{"id":"c17307eb-5104-46b2-b7f6-92a058fe8b84","arxiv_id":"2501.02238","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Quasi-retracts give a new coarse-splitting criterion: a subgroup is a quasi-retract exactly when the ambient group is strictly quasi-isomorphic to a direct product over the quotient.","lead":"This paper introduces quasi-retracts, approximate versions of retractions in which the retracting map is only required to behave like a homomorphism up to bounded error. It proves a coarse analogue of the classical splitting theorem for group extensions and shows several geometric properties are preserved under these approximations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 4.7(2) relies on an unproved one-line implication: finiteness of C(<f^k>, s(G/H)) is asserted to force s(G/H)⊂E(f^k). This is load-bearing for the classification of normal quasi-retracts and needs a geometric axis argument; Theorem 4.5 itself appears sound.","rationale":"I read the whole paper with Theorem 4.5 as the central object. The three equivalences are supported by explicit constructions; the finite-index and defect-set bookkeeping in Lemma 4.3 and in the proof of (2)=>(3) is consistent, and the strict inverse phi^{-1}=phi' is verified. The applications to (QF A), (QT'), (PH') and hyperbolic structures all reduce to Theorem 4.5 plus standard quasi-action results, so I do not see a challenge to the central claim. The reader's conditional verdict is therefore appropriate. The single soft spot I find is exactly the one flagged by the reader: Proposition 4.7(2) has an unproved geometric implication. I also note a secondary issue in Proposition 4.8: the claimed unique representation g=f c_1^{n_1}...c_m^{n_m} for a merely quasi-homomorphic section s is not literally true when the finite-by-Z^m extension is non-split; however the proof is repairable by using the quotient projection and the finiteness of F and D(s) to compare h with c_1^{φ_1(h)}...c_m^{φ_m(h)}. This reinforces CONDITIONAL but does not affect Theorem 4.5. Since my read agrees with the reader's weakest assumption and recommended verdict, I set verdict_should_be=UNCHANGED.","tokens_in":30808,"tokens_out":19986,"duration_ms":203705,"concrete_test":"Isolate the assertion as a lemma: let G be a non-elementary torsion-free hyperbolic group, f hyperbolic, F finite, and T a set with [f^{kn}, t]∈F for all n∈Z and t∈T. Prove or disprove that T⊂E(f^k). A concrete way to test it is to use the axis characterisation: choose n large enough that the word length of every element of F is smaller than the translation length of f^{kn}; then t f^{kn} t^{-1} is a bounded perturbation of f^{kn}. By the fellow-traveller property for axes, t(axis f^k) is at bounded Hausdorff distance from axis f^k. Check whether this forces t to fix the two endpoints setwise; if yes, the standard cyclic commensurator theorem for torsion-free hyperbolic groups gives t∈E(f^k). If the argument requires an orientation-preservation step, verify that large n rules out swapping the endpoints.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central equivalence Theorem 4.5 survives my reading: Lemma 4.3's transversal construction and the proof of (2)=>(3) both check out, including the outer-conjugation step via Lemma 4.2. The genuinely load-bearing gap is in Proposition 4.7(2). After choosing k with g^k∈H for all g, the paper has f^k∈H, so finiteness of C(H,s(G/H)) gives finiteness of C(<f^k>,s(G/H)). The next sentence, 'This implies that s(G/H)⊂E(f^k)', is doing a lot of work. It is not a formal consequence of the finiteness of a commutator set; it uses hyperbolicity in an essential way. For each t∈s(G/H) and each n, finiteness gives t f^{kn} t^{-1} = c_n f^{kn} with c_n in a fixed finite set. One must then prove, from the fellow-traveller property for the axes of f^{kn} and t f^{kn} t^{-1}, that t fixes the endpoints of f^k, and that in a torsion-free hyperbolic group this implies t f^{km} t^{-1}=f^{km} for some m. The paper supplies neither a proof nor a reference for this implication. The step is essential: it converts finite index of H into s(G/H)⊂⋂_f E(f^k)={1}, which is exactly the claimed triviality of normal quasi-retracts in non-elementary torsion-free hyperbolic groups. If this implication is false, Proposition 4.7(2) and the rigidity conclusion are unsupported. I regard it as likely true but not established as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces the notion of a quasi-retract of a group, a subgroup H of G for which there exists a quasi-homomorphism r:G→H that is the identity on H (up to bounded error). The main structural result is Theorem 4.5 (Theorem 1.4), which gives three equivalent conditions for a short exact sequence 1→H→G→Q→1 to be left quasi-split: (1) H is a quasi-retract of G; (2) there is a normalized quasi-homomorphic section s:Q→G with H almost commuting with s(Q); (3) there is a strict quasi-isomorphism G→H×Q making the diagram commute. The paper then applies this theorem to classify quasi-retracts in free groups and normal quasi-retracts in hyperbolic groups, to show that several geometric properties (QFA, QT', PH', and PH under an FZ hypothesis) are stable under left quasi-split extensions, and to show that quasi-isomorphic groups have isomorphic hyperbolic structures. It also develops a theory of induced quasi-actions from quasi-homomorphisms.","tokens_in":31186,"tokens_out":24461,"duration_ms":214955,"significance":"Theorem 4.5 is a clean and useful coarse analogue of the classical split extension theorem; if it holds, it provides a unified way to transfer properties between a group and its quasi-retracts. The applications to property inheritance and hyperbolic structures are novel and likely to be cited. The paper is carefully structured, and the proof of the central theorem is essentially direct and convincing. The main caveat is that a few geometric facts in the applications are asserted without proof, most notably in Proposition 4.7(2) and Lemma 5.8.","major_comments":[{"comment":"The implication from finiteness of C(<f^k>, s(G/H)) to s(G/H) ⊂ E(f^k) is not justified in the text. It is, however, true: for each t∈s(G/H), the finiteness of { [f^{km}, t] : m∈Z } yields m≠n with [f^{km},t]=[f^{kn},t], which implies t f^{k|m-n|} t^{-1}=f^{k|m-n|}, so t∈E(f^k). I recommend adding this one-line argument (or a reference) because the step is load-bearing for the rigidity conclusion.","section":"Proposition 4.7(2)"},{"comment":"The proof assumes the existence of a hyperbolic element f with stable length τ(f)>M in a cobounded focal or general type action. This is standard (e.g., from the dynamics of hyperbolic isometries on the boundary), but it is not stated or referenced. Since the inequality τ(f)−M ≤ τ(f_H)+τ(f_T) is used to ensure that one of f_H, f_T is hyperbolic, a brief justification of the unboundedness of translation lengths in this situation should be supplied.","section":"Lemma 5.8"}],"minor_comments":[{"comment":"There is a typo: 'it is it is' should be 'it is'.","section":"Corollary 1.6"},{"comment":"Lemma 2.1(2) gives a D^2 estimate, but several later arguments use the stronger estimate φ(x^{-1}) ≈_{D^{-1}} φ(x)^{-1} for normalized quasi-homomorphisms. Please clarify that this follows from φ(1)=1.","section":"Lemma 2.1"},{"comment":"Proposition 5.20 cites [20, Proposition 4.4]; the result about quasi-conjugacy to graph actions is likely in [21] (Manning's QFA paper). Please check the reference.","section":"Proposition 5.20"},{"comment":"In Lemma 5.7, the proof would be easier to follow if the two cases were separated more explicitly, especially the passage from finiteness of C(H,T) to T⊂E(f).","section":"Lemma 5.7"}],"recommendation":"major_revision","confidential_remarks":"The paper relies on the author's own preprint [30] for several key results (Lemma 5.22, [30, Lemma 7.1], and the properties (PH)/(QT) equivalences). The editor may want to ensure that [30] is available and in a stable form before publication. The present manuscript is largely self-contained for its central theorem."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Central Theorem 4.5 is the real contribution, and it looks right: the equivalence between left quasi-split, almost-commuting quasi-section, and strict quasi-isomorphism to a direct product is a clean coarse analogue of the classical split-extension theorem, and the proof via the transversal construction checks out. The notion of a quasi-retract with finite defect set, distinguished from HS-quasi-retractions, is genuinely new, and it earns its keep: it gives a unified explanation for why bounded Euler classes, (QFA), (QT'), (PH'), and hyperbolic structure posets behave well under these extensions. The paper is honest about relying on the author's own preprint [30] for some applications; that is a real dependency but not a flaw, since those results are external.\n\nThe soft spot is Proposition 4.7(2), exactly where the stress-test lands. After choosing k with g^k in H, the one-line implication 'finiteness of C(<f^k>, s(G/H)) implies s(G/H) ⊂ E(f^k)' is doing the entire job of converting finite index into triviality of normal quasi-retracts in non-elementary torsion-free hyperbolic groups. It is not a formal consequence of finite commutator set; it needs a geometric axis/fellow-traveller argument for hyperbolic elements. The paper gives neither a proof nor a reference. I think the statement is likely true and fixable with a standard hyperbolic argument, but as written it is unsupported.\n\nThere are also minor defect-set bookkeeping issues: Lemma 2.1(2) gives a D^2 approximation for inverses, while some uses in Lemma 4.3 and Theorem 4.5 assume a D^{-1} approximation. These are fixable and do not threaten the main equivalence, but they should be cleaned up.\n\nWho is this for: geometric group theorists working on coarse geometry of group extensions, bounded cohomology, and (QT)/(PH)-type properties. The central theorem and framework deserve referee time. The gap in 4.7(2) is localized and probably repairable, so I would send it to a serious referee with a request to scrutinize that step.","headline":"A genuinely new quasi-retract framework with a solid central equivalence, but a load-bearing unproved step in the hyperbolic rigidity application needs real proof.","tokens_in":31702,"tokens_out":1671,"would_cite":true,"duration_ms":16537,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20F65","20F67"],"pacs":[],"model":"deepseek-v4-flash","headline":"A subgroup of a group is a quasi-retract exactly when the ambient group has a normalized quasi-homomorphic section that almost commutes with it, equivalently when the group is strictly quasi-isomorphic to the direct product.","keywords":["quasi-homomorphism","quasi-retract","quasi-split exact sequence","quasi-isomorphism","hyperbolic group","quasi-action","hyperbolic structure","bounded Euler class"],"falsifier":"Take a non-elementary torsion-free hyperbolic group and a hyperbolic element $f$. Search for a finite set $T$ with $C(\\langle f^k\\rangle,T)$ finite yet some $t\\in T$ does not conjugate any positive power of $f$ to another power of $f$; finding one would disprove the unproved step behind Proposition 4.7(2). A second test of Theorem 4.5 itself: exhibit a short exact sequence satisfying condition (2) but admitting no strict quasi-isomorphism to $H\\times Q$; such an example would refute the claimed equivalence.","tokens_in":30603,"feed_emoji":"🔗","tokens_out":14253,"duration_ms":119454,"temperature":0.7,"pith_summary":"This paper studies quasi-retracts: subgroups $H$ of a group $G$ for which there exists a quasi-homomorphism $r:G\\to H$ that is the identity on $H$ up to bounded error. Its central result, Theorem 4.5, is a coarse analogue of the classical split-extension theorem: for a short exact sequence $1\\to H\\to G\\to Q\\to 1$, the subgroup $H$ is a quasi-retract if and only if there is a normalized section $s:Q\\to G$ that is a quasi-homomorphism and whose image almost commutes with $H$, meaning the commutator set $C(H,s(Q))$ is finite, and if and only if $G$ is strictly quasi-isomorphic to the direct product $H\\times Q$ through a map commuting with the exact sequences. This matters because the criterion turns a coarse algebraic condition into a usable splitting tool: the paper derives from it that normal quasi-retracts inherit cobounded actions on hyperbolic spaces, that properties $(QFA)$, $(QT')$, and $(PH')$ are stable under left quasi-split extensions, and that quasi-isomorphic groups have isomorphic posets of hyperbolic structures. A sympathetic reading is that the paper establishes quasi-retracts as the right coarse analogue of retracts and shows the equivalence holds constructively.","feed_headline":"Quasi-retracts exist exactly when a coarse section almost commutes","feed_subtitle":"A bounded-error section criterion makes quasi-retracts a tool for proving stability of group actions and rigidity","key_machinery":"The load-bearing object is the defect set $D(\\varphi)=\\{\\varphi(y)^{-1}\\varphi(x)^{-1}\\varphi(xy)\\}$ of a quasi-homomorphism, which is finite by definition, together with the almost-commutation set $C(H,s(Q))=\\{h s(q) h^{-1}s(q)^{-1}\\}$. Theorem 4.5 runs on converting these two finiteness conditions into each other: a finite defect set for the section plus a finite commutator set makes the projection map $g_Hg_T\\mapsto g_H$ a quasi-homomorphism, and a quasi-retraction's finite defect set forces the existence of a transversal $T$ with $C(H,T)$ finite. The same conversion produces the strict quasi-isomorphism to $H\\times Q$ and its quasi-inverse $(h,q)\\mapsto h s(q)$.","core_discovery":"The central discovery is that being a quasi-retract is not merely a property of a subgroup but a structural statement about the ambient group. For any short exact sequence $1\\to H\\to G\\to Q\\to 1$, Theorem 4.5 shows the following are equivalent: $H$ is a quasi-retract of $G$; there is a normalized section $s:Q\\to G$ that is a quasi-homomorphism and whose image almost commutes with $H$, meaning the commutator set $C(H,s(Q))$ is finite; and there is a strict quasi-isomorphism $\\varphi:G\\to H\\times Q$ (a quasi-isomorphism whose quasi-inverse is the set-theoretic inverse) making the two short exact sequences commute. The equivalence is proved constructively: the projection $g=g_Hg_T\\mapsto g_H$ built from the section is the quasi-retraction, and conversely any quasi-retraction produces such a section. From this coarse splitting criterion the paper derives the stability results for group actions and hyperbolic structures.","pith_inferences":["Because Lemma 2.10 identifies quasi-isomorphisms with quasi-isometries for finitely generated groups, Theorem 4.5 gives an explicit coarse model $G\\approx H\\times Q$; one could use it to compute coarse invariants such as asymptotic dimension from the two factors.","The unproved step in Proposition 4.7(2) suggests a standalone lemma about finite sets almost commuting with powers of hyperbolic elements in non-elementary torsion-free hyperbolic groups; proving or disproving that lemma would settle the classification of normal quasi-retracts of hyperbolic groups.","The induced-quasi-action construction via coarsely surjective quasi-homomorphisms is not tied to hyperbolicity, so the stability arguments for $(PH')$ and $(QT')$ may extend to other classes of actions on products of hyperbolic spaces or quasi-trees."],"forward_implications":["If $H$ is a normal quasi-retract of $G$, every cobounded action of $G$ on a hyperbolic space restricts to an action of $H$ that is either elliptic or cobounded; in particular property $(PH')$ passes to normal quasi-retracts.","For left quasi-split extensions of finitely generated groups, $G$ has property $(QFA)$ if and only if both $H$ and $Q$ do, and the same two-way stability holds for $(PH')$ and $(QT')$.","A coarsely surjective quasi-homomorphism $G\\to H$ embeds the poset of hyperbolic structures of $H$ into that of $G$, and quasi-isomorphic groups have isomorphic posets of hyperbolic structures.","A finite-by-$\\mathbb{Z}^m$ hyperbolically embedded subgroup is always a quasi-retract, so quasimorphisms on it extend to the ambient group; in $F_2$, nontrivial quasi-retracts are exactly cyclic subgroups."],"supporting_citations":[{"why":"It supplies the foundational theory of quasi-homomorphisms with noncommutative targets and the equivalence between right quasi-split central extensions and bounded Euler classes.","marker":"[12]"},{"why":"It provides the extension theorem for homomorphisms on hyperbolically embedded subgroups, used to build the quasi-retraction in Proposition 4.8.","marker":"[17]"},{"why":"It defines hyperbolically embedded subgroups and their basic properties, used to identify which subgroups admit quasi-retractions.","marker":"[9]"},{"why":"It defines hyperbolic structures and the trichotomy of actions on hyperbolic spaces, used in the inheritance and rigidity results of Section 5.","marker":"[1]"},{"why":"It introduces quasi-actions and property (QFA), and supplies the quasi-conjugacy machinery used to induce actions from quasi-homomorphisms.","marker":"[21]"},{"why":"It establishes properties (PH') and (QT') and their stability under left quasi-split central extensions, which Theorem 1.10 generalizes.","marker":"[30]"},{"why":"It defines property (QT) and gives examples of groups admitting proper actions on finite products of quasi-trees.","marker":"[5]"},{"why":"It provides the result that quasi-actions on geodesic metric spaces are quasi-conjugate to isometric actions on graphs, used to turn induced quasi-actions into genuine actions.","marker":"[20]"}],"fun_headline_variants":["Quasi-retracts exist when a section almost commutes","Coarse section criterion reveals group rigidity","Almost commuting sections define quasi-retracts","Group quasi-retracts: a bounded-error section test","Quasi-split extensions yield stable actions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing step the proof of Proposition 4.7(2) does not justify is this: in a non-elementary torsion-free hyperbolic group, if a finite set almost commutes with the cyclic subgroup generated by $f^k$, then every element of that set must conjugate some positive power of $f$ to itself. If that step fails, the proof that a normal quasi-retract of such a group is trivial collapses.","fun_headline_variants_meta":{"raw":{"variants":["Quasi-retracts exist when a section almost commutes","Coarse section criterion reveals group rigidity","Almost commuting sections define quasi-retracts","Group quasi-retracts: a bounded-error section test","Quasi-split extensions yield stable actions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000164,"raw_usage":{"total_tokens":1198,"prompt_tokens":851,"completion_tokens":347,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":467,"completion_tokens_details":{"reasoning_tokens":278}},"tokens_in":467,"tokens_out":347,"duration_ms":4350,"temperature":1.0,"reasoning_tokens":278,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:14:05.778024+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a non-elementary torsion-free hyperbolic group and a hyperbolic element $f$. Search for a finite set $T$ with $C(\\langle f^k\\rangle,T)$ finite yet some $t\\in T$ does not conjugate any positive power of $f$ to another power of $f$; finding one would disprove the unproved step behind Proposition 4.7(2). A second test of Theorem 4.5 itself: exhibit a short exact sequence satisfying condition (2) but admitting no strict quasi-isomorphism to $H\\times Q$; such an example would refute the claimed equivalence.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the foundational theory of quasi-homomorphisms with noncommutative targets and the equivalence between right quasi-split central extensions and bounded Euler classes."},{"cited_title":"245, American Mathematical Society, 2017","cited_arxiv_id":null,"evidence_quote":"It defines hyperbolically embedded subgroups and their basic properties, used to identify which subgroups admit quasi-retractions."},{"cited_title":"4, 1747–1835","cited_arxiv_id":null,"evidence_quote":"It defines hyperbolic structures and the trichotomy of actions on hyperbolic spaces, used in the inheritance and rigidity results of Section 5."},{"cited_title":"1, 84–108","cited_arxiv_id":null,"evidence_quote":"It introduces quasi-actions and property (QFA), and supplies the quasi-conjugacy machinery used to induce actions from quasi-homomorphisms."},{"cited_title":"Central extensions and proper actions on products of hyperbolic spaces","cited_arxiv_id":"2506.04856","evidence_quote":"It establishes properties (PH') and (QT') and their stability under left quasi-split central extensions, which Theorem 1.10 generalizes."},{"cited_title":"Bestvina, K","cited_arxiv_id":null,"evidence_quote":"It defines property (QT) and gives examples of groups admitting proper actions on finite products of quasi-trees."},{"cited_title":"MR 2174263","cited_arxiv_id":null,"evidence_quote":"It provides the result that quasi-actions on geodesic metric spaces are quasi-conjugate to isometric actions on graphs, used to turn induced quasi-actions into genuine actions."}],"review_version":1}