{"id":"b83b9239-e03e-4002-9e1c-c318aa12f192","arxiv_id":"2510.04350","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Surface-side random measures pushed through the Cannon–Thurston map are mutually singular with every natural 3-manifold measure on the boundary sphere of a fibered hyperbolic 3-manifold, with exponential effective rates.","lead":"Randomness attached to the surface that builds a fibered 3-manifold, and randomness attached to the 3-manifold itself, never mix on the boundary sphere: pushed there by the Cannon–Thurston map, the two kinds of random points behave completely differently for almost every choice. This paper proves the mutual-singularity result and gives explicit exponential rates, using a new family of uniform quasigeodesics that straighten badly distorted surface geodesics.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 6's full-surface-measure claim needs finite-moment estimates that Definition 2 does not provide; log-k rescaling in Prop 52 is secondary.","rationale":"The reader's CONDITIONAL verdict is appropriate and I would not change it. The strongest issue is not the log-k constant in Proposition 52, which is a harmless rescaling, but the mismatch between the class of 'full surface measures' in Definition 2 and the finite-moment hypotheses in the random-walk propositions used by Theorem 6. If Theorem 6 is intended only for geometric measures (or full measures with finite exponential moment), the central qualitative conclusion for Corollary 5 still goes through; if it is intended for all full measures, an additional argument is required. The paper's skeleton is sound and the issue is repairable, hence no move to REJECT. I mark partial agreement because the reader identified this concern as one of two, alongside the log-k point; I view the moment gap as the load-bearing one.","tokens_in":60603,"tokens_out":15015,"duration_ms":127788,"concrete_test":"Audit the proof of Lemma 51 line-by-line and list every use of Proposition 32/35/37 or an a.s. finite upper bound on t_n/n. If any such use occurs, construct a full nonelementary μ with infinite first moment on π1(S) (e.g., P(length>R) ~ 1/R) and check whether t_n/n → ∞ along sample paths; if so, the 'linearly often' step in Lemma 51 fails, and Theorem 6 must be restricted to surface measures with finite exponential moment (which still covers Corollary 5). If no such dependency is found, the theorem survives.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central Theorem 4 depends on Theorem 6 for every measure in Definition 2, whose 'full surface measures' include nonelementary, full probability measures with no finite exponential moment. The random-walk estimates in §2.8 are heavily stated under finite exponential moment: Proposition 32, Corollary 33, and Proposition 35 assume it, and Proposition 37 (stated for nonelementary μ) is proved using Lemma 36, which is stated only for geometric μ. Lemma 51's key step says that a geodesic has a large projection onto an axis α 'linearly often'; converting a positive frequency of good random-walk times into a positive proportion of geodesic time requires the tracked-geodesic time t_n to grow at a finite positive linear rate, or at least to have controlled gaps. Gou22 gives a lower escape rate for all nonelementary μ, but the upper linear rate for t_n and O(log n) gap bounds require finite exponential moment. For a heavy-tailed full μ, t_n/n may diverge, so the claim can fail even if hitting measure is full. Proposition 41's use of double ergodicity (Kaimanovich) is likewise a theorem normally stated with finite-first-moment hypotheses. The log k factor in Prop 52 (Eq. (1) makes adjacent fibers distance log k, not 1) only rescales the constant A and is easily repaired; it is not the main obstruction.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a measure-singularity theorem for Cannon–Thurston maps of closed hyperbolic fibered 3-manifolds. The main result, Theorem 4, states that the pushforward to S²∞ of any \"full surface measure\" on S¹∞ (including Lebesgue measure and hitting measures of nonelementary, full random walks on π₁(S)) is mutually singular with any \"3-manifold measure\" on S²∞ (Lebesgue measure or hitting measures of geometric random walks on π₁(M)). The proof is organized around two geodesic-statistics assertions: Theorem 6, that geodesics sampled by pushforwards of surface measures spend a definite positive proportion of time near the base fiber S₀, and Theorem 8, that geodesics sampled by 3-manifold measures spend asymptotically negligible time near S₀. The effective Theorem 7 gives exponential rates for geometric surface measures, using an explicit construction of quasigeodesic test paths in Section 6. The paper is well structured and the overall architecture—deducing singularity from incompatible typical behavior of geodesics—is sound.","tokens_in":60795,"tokens_out":14201,"duration_ms":141441,"significance":"If the theorem holds in the stated generality, it provides a unified geometric explanation for Tukia's Lebesgue singularity theorem and extends singularity statements to random-walk hitting measures. The construction of uniform quasigeodesic test paths with a height function and tame bottlenecks is substantial and of independent interest, and the effective exponential bounds in Theorem 7 are a genuine strengthening. The paper also correctly identifies the key geometric inputs: ladders over lamination leaves, separation properties, and quasigeodesic stability. The manuscript is careful in many places, especially in isolating the suited-lamination hypothesis and in Remark 78 concerning bounded geometry. However, the full generality of Theorem 4 currently rests on moment-free random-walk estimates that are neither stated nor proved, and the random-walk half of Theorem 8 contains a metric-scaling error; these issues need to be addressed before the central claim is fully supported.","major_comments":[{"comment":"Theorem 4 and Theorem 6 quantify over all full surface measures, which by Definition 2 include nonelementary full probability measures with no finite exponential moment. However, the random-walk estimates used in the proof assume geometric or finite-exponential-moment measures: Proposition 32, Corollary 33, Proposition 35, and Proposition 37 are stated with finite exponential moment; Lemma 36 is stated for geometric μ; and Corollary 34's upper linear-rate conclusion also uses finite exponential moment. Lemma 51's argument that good events occur \"linearly often\" needs not only ergodicity of the shift map but also control of the projected times t_n in order to convert event frequencies into a positive proportion of geodesic time. For a heavy-tailed μ, t_n/n may diverge, so the conversion can fail. No moment-free version of these estimates is supplied. This is load-bearing because Theorem 4","section":"Definition 2 and §3.3 (Lemma 51); §2.8"},{"comment":"The proof states that the distance between adjacent fibers S×{n} and S×{n+1} is equal to one in the Cannon–Thurston metric. This contradicts Eq. (1): the vertical term is (log k)² dz², so the metric distance between fibers with z-coordinates n and n+1 is log k, not 1 (unless k=e). Consequently, the implication |φ(w_n)| ≥ 2A log n ⇒ d(w_n x₀, S₀) ≥ 2A log n needs a factor 1/log k in the choice of A. The error appears repairable by rescaling A, but as written the proof contains a false metric statement in a central step of Theorem 8.","section":"§3.3, proof of Proposition 52"},{"comment":"The proof that ν-almost all geodesics are non-exceptional for full surface measures invokes double ergodicity of the boundary action, citing Kaimanovich [Kai03, Theorem 17]. Double ergodicity is typically stated under finite-first-moment or finite-exponential-moment hypotheses, and the manuscript does not verify that Definition 2's full measures satisfy them. Similarly, Lemma 51's use of \"ergodicity\" to get linearly many translates of α along a random-walk-sampled geodesic is not tied to a specific ergodic theorem; for finite-moment measures one could use Corollaries 33–35, but those are exactly the inputs missing in the no-moment case. This is part of the same gap as the first major comment, but it is important enough to flag separately because Proposition 41 is needed already to define the test paths and to rule out exceptional geodesics.","section":"§3, Proposition 41 and §3.3, Lemma 51"}],"minor_comments":[{"comment":"The arXiv title is \"Quasi-geodesics in the Cannon-Thurston metric,\" while the manuscript itself is titled \"Singularity of Cannon–Thurston maps\" and the abstract is correspondingly different. Please align the title and abstract for the intended submission.","section":"Title/abstract"},{"comment":"The height function is defined twice, in Definition 54 and again in Definition 94, with slightly different wording (one says \"regular pair,\" the other \"suited pair\"). Merge these into a single definition to avoid confusion.","section":"Definitions 54 and 94"},{"comment":"The effective bound Theorem 7 relies on Theorem 61, imported from [GH24], which states that d(ι(γ(0)), ι(γ(T))) grows linearly for Lebesgue-almost all γ. If [GH24] is a companion or unpublished manuscript, please state the exact theorem and, ideally, sketch the proof; as written, this is an external dependency for the effective half.","section":"§4, Theorem 61"},{"comment":"The phrase \"the forward and backward projections of φ'_t to eS_h × R converge to Lebesgue measure\" is vague; these are pushforwards, not projections. Please clarify the wording.","section":"§3.2, proof of Lemma 49"}],"recommendation":"major_revision","confidential_remarks":"The core geometric architecture is convincing and the paper is likely correct after fixing the two main issues: the missing finite-moment assumptions behind the full-surface-measure claims, and the log k scaling in Proposition 52. The definition of full surface measures may need to be weakened if the moment-free estimates cannot be supplied; that would still leave a substantial theorem for geometric measures. I see no grounds for rejection, but the current statements overreach the proof in a load-bearing way."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it closely. The main result is genuinely new: it upgrades Tukia's Lebesgue-vs-Lebesgue singularity, and the recent conformal variants, into a geometric criterion that covers random-walk hitting measures on both sides and gives explicit exponential rates (Theorem 7). The architecture is clean: Theorem 4 falls out of Theorems 6 and 8 by a full-measure-set singularity argument, and the surface-side proofs rest on standard geodesic-flow ergodicity. The height-function/quasi-geodesic machinery in Part II is substantial and appears to be doing real work, not just decoration. The paper also cites Tukia, Kim–Oh, and Kim–Zimmer properly; the novelty claim does not depend on ignoring them.\n\nThe soft spot is the definition of 'full surface measures' (Definition 2), where the random-walk estimates in §2.8 are stated under finite exponential moment. Propositions 32, 35, and Corollary 33 need it; Proposition 37 is stated for nonelementary μ, but its proof cites Lemma 36, which is only proved for geometric μ. Lemma 51 and Proposition 41 then invoke ergodicity and double ergodicity for the full class without reconciling this. That is load-bearing for Theorem 6, and it is not a cosmetic gap: for heavy-tailed μ with no finite exponential moment the tracked-geodesic times t_n can have uncontrolled gaps, so the 'linearly often' step can fail. The fix is either to restrict Theorem 6 to geometric surface measures or to prove the needed estimates for the full class. My guess is the qualitative result is true, and the argument can be repaired, but the statement as written overreaches.\n\nThere is a smaller, easily fixed error in Proposition 52: adjacent fibers in the Cannon–Thurston metric are log k apart by Eq. (1), not distance one. That only rescales the constant A, but the text should say so.\n\nWho is this for: people working on Cannon–Thurston maps, stationary measures on boundaries, and random walks on surface/manifold groups. They will want the effective theorem and the quasi-geodesic construction regardless of the measure-class issue. I would cite it, with a caveat until the full-measure case is sorted. It deserves a serious referee: the central idea is sound, the work is honest, and the problems are repairable. Recommend conditional acceptance subject to those fixes.","headline":"A real advance on Cannon–Thurston measure singularity, with a load-bearing gap in the 'full surface measure' statement and a small rescaling error in Prop 52.","tokens_in":61487,"tokens_out":3776,"would_cite":true,"duration_ms":34441,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["57K32","37D40","60B15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The Cannon–Thurston pushforwards of surface measures — Lebesgue and random-walk hitting measures on the circle — are mutually singular with every natural 3-manifold measure, because typical geodesics spend different amounts of time near a f","keywords":["Cannon–Thurston map","fibered hyperbolic 3-manifolds","mutual singularity of measures","hitting measures","random walks on hyperbolic groups","quasi-geodesics","measured laminations","geodesic flow"],"falsifier":"Compute the Cannon–Thurston distance between adjacent fibers S×{n} and S×{n+1} in the universal cover of the mapping torus. If it equals log k rather than 1, the estimate d(w_n x0, S0) ≥ 2A log n in Proposition 52 must be rescaled by log k, and for k < e the claimed 1/√n decay (and hence zero limiting fiber-time) may fail for the hitting-measure half of Theorem 8.","tokens_in":60332,"feed_emoji":"🕸️","tokens_out":5439,"duration_ms":44141,"temperature":0.7,"pith_summary":"This paper proves that, for a closed hyperbolic 3-manifold fibering over the circle, the Cannon–Thurston map pushes forward the natural measures on the fiber's circle at infinity (Lebesgue measure, and hitting measures of full random walks on the surface group) to measures on the sphere at infinity that are mutually singular with the natural 3-manifold measures there (Lebesgue measure, and hitting measures of geometric random walks on the 3-manifold group). The mechanism is a geodesic-statistics dichotomy: almost every geodesic sampled by a pushed-forward surface measure spends a definite positive fraction of its length near the lifted base fiber, while almost every geodesic sampled by a 3-manifold measure spends a fraction that tends to zero. For geometric surface measures the positive fraction is upgraded to an explicit exponential rate 1 − K e^{−αkR}. The proof builds explicit quasi-geodesics in the Cannon–Thurston metric, controlled by a height function measuring distance to the invariant laminations, which may be useful beyond the measure-singularity question.","feed_headline":"Pushforwards to the sphere are singular to 3-manifold measures","feed_subtitle":"Two statistics of typical geodesics, positive vs zero time near a fiber, prove the singularity and give an explicit dichotomy.","key_machinery":"The central object is the height function hθ(v) = log_k⌊log(1/d(v, Λ1+)) − log(1/θ)⌋_1 − log_k⌊log(1/d(v, Λ1−)) − log(1/θ)⌋_1 on the unit tangent bundle of the surface, where Λ1± are the extended invariant laminations. For a non-exceptional geodesic γ the test path τγ(t) = (γ(t), hθ(γ1(t))) is shown to be a uniform unparametrized quasigeodesic in the Cannon–Thurston metric with the same endpoints as ι(γ). Combined with rectangle/bottleneck estimates (optimal-height rectangles are bottlenecks for opposite quadrants), this yields the effective control of time spent near the base fiber. The non-effective half uses ergodicity of the geodesic flow and, for the 3-manifold side, a mixing estimate f","core_discovery":"Theorem 4: pushforwards of full surface measures by the Cannon–Thurston map are mutually singular with every 3-manifold measure. The proof rests on two statistical statements: Theorem 6 (a ι∗ν-typical geodesic has liminf T→∞ (1/T)|γ([0,T]) ∩ N_R(S0)| ≥ ε > 0) and Theorem 8 (a ν-typical geodesic for any 3-manifold measure has the same proportion tend to 0). The dichotomy gives an explicit description of sets witnessing singularity. Theorem 7 strengthens the surface half, for geometric surface measures, to a lower bound 1 − K e^{−αkR}.","pith_inferences":["The dichotomy likely extends to broader classes of stationary measures: if the paper's mechanism transfers, any stationary measure on the sphere whose geodesics avoid fibers would be singular to any surface-pushforward with positive fiber time.","The log k vs 1 fiber-separation issue in Proposition 52 suggests the random-walk half of Theorem 8 needs a rescaling by log k; if k