariant, Borel boundary extension, the reduction to the RAAG case fails. The stress-test note is right that the statement is true—convex subcomplex, each ambient hyperplane either meets the subcomplex or lies entirely on one side, so a Roller boundary point of the domain has a canonical extension; injectivity comes from separation by hyperplanes of the subcomplex; Borelness comes from Lusin-Souslin—but none of that is written. A careful reader cannot verify the main theorem without supplying this. It is a missing lemma, not a flaw in the strategy.\n\nMinor: Section 5 leans on two of the author's preprints [Oya25] and [FLM24] for the comparison with Gromov boundaries. That is acceptable, but the referee should ask for those results to be stated precisely or made available in a checkable form. Lemma 5.4 is also quite compressed; surjectivity and continuity deserve a few more lines.\n\nOverall: the central argument holds together, the new material is real, and the gap is fillable. I'd send it to a good referee and expect a revision rather than a rejection. The paper will be useful to anyone working on Borel complexity of group boundary actions.","headline":"New, correct-in-outline proof that virtually special cubical actions have hyperfinite Roller boundary orbit equivalence relations, with one omitted boundary-extension lemma that should be added before publication.","tokens_in":20977,"tokens_out":1869,"would_cite":true,"duration_ms":17476,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20F65","03E15","37A20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every virtually special cubical action has a hyperfinite orbit equivalence relation on the Roller boundary.","keywords":["hyperfinite equivalence relation","Roller boundary","CAT(0) cube complex","virtually special groups","right-angled Artin groups","orbit equivalence relation","cubulated hyperbolic groups","descriptive set theory"],"falsifier":"Find a convex embedding $X \\hookrightarrow Y$ of CAT(0) cube complexes such that two distinct points of $\\mathrm{BR}X$ are identified in $\\mathrm{BR}Y$, or such that the induced boundary map fails to be equivariant with respect to the relevant subgroups; if such an example exists, equation (13) and the reduction to RAAG boundary actions collapse. If no such example exists, a proof of boundary injectivity for convex embeddings would close the gap.","tokens_in":19929,"feed_emoji":"🧊","tokens_out":15253,"duration_ms":138628,"temperature":0.7,"pith_summary":"The paper proves that for any countable group G acting virtually specially on a CAT(0) cube complex X, the orbit equivalence relation induced by G on the Roller boundary of X is hyperfinite: it can be written as an increasing union of finite Borel equivalence relations. This is the first hyperfiniteness result for Roller boundaries of non-hyperbolic cubical actions, and it generalizes the Huang–Sabok–Shinko theorem that cubulated hyperbolic groups act hyperfinitely on their Gromov boundaries. The proof proceeds by encoding Roller boundary points of a right-angled Artin group's cube complex as ordered sequences of labeled hyperplanes, reducing the orbit relation to a tail equivalence relation, and then transferring hyperfiniteness from RAAG actions to virtually special actions through a convex embedding into a RAAG cube complex. If correct, the theorem gives a uniform answer for a large class of nonpositively curved group actions: their natural boundary orbit relations are as simple as Borel equivalence relations can be in this hierarchy.","feed_headline":"Virtually special groups act hyperfinitely on Roller boundaries","feed_subtitle":"The theorem makes these geometric boundary actions as simple as Borel equivalence relations can be.","key_machinery":"The machine is the boundary coding for right-angled Artin groups. Let $\\Gamma$ be a finite graph, $A(\\Gamma)$ the right-angled Artin group, $X(\\Gamma)$ its standard CAT(0) cube complex, and $\\mathrm{BR}X(\\Gamma)$ the Roller boundary. Hyperplanes of $X(\\Gamma)$ are partitioned into finitely many label classes $H_i$, one per generator; hyperplanes in one class never cross, so the hyperplanes in $H_i$ separating a fixed vertex $o$ from a boundary point $\\xi$ can be listed in a canonical order by distance from $o$. The proof shows that the $A(\\Gamma)$-orbit of a Roller boundary point is determined by the tail of this ordered hyperplane sequence. The tail equivalence relation on the countable set of hyperplanes is hyperfinite by [DJK94, Theorem 8.1], giving Proposition 3.12. The convex projections $\\psi_A^B$ between convex sets and the stabilization of cosets $C_n \\subseteq C_{n-1} \\subseteq \\cdots$ (finite graph forces eventual constancy) provide the invariant used to prove smoothness of the orbit relation on hyperplane sequences. For the general theorem, a local isometry $X/H \\to R(\\Gamma)$ lifts to a convex embedding $X \\to X(\\Gamma)$, and the induced injection $p_\\psi \\colon \\mathrm{BR}X \\to \\mathrm{BR}X(\\Gamma)$ carries the RAAG hyperfiniteness back.","core_discovery":"On the paper's own terms, the central discovery is Theorem 1.2: if X is a CAT(0) cube complex and G is countable and acts virtually specially on X, then the orbit equivalence relation $E^{{BRX}}$_G on the Roller boundary is hyperfinite. The stronger Theorem 4.1 isolates the mechanism: a finite-index subgroup H acting freely with special compact quotient X/H suffices, because Proposition 2.14 transfers hyperfiniteness across finite-index supergroups. The RAAG case (Proposition 3.12) is the engine: for a finite graph Γ, the A(Γ)-action on the Roller boundary of X(Γ) is hyperfinite. Section 5 then verifies the advertised generalization by exhibiting a continuous, surjective, finite-to-one, equivariant map from the Roller boundary of a uniformly locally finite hyperbolic CAT(0) cube complex to its Gromov boundary, so hyperfiniteness of the former is equivalent to hyperfiniteness of the latter.","pith_inferences":["The transition from the convex embedding of complexes to the injection $p_\\psi$ of Roller boundaries in Section 4 is asserted rather than proved; a testable question is whether every convex subcomplex inclusion of CAT(0) cube complexes induces an injective Roller-boundary map, and if so, whether the same argument extends to arbitrary convex subcomplex inclusions.","The RAAG coding depends only on finitely many pairwise non-crossing label classes and on eventual stabilization of convex-projection cosets, so the same tail-equivalence argument should apply to any cubical action whose hyperplane system admits such a finite labeling, a class potentially wider than virtually special groups.","Proposition 5.5 sets up a transfer principle: for hyperbolic CAT(0) cube complexes, hyperfiniteness of the Roller-boundary action is equivalent to hyperfiniteness of the Gromov-boundary action. One could test whether a similar finite-to-one equivariant transfer holds for other pairs of boundary constructions, such as the regular boundary and the full Roller boundary."],"forward_implications":["Every cocompactly cubulated hyperbolic group acts hyperfinitely on the Roller boundary of its cube complex; by the Section 5 equivalence, this recovers the Huang–Sabok–Shinko theorem on Gromov boundaries.","For any virtually special action, the induced action on the Gromov boundary of the contact graph is hyperfinite (Corollary 4.3).","For every right-angled Artin group $A(\\Gamma)$, the Roller-boundary action on $X(\\Gamma)$ is hyperfinite, and so is the action on the Gromov boundary of its extension graph (Proposition 3.12 and Corollary 4.4).","Since hyperfinite implies measure-hyperfinite, every boundary action covered by the theorem is measure-hyperfinite with respect to every Borel probability measure."],"supporting_citations":[{"why":"Characterizes special cube complexes by the existence of a local isometry into a RAAG Salvetti complex, providing the bridge from special quotients to right-angled Artin groups in Theorem 4.1.","marker":"[HW08, Theorem 1.1]"},{"why":"Gives that the lift of this local isometry embeds the universal cover as a convex subcomplex, the embedding that the proof then extends to a Roller-boundary injection.","marker":"[Wis12, Lemma 3.12]"},{"why":"Supplies hyperfiniteness of tail equivalence relations on countable sets, the endpoint of the RAAG boundary coding.","marker":"[DJK94, Theorem 8.1]"},{"why":"Used to transfer hyperfiniteness from a finite-index subgroup action and from a subrelation with finitely many classes to the full orbit relation.","marker":"[JKL02, Proposition 1.3.(vii)]"},{"why":"Provides the disk-diagram corner-move argument behind Lemma 3.2, from which the convexity and coset invariants used in the RAAG coding are derived.","marker":"[Sag95, Lemma 4.2]"},{"why":"The result on cubulated hyperbolic groups' Gromov boundaries that the paper generalizes and later recovers as Corollary 5.6.","marker":"[HSS20, Theorem 1.1]"},{"why":"Shows that cocompactly cubulated hyperbolic groups are virtually special, letting Corollary 4.2 apply the main theorem to them.","marker":"[Ago13, Theorem 1.1]"}],"fun_headline_variants":["Hyperfinite action: virtually special groups tame Roller boundaries","Roller boundary actions of virtually special groups are hyperfinite","Virtually special groups: Roller boundary actions hyperfinite","Hyperfiniteness for virtually special group actions on Roller boundaries","Cubulated groups: Roller boundary actions are hyperfinite"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a convex copy of one cube complex inside another sends distinct Roller-boundary points to distinct boundary points and respects the group action; the cited lemma establishes only the embedding at the level of complexes, not this boundary extension.","fun_headline_variants_meta":{"raw":{"variants":["Hyperfinite action: virtually special groups tame Roller boundaries","Roller boundary actions of virtually special groups are hyperfinite","Virtually special groups: Roller boundary actions hyperfinite","Hyperfiniteness for virtually special group actions on Roller boundaries","Cubulated groups: Roller boundary actions are hyperfinite"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001082,"raw_usage":{"total_tokens":4442,"prompt_tokens":782,"completion_tokens":3660,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":398,"completion_tokens_details":{"reasoning_tokens":3575}},"tokens_in":398,"tokens_out":3660,"duration_ms":24545,"temperature":1.0,"reasoning_tokens":3575,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:30:49.533891+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a convex embedding $X \\hookrightarrow Y$ of CAT(0) cube complexes such that two distinct points of $\\mathrm{BR}X$ are identified in $\\mathrm{BR}Y$, or such that the induced boundary map fails to be equivariant with respect to the relevant subgroups; if such an example exists, equation (13) and the reduction to RAAG boundary actions collapse. If no such example exists, a proof of boundary injectivity for convex embeddings would close the gap.","supporting_citations":[],"review_version":2}