{"id":"e5f4f228-3d0f-4488-9de3-cf09c83ffcf1","arxiv_id":"2501.13600","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Every residually finite hyperbolic group, and every curve graph of a finite-type surface, admits globally stable cylinders.","lead":"This mathematics paper proves that every residually finite hyperbolic group has globally stable cylinders, a sharp structural property asked about by Rips and Sela in 1995, and proves the same for curve graphs of surfaces. It also embeds curve graphs equivariantly into finite products of quasitrees, a question left open by Bestvina, Bromberg, and Fujiwara.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 6.2 is the load-bearing bridge from BBF's (QT) to strong (QT); it is only a proof outline, so Theorem 1.1 is not fully established as written.","rationale":"I read the paper in good faith. The framework is coherent: the dualisable-system construction in Sections 3–4 is carefully developed, Theorem 5.8 has a detailed proof, and the paper is honest about what is imported and what is only outlined. I found no internal contradiction and no post-hoc parameter fitting. The single most load-bearing concern is indeed the quasimedian upgrade of the Bestvina–Bromberg–Fujiwara embedding: Proposition 6.2 is presented as a proof outline, yet Theorem 1.1 depends on it entirely. My concern is not that the statement is false; it is that the paper makes its headline theorem rest on a nontrivial proposition whose proof is deferred to papers that do not explicitly contain it. The proposed concrete check—a complete proof with explicit constants—would settle whether the argument transfers. Because this matches the reader's identified weakest assumption and the reader's conditional verdict, I recommend no change to the verdict; the paper should be accepted only after the proof of Proposition 6.2 is completed.","tokens_in":23284,"tokens_out":10533,"duration_ms":99541,"concrete_test":"Write out the complete proof of Proposition 6.2 by transcribing the argument of [HP22, Prop. 3.9] and [Pet21, Prop. 3.2] into the BBF setting. In particular, give explicit constants showing that for a geodesic γ in a finite-index factor-preserving subgroup H and any BBF factor T_i, the projected path f_iγ decomposes into subintervals whose projected images have uniformly bounded backtracking, and verify that every hypothesis used in that argument holds for the BBF quasitrees of arbitrary residually finite hyperbolic groups. If any step relies on data specific to mapping class groups or hierarchically hyperbolic groups, Proposition 6.2 fails as written; if all steps transfer, the proof gap is closed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central theorem, Theorem 5.8, assumes strong (QT), and Theorem 1.1 obtains strong (QT) for residually finite hyperbolic groups from Theorem 6.1 plus Proposition 6.2. But Proposition 6.2 is not proved in the paper: the text explicitly says that the quasimedian upgrade of BBF orbit maps 'has not been explicitly stated in the literature' and gives only a proof outline, deferring to [HP22, Prop. 3.9] and [Pet21, Prop. 3.2]. The load-bearing step is the assertion that, for a geodesic γ in a finite-index factor-preserving subgroup H, each component orbit map f_i sends γ to an unparametrised quasigeodesic in the BBF quasitree T_i. This is exactly what prevents backtracking and makes the embedding quasimedian; it is not a formality. Lemma 2.4 cannot supply it, because the product of m>1 quasitrees is not hyperbolic. The outline's final sentence—'By a careful choice of which subintervals to consider, one can then realize f_iγ as a concatenation of unparametrised quasigeodesics that do not fellow-travel one another'—states the desired conclusion rather than proving it. Since every advertised application to residually finite hyperbolic groups passes through this proposition, the central new result is conditional on a published-but-not-quoted argument. The curve-graph application has a second compressed input (Proposition 7.3), but Proposition 6.2 is the more fundamental bottleneck for Theorem 1.1.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims to prove that every residually finite hyperbolic group and every curve graph of a finite-type surface (with mapping class group action) admits globally stable cylinders, resolving the Rips–Sela question in these cases and giving the first equivariant quasiisometric embeddings of curve graphs into finite products of quasitrees. The architecture is: define strong (QT) (equivariant, quasimedian quasiisometric embedding into a finite product of quasitrees), thicken such a space using a generalised Sageev wall-space construction to obtain a median, Helly, hyperbolic space with controlled wall combinatorics, and then construct globally stable cylinders directly in that thickening (Theorem 5.8). The applications then reduce to establishing strong (QT): for residually finite hyperbolic groups via Bestvina–Bromberg–Fujiwara plus a quasimedian upgrade (Proposition 6.2), and for curve graphs via a similar upgrade plus a comparison between the thickened product and the curve graph (Proposition 7.3).","tokens_in":23545,"tokens_out":2726,"duration_ms":24832,"significance":"If the main claims hold, the paper resolves a long-standing question of Rips–Sela for a large class of hyperbolic groups and introduces a new structural property, strong (QT), that cleanly implies globally stable cylinders. The main cylinder construction (Theorem 5.8) is original and carefully organised: it develops a wall-space analogue of the Lazarovich–Sageev cube-complex argument without needing cubulation, which is essential because residually finite hyperbolic groups with Property (T) are not cubulated. The paper also gives a unified treatment that yields cylinders for curve graphs and, conditionally, for other hierarchically hyperbolic groups. The proofs are largely self-contained for the core cylinder theorem, and the conceptual framework is likely to be influential. However, as written, the two load-bearing bridges to the advertised applications are not fully demonstrated.","major_comments":[{"comment":"Proposition 6.2 is the only step that upgrades the Bestvina–Bromberg–Fujiwara property (QT) to strong (QT) for residually finite hyperbolic groups, and Theorem 6.3 and Theorem 1.1 depend on it. The proof is explicitly only an outline, and the text states that the quasimedian upgrade 'has not been explicitly stated in the literature' and refers to [HP22, Prop. 3.9] and [Pet21, Prop. 3.2] for details. The critical assertion—that for a geodesic γ in the finite-index factor-preserving subgroup H, each component orbit map f_i sends γ to an unparametrised quasigeodesic in the quasitree T_i—is not proved here, and the final sentence of the outline ('By a careful choice of which subintervals to consider, one can then realize f_i γ as a concatenation...') states exactly the desired conclusion. Lemma 2.4 cannot replace this step because the product of m>1 quasitrees is not hyperbolic. As written, Theorem 1.1 is conditional on an unstated published argument, so this is a load-bearing gap that must be filled by a complete proof or a precise quotation of a result that implies the quasimedian property of orbit maps.","section":"§6, Proposition 6.2"},{"comment":"Proposition 7.3 is the second load-bearing bridge: it asserts that the thickened dual space (S, d_{E^K}) is quasiisometric to the curve graph CΣ, which is necessary for Theorem 7.6 and consequently for Theorems 1.2 and 1.3. The proof is a compressed sketch that relies on several external results ([DZ22, Lem. 7.10], [MM99, Thm 1.3]) and on informal claims about hulls, coarse gates, and squares in product regions. In particular, the step producing chains c_i separating H_i from H_{i+1} and the subsequent 'almost all elements cross' arguments are not given in enough detail to verify the uniform constants or the existence of the K–chains. Since the entire curve-graph application passes through this proposition, the manuscript should either provide a complete proof or isolate a precise statement with full hypotheses and refer to a proof that is publicly available in the same form.","section":"§7, Proposition 7.3"},{"comment":"The central cylinder theorem is well structured, but one point in the proof deserves clarification: in the final paragraph, the argument distinguishes the case d_C(x, q) ≥ r − mL − L and then handles the complementary case using Corollary 5.7, but the transition to 'at most m+1 balls' relies on bounding d_C(q, g_{[x,y]}(q)) by 2 in the second case. The text says 'all but one elements of c cross every element of b, so the fact that c ∈ C implies that |c| ≤ 2'; this is plausible given Lemma 4.4, but the deduction that the remaining elements of c cross b is stated without a full justification of why they cannot separate x from y. Since this is the key estimate controlling the number of exceptional balls, a short extra explanation would remove a potential source of error.","section":"§5, Theorem 5.8"}],"minor_comments":[{"comment":"The section heading contains a typo: 'Cur ve graphs' should be 'Curve graphs'.","section":"§7, heading"},{"comment":"The proof outline does not specify which finite-index subgroup H is used or how the constants in the quasimedian estimate depend on the index; this should be made explicit when the proof is completed.","section":"§6, Proposition 6.2"},{"comment":"In the displayed stability condition, the notation '∖' is used for set difference; it would be clearer to use the standard '\\setminus' symbol consistently.","section":"§2, Definition 2.2"},{"comment":"In the proof of Proposition 2.5, the notation 'd_X(x_1, φ(q)) > ⟨y_1, z_1⟩_{x_1}' is meaningful but could be confused with an open interval; consider using '\\langle y_1, z_1\\rangle_{x_1}' with a word of explanation, as is done elsewhere.","section":"§2, Proposition 2.5"},{"comment":"The proof that π_iγ is an unparametrised rough geodesic in T_{D_i} cites a fact that is stated later ([Pet22, Lem. 2.16]); adding an explicit parenthetical reference at this point would improve readability.","section":"§4, Lemma 4.2"}],"recommendation":"major_revision","confidential_remarks":"The two load-bearing gaps (Proposition 6.2 and Proposition 7.3) both concern arguments that the authors appear to have in related work, but they are not presented in a form that allows a referee to verify them from the manuscript alone. The editor may wish to ask the authors to provide a fully detailed proof of Proposition 6.2 in particular, since Theorem 1.1 is the headline result and the current text explicitly says the needed statement is not in the literature. The overall framework is promising and the main cylinder construction appears sound, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe headline: this is a substantial paper. It answers the 1995 Rips–Sela question for all residually finite hyperbolic groups and settles the BBF open question on equivariant embeddings of curve graphs into products of quasitrees. The main engine, Theorem 5.8, is a genuine architectural step: strong (QT) plus a generalized Sageev thickening gives globally stable cylinders. The colored-walls surrogate for cubical dimension is clever, and the proof of Theorem 5.8 is earnest and detailed—Lemmas 4.4, 5.4, and Proposition 5.6 do real work.\n\nWhere the paper is soft: exactly where the stress-test puts it. Proposition 6.2, which upgrades the BBF orbit maps to quasimedian embeddings, is load-bearing for Theorem 1.1, and it is not proved. The text says the result hasn't been stated in the literature, gives an outline, and refers to [HP22, Prop. 3.9] and [Pet21, Prop. 3.2]. The crucial step—geodesics in H map to unparametrised quasigeodesics in each quasitree, i.e., no long back-tracking—is asserted in the outline rather than demonstrated. Lemma 2.4 can't supply it in a product of more than one quasitree. So as written, the residually finite hyperbolic group theorem is conditional on a published-but-not-quoted argument. I don't think it's circular—the prior work is background, not the target result—but the referee will need to verify that step.\n\nThe curve graph application has a second compressed input, Proposition 7.3, which imports heavy machinery (hulls, projection complexes, [MM99]'s curve stabilizer products) to show S is quasiisometric to the curve graph. That's a lot to ask of a few paragraphs. Still, it is a separate application, and the main conditional bottleneck for Theorem 1.1 is Proposition 6.2.\n\nOverall: no internal contradiction, no parameter fitting. The results are new and significant. The paper is not fully self-contained, and the two compressed arguments need to be expanded or quoted precisely before the central claims are fully established. This deserves a serious referee—not a desk reject. The referee should be asked to check Prop 6.2 and Prop 7.3 rather than the whole architecture.","headline":"Strong new architecture, but Theorem 1.1 rests on a load-bearing proposition that is only sketched, so the paper is conditional as written yet well worth a serious referee.","tokens_in":24156,"tokens_out":2221,"would_cite":true,"duration_ms":19479,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20F65","20F67","57K20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every residually finite hyperbolic group admits globally stable cylinders, and curve graphs of finite-type surfaces do too, via a wall-dual upgrade into median spaces.","keywords":["globally stable cylinders","hyperbolic groups","curve graphs","quasitrees","quasimedian maps","median algebras","dualisable systems","mapping class groups"],"falsifier":"Take a residually finite hyperbolic group, build its quasitree product factors, and check whether every geodesic in the group projects to an unparametrised quasigeodesic in each factor; any geodesic whose projection backtracks by an amount growing with length would falsify Proposition 6.2 and remove the input needed for Theorem 1.1.","tokens_in":23005,"feed_emoji":"📐","tokens_out":18013,"duration_ms":148547,"temperature":0.7,"pith_summary":"This paper answers a 1995 question in geometric group theory by proving that every residually finite hyperbolic group has globally stable cylinders: a choice of tube around each geodesic that is reversible and agrees for all triples except inside a bounded number of small balls. The same conclusion is proved for the curve graph of every finite-type surface, under the action of the mapping class group. The route is to first embed the space equivariantly and quasimedianly into a finite product of quasitrees, and then apply a generalised wall-dual construction that replaces the coarse hyperbolic space by a finer median space with exact gates. On that finer space, cylinders built from 'distant' walls are shown to be globally stable, and stability transfers back along the quasiisometry. The result gives hyperbolic groups a globally stable bicombing and simplifies the algorithmic solution of equations over such groups.","feed_headline":"All residually finite hyperbolic groups get stable cylinders","feed_subtitle":"A wall-dual construction also gives curve graphs equivariant quasiisometric embeddings into products of quasitrees","key_machinery":"The load-bearing machinery is a dualisable system of chains of walls. Starting from a product of quasitrees, walls are induced by balls in each factor; a chain is a sequence of walls each separating its predecessor from its successor, and the system keeps only chains whose crossing patterns avoid large square grids. The dual space, the set of ultrafilters on walls at finite distance from the original space, is a roughly geodesic hyperbolic median space whose balls and halfspaces are gated, with a gate map for every gated subset. The bound on grid size supplies a surrogate for finite dimensionality, replacing the hyperplane-dimension argument used for cube complexes. A wall is called distant from $x$ and $y$ when every monochromatic chain separating $x$ from $y$ and crossing it is short; the cylinder between $x$ and $y$ is a neighbourhood of the intersection of all halfspaces of distant walls that contain both points.","core_discovery":"At the centre of the paper is Theorem 5.8: if a roughly geodesic hyperbolic space $X$ with a group action has finite strong quasitree rank $m$ — meaning a $G$-equivariant, quasimedian, quasiisometric embedding into a product of $m$ quasitrees — then the pair $(G,X)$ admits globally $(2m+1,R)$-stable cylinders for some constant $R$. The cylinders are defined on the dual wall-space $S^C$, which is quasiisometric to $X$ and carries an exact median structure; the proof shows that the interval of distant walls between two points is a uniform quasiline, and that the symmetric difference of two such intervals near their common basepoint is covered by at most $m+1$ uniformly bounded balls. Applying this to residually finite hyperbolic groups and to curve graphs yields the headline theorems, and, as a byproduct, an equivariant quasimedian quasiisometric embedding of each curve graph into a finite product of quasitrees.","pith_inferences":["The paper's architecture exposes one missing input for a full solution of the original 1995 question: a quasimedian version of property (QT) for all hyperbolic groups. The authors pose this as a question, but the conditional conclusion is direct from Theorem 5.8.","Because the stability bound is $2m+1$ in terms of quasitree rank $m$, sharper rank bounds would directly improve cylinder quality; for rank-one spaces the construction suggests at most three exceptional balls per triple, a quantitative prediction that could be tested on concrete groups.","The wall-dual thickening does not use residual finiteness, so any hyperbolic space with a quasimedian quasitree model is a candidate for stable cylinders; deformation spaces of surfaces and other hierarchically hyperbolic spaces would be natural next cases once equivariant quasimedian embeddings are known.","The exact gates and finite-intersection properties produced by the wall-dual construction may be reusable beyond cylinders, for example to build canonical projections or exact medians in hyperbolic spaces."],"forward_implications":["Every residually finite hyperbolic group admits globally stable cylinders, hence a globally stable bicombing and coarsely canonical representatives for group elements.","The curve graph of every finite-type surface admits globally stable cylinders that are equivariant under the mapping class group.","Curve graphs admit equivariant, quasimedian, quasiisometric embeddings into finite products of quasitrees, resolving the open point left by earlier non-equivariant embeddings.","Any hierarchically hyperbolic group with strong (QT) has its largest hyperbolic space with strong (QT) and with globally stable cylinders; residually finite Artin groups of large and hyperbolic type are one concrete class.","If all hyperbolic groups were shown to have strong (QT), the same theorem would give them globally stable cylinders without further work."],"supporting_citations":[{"why":"Defines stable cylinders, proves the finite-set version for torsionfree hyperbolic groups, and poses the global question this paper answers.","marker":"[RS95]"},{"why":"Proves globally stable cylinders for hyperbolic cube complexes and supplies the stability-transfer proposition adapted here.","marker":"[LS24]"},{"why":"Supplies equivariant quasiisometric embeddings into finite products of quasitrees for residually finite hyperbolic groups and mapping class groups.","marker":"[BBF21]"},{"why":"Provides the quasimedian results, including Lemma 2.4 and the upgrades used as Propositions 6.2 and 7.1, that turn property (QT) into strong (QT).","marker":"[Pet21]"},{"why":"Develops dualisable systems and dual spaces, the wall-dual construction that yields the fine median space $S^C$.","marker":"[PZ24]"},{"why":"Establishes the hyperbolic wallspace with finite-intersection, gate, and convexity properties used in the thickening step.","marker":"[PSZ24]"},{"why":"Proves that the curve graph is obtained from the mapping class group by coning off curve stabilisers, a step used in Proposition 7.3.","marker":"[MM99]"}],"fun_headline_variants":["Stable cylinders for all residually finite hyperbolic groups","1995 Rips–Sela question solved for residually finite hyperbolic groups","Stable cylinders for curve graphs and residually finite hyperbolic groups","Wall-space duality yields stable cylinders and quasitree embeddings","Stable cylinders for all residually finite hyperbolic groups and curve graphs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction needs the known embedding of the group or space into a finite product of quasitrees to be quasimedian; for residually finite hyperbolic groups this upgrade is only sketched, and for curve graphs a related quasiisometry is compressed, so if either gap is real the cylinder proof collapses.","fun_headline_variants_meta":{"raw":{"variants":["Stable cylinders for all residually finite hyperbolic groups","1995 Rips–Sela question solved for residually finite hyperbolic groups","Stable cylinders for curve graphs and residually finite hyperbolic groups","Wall-space duality yields stable cylinders and quasitree embeddings","Stable cylinders for all residually finite hyperbolic groups and curve graphs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001682,"raw_usage":{"total_tokens":6607,"prompt_tokens":820,"completion_tokens":5787,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":436,"completion_tokens_details":{"reasoning_tokens":5700}},"tokens_in":436,"tokens_out":5787,"duration_ms":36474,"temperature":1.0,"reasoning_tokens":5700,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T15:47:26.103683+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a residually finite hyperbolic group, build its quasitree product factors, and check whether every geodesic in the group projects to an unparametrised quasigeodesic in each factor; any geodesic whose projection backtracks by an amount growing with length would falsify Proposition 6.2 and remove the input needed for Theorem 1.1.","supporting_citations":[],"review_version":1}