{"id":"16af4330-fdb5-4578-a8f8-495efb20ac6d","arxiv_id":"1908.08072","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A continuous flow's Bowen entropy on the generic-point set of an invariant measure is at most that measure's metric entropy, with equality when the measure is ergodic.","lead":"The authors define a Bowen entropy for subsets of a continuous flow and prove a flow version of Bowen's inequality: the entropy of any measure's generic points is bounded by the measure's own entropy, with equality for ergodic measures. This gives a route to transfer saturation and irregular-set results from discrete maps to continuous flows.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem B rests on Theorem 2.4, whose proof uses an unproved and in general false continuous-selection step; this gap is load-bearing as written.","rationale":"The reader's verdict of CONDITIONAL is appropriate. The central inequality Theorem B is plausible and the surrounding structure—Theorem A and the entropy comparison via Lemma 3.3—is coherent. The decisive weakness is exactly the one the reader identified: Theorem 2.4's proof contains an unsupported continuous-selection assertion, and the needed implication Q(Φ)⊆Q(φ1) is not established by the text. I found the same load-bearing concern, and I would keep the verdict CONDITIONAL rather than ACCEPT or REJECT: the gap is specific and potentially repairable, but as written the proof of Theorem B depends on it. I do not see a deeper fatal flaw in the main theorem independent of this step, so no change to the reader's verdict is needed.","tokens_in":11415,"tokens_out":28943,"duration_ms":314511,"concrete_test":"Re-derive Q(Φ)⊆Q(φ1) without the continuous-selection device: for x∈Q(Φ), prove directly that the time-one empirical measures ν_n=(1/n)Σ_{j=0}^{n-1}δ_{φ^j x} have a unique limit point, using only the existence of the flow averages for all continuous observables. If this direct proof can be completed, Theorem 2.4 is repaired and the main reduction stands. As an intermediate check, test the selection lemma itself on the period-1 rotation flow with C1=δ_0 and C2=δ_{1/2}: the equal-integral hypothesis is satisfied, while no continuous selection t(φ) can satisfy the fixed-point equation, showing the step in the current proof is not a valid general lemma.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem B reduces the flow statement to Bowen's discrete inequality via Theorem 2.4, which asserts Q(Φ)=Q(φ_t). The critical direction Q(Φ)⊆Q(φ1) is proved by taking two subsequential limits C1,C2 of time-one empirical measures and observing only that ∫_0^1 C1(φ∘φ^t)dt=∫_0^1 C2(φ∘φ^t)dt for every continuous φ. The text then says: 'Since C1 and C2 are continuous, there is t=t(φ) with C1(φ∘φ^t)=C2(φ∘φ^t) and the map t(φ) varies continuously with φ.' No proof of the continuous selection is supplied, and the selection claim is false in general: for the period-1 rotation flow on S^1 with C1=δ_0 and C2=δ_{1/2}, the integral equality holds for every φ but no continuous selection t(φ) can lead to the fixed-point conclusion C1=C2. Consequently, as written, the argument does not establish that a flow-quasi-regular point is time-one-quasi-regular. Since Theorem B uses Theorem 2.4 to place G_μ(Φ) inside QR(h_μ(Φ)), this is the load-bearing step; without it the proof of the main inequality lacks a bridge. The theorem may be true and repairable, but the current text does not prove it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper defines a Bowen-type topological entropy for arbitrary subsets of a compact metric space under a continuous flow and proves an Abramov-type scaling identity: h(Φ,Y)=|t|^{-1}h(φ_t,Y) for every Y⊆X and t≠0 (Theorem A). It then claims a flow analogue of Bowen's inequality: for every Φ-invariant Borel probability measure μ, h(Φ,G_μ(Φ))≤h_μ(Φ), with equality when μ is ergodic (Theorem B). The proof proceeds by comparing quasi-regular points of the flow with those of the time-one map (Theorem 2.4), then applying Bowen's discrete inequality to a set QR(h_μ(Φ)). The paper also derives consequences: if the time-one map is saturated, then the flow is saturated (Theorem C); flows with almost specification are saturated (Corollary 5.4); and under almost specification the irregular set carries full topological entropy (Theorem 5.8).","tokens_in":11734,"tokens_out":17565,"duration_ms":159740,"significance":"If the results are correct, this is a useful contribution: it provides a workable Bowen entropy for flows, a clean scaling identity, and a flow-level Bowen inequality that transfers saturation and irregular-set results from discrete dynamics to continuous flows, including geodesic flows in negative curvature. The overall strategy is attractive and not circular: it reduces the flow statement to Bowen's discrete theorem and Walters' isomorphism lemma. However, as the manuscript stands, the key bridge Theorem 2.4 rests on an unproved continuous-selection step, and several corollaries depend on unproved equivalences or nonemptiness assertions. The main theorem is plausible and likely repairable, but the written proof does not fully support all claims.","major_comments":[{"comment":"The proof that Q(Φ)⊆Q(φ_1) assumes that from the identity ∫_0^1 C_1(ϕ∘φ^t)dt = ∫_0^1 C_2(ϕ∘φ^t)dt for every continuous ϕ and the continuity of t↦C_i(ϕ∘φ^t), one can choose t(ϕ) such that C_1(ϕ∘φ^{t(ϕ)})=C_2(ϕ∘φ^{t(ϕ)}) and such that the map t(ϕ) is continuous in ϕ, after which a fixed-point argument gives C_1=C_2. No proof of the existence of such a continuous selection is supplied, and the displayed integral equality does not imply it. For example, for the rotation flow R_{t/2} on S^1, the two distinct φ_1-invariant measures C_1=(δ_0+δ_{1/2})/2 and C_2=(δ_{1/4}+δ_{3/4})/2 satisfy ∫_0^1 C_1(ϕ∘φ^t)dt=∫_0^1 C_2(ϕ∘φ^t)dt for every continuous ϕ, yet C_1≠C_2. Since this step is the only bridge from flow quasi-regularity to time-one quasi-regularity, Theorem 2.4 is not proved as written, and the subsequent inclusion G_μ(Φ)⊆QR(h_μ(Φ)) in Section 4 lacks support.","section":"Section 2, proof of Theorem 2.4 (around equation (3))"},{"comment":"The displayed line 'By (4) and Teorema A, we obtain h(Φ,QR(h_μ(Φ))) = h(φ_1,QR(h_μ(φ_1))) ≤ h_μ(φ_1) = h_μ(Φ)' uses a QR-entropy bound that is not literally Bowen's inequality (4). Inequality (4) is for the set of generic points of a single measure, whereas QR(h_μ(Φ)) is a union over the uncountable family of measures ν with h_ν(φ_1)≤h_μ(Φ). The standard estimate h(f,QR(c))≤c is true but requires a separate argument (or a precise citation); as written, the upper bound in Theorem B relies on an unproved extension of (4).","section":"Section 4, first paragraph"},{"comment":"The proof asserts 'Φ has the almost specification property if and only if φ_1 has this property' without proof or reference. This equivalence is not immediate: the flow definition allows omitting a set of times of small Lebesgue measure inside each interval, while the discrete definition for φ_1 only controls omission of integer times. The equivalence is load-bearing because it transfers the Mesón–Vericat discrete saturation theorem to flows, and it is also used in Theorem 5.8.","section":"Section 5, Corollary 5.4"},{"comment":"The opening assertion 'Observe that I_ϕ(φ_t) is not empty, for some t≠0' is nontrivial and unproved. Non-convergence of the flow averages of ϕ along a point does not automatically imply non-convergence of the Birkhoff sums of φ_t for some fixed t. Moreover, the inclusion I_ϕ(φ_t)⊆I_ϕ(Φ) used in the display comes from Corollary 5.7, whose proof depends on the unproved equation (3) in Theorem 2.4. Thus Theorem 5.8 is not established as written.","section":"Section 5, Theorem 5.8"}],"minor_comments":[{"comment":"The maps C_i are defined as 'C_i: C(X) → X'; they should be 'C_i: C(X) → R'. Also, the formula 'C_1(1_X)=C_2(1_X)=1_X' should use the real number 1, not the constant function 1_X.","section":"Section 2, proof of Theorem 2.4"},{"comment":"The quantities n(B_i) and n(φ_τ,B_i) are used before they are defined; the discrete Bowen-entropy analogues of N(B) should be defined explicitly.","section":"Section 3, proof of Theorem A"},{"comment":"In the displayed equation, the set 'QR(h_μ(φ_1))' on the right-hand side should be 'QR(h_μ(Φ))' (or the equality h_μ(φ_1)=h_μ(Φ) should be stated immediately before the display).","section":"Section 4, first paragraph"},{"comment":"The phrasing 'there are T_g(ε_1),...,T_g(ε_k)>0' should clarify that T_g(ε_i) is associated to each ε_i, and the concatenation times in (6) should be written more carefully (for instance, specifying that the inequality holds for each j=1,...,k).","section":"Section 5, Definition 5.3"},{"comment":"There are numerous typographical errors, including 'Corolary', 'Lema', 'Teorema', and 'Pfiste r'; the manuscript should be proofread. The sentence in Section 2 that the results are 'probably folklore' should be replaced by a precise statement about novelty or by references.","section":"Throughout"},{"comment":"Reference [19] is a closely related preprint; the paper should discuss explicitly how Theorem B compares with the results in [19], particularly since the abstract and introduction advertise a generalization of that work.","section":"Introduction and references"}],"recommendation":"major_revision","confidential_remarks":"The main gap is localized to the proof of Theorem 2.4 and the QR-entropy step in Section 4. If the authors can supply a correct bridge from flow quasi-regularity to time-one quasi-regularity (or replace it with a different argument), the paper is likely publishable. The almost-specification equivalence in Corollary 5.4 should also be checked against the literature, as it may be known but is not cited here."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a useful flow-adaptation paper with a real hole in the middle. The main inequality (Theorem B) is probably true, and the authors deserve credit for seeing how to extend the fixed-point-free result of Wang–Chen–Lin–Wu to arbitrary continuous flows and for stating the corollaries on saturated systems and irregular points. But as written, the proof does not support the claim.\n\nWhat's genuinely good: Theorem A is a clean and useful result: the Bowen entropy of a flow on a subset equals the Bowen entropy of the time-one map scaled by 1/|t|. It makes the flow entropy for noncompact sets tractable and matches Shen–Zhao's earlier definition. The application program is sensible, and the paper is honest about what it imports from [14] and [19]; there were no invented quantities or fitted constants.\n\nThe soft spot is Theorem 2.4. The direction Q(Φ)⊆Q(φ_1) needs the claim that from the integral equality ∫_0^1 C1(φ∘φ^t)dt = ∫_0^1 C2(φ∘φ^t)dt one can choose t(φ) with C1(φ∘φ^t)=C2(φ∘φ^t) and, crucially, that t(φ) varies continuously with φ. That is not proved, and as a general statement about continuous functions on a compact space it is false; the rotation-flow example with δ_0 and δ_{1/2} shows the selection can fail to be continuous. Since Section 4 uses Theorem 2.4 to place G_μ(Φ) inside QR(h_μ(Φ)), this is not a cosmetic gap. The theorem may be standard folklore, but the proof given doesn't establish it.\n\nThere's a second, smaller problem: Corollaries 5.4 and 5.8 rely on the equivalence 'Φ has the almost specification property iff φ_1 does.' That is asserted in one sentence with no proof, and the definitions (Lebesgue-measurable mistakes for flows vs. discrete gaps for maps) are different enough that the equivalence needs a real argument. If the authors fix Theorem 2.4 and add that argument, the paper will be solid.\n\nWho this is for: researchers working on entropy and multifractal analysis of continuous-time systems. It deserves a serious referee—the flaws are repairable and the results are likely publishable. If I were the editor, I'd send it out and ask the referee to focus on Theorem 2.4.","headline":"A useful flow-adaptation of Bowen's entropy inequality, but Theorem B's proof has a load-bearing unproved continuous-selection step.","tokens_in":12188,"tokens_out":6620,"would_cite":true,"duration_ms":66822,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37B40","37A35","37C45","37D40"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a continuous flow on a compact metric space, the noncompact-set topological entropy of the set of generic points of any invariant measure is bounded above by that measure's metric entropy, with equality for ergodic measures.","keywords":["Bowen topological entropy","continuous flows","generic points","metric entropy","almost specification property","saturated systems","irregular points","time-one map"],"falsifier":"Construct a continuous flow, an observable \\(\\varphi\\), and a point \\(x\\) for which the set of \\(t\\in[0,1]\\) with \\(C_1(\\varphi\\circ\\varphi^t)=C_2(\\varphi\\circ\\varphi^t)\\) has no continuous branch \\(t(\\varphi)\\); if such a construction also makes \\(G_\\mu(\\Phi)\\not\\subseteq QR(h_\\mu(\\Phi))\\) for some invariant measure \\(\\mu\\), then Theorem B's reduction to the discrete inequality fails. A direct computation of \\(h(\\Phi,G_\\mu(\\Phi))>h_\\mu(\\Phi)\\) for any continuous flow would refute the theorem itself.","tokens_in":11203,"feed_emoji":"🌀","tokens_out":11393,"duration_ms":95235,"temperature":0.7,"pith_summary":"This paper proves a flow version of a classical inequality from discrete ergodic theory: for every invariant probability measure of a continuous flow on a compact metric space, the Bowen topological entropy of the set of points whose flow averages converge to that measure is no larger than the metric entropy of the measure. When the measure is ergodic, the two entropies coincide. The proof rests on a rescaling theorem showing that the Bowen entropy of any set under a continuous flow is exactly one over |t| times the Bowen entropy of the time-t map on that same set, together with a comparison between flow-generic and time-one-generic points. The authors then use the inequality to show that flows inheriting the almost specification property from their time-one maps are saturated, that geodesic flows on negatively curved closed manifolds are saturated, and that nonempty irregular sets of such flows carry full flow entropy.","feed_headline":"Flow entropy obeys the same generic-point bound","feed_subtitle":"For every invariant measure, generic-point entropy is at most measure entropy; ergodic measures give equality.","key_machinery":"The load-bearing mechanism is the time-one comparison. Theorem A rescales Bowen topological entropy across flow time, \\(h(\\Phi,Y)=\\frac1{|t|}h(\\varphi^t,Y)\\), proved by comparing spanning covers at the scale of a Lebesgue number. Theorem 2.4 then identifies the set of quasi-regular points of the flow with that of the time-one map; its proof constructs, for each point, the \\(\\Phi\\)-invariant measure \\(\\bar\\mu_x=\\$int_0^{1}$(\\varphi^s)_*\\mu_x\\,ds\\) from the time-one empirical measure \\(\\mu_x\\), and uses Riesz representation to extract invariant measures from subsequential Birkhoff limits. Lemma 3.3, proved with Jensen's inequality on the concave function \\(-x\\log x\\), gives \\(h_\\mu(\\$varphi^{1}$)\\ge h_{\\bar\\mu}(\\$varphi^{1}$)\\), which converts the discrete bound \\(h(\\$varphi^{1}$,G_{\\mu_x}(\\$varphi^{1}$))\\le h_{\\mu_x}(\\$varphi^{1}$)\\) into the flow bound. For the applications, the named object is the almost specification property for flows, defined through a mistake function \\(g(t,\\epsilon)\\) with \\(g(t,\\epsilon)/t\\to0\\); the paper proves that a flow has this property exactly when its time-one map does, so discrete consequences transfer.","core_discovery":"The central claim is Theorem B: for a continuous flow \\(\\Phi=\\{(\\varphi^t)\\}_{t\\in\\mathbb R}\\) on a compact metric space \\(X\\), every \\(\\Phi\\)-invariant Borel probability measure \\(\\mu\\) satisfies \\(h(\\Phi, G_\\mu(\\Phi))\\le h_\\mu(\\Phi)\\), where \\(h(\\Phi, G_\\mu(\\Phi))\\) is the Bowen topological entropy of the set of \\(\\mu\\)-generic points, and equality holds whenever \\(\\mu\\) is ergodic. The companion Theorem A is equally structural: for any subset \\(Y\\subseteq X\\), the flow entropy is exactly \\(h(\\Phi,Y)=\\frac1{|t|}h(\\varphi^t,Y)\\), so flow entropy and time-\\(t\\) entropy carry the same information up to a constant rescaling. The proof strategy is to show that every flow-generic point is quasi-regular for the time-one map, that the induced time-one empirical measures have metric entropy no larger than the flow's metric entropy of \\(\\mu\\), and then to invoke the discrete version of the inequality.","pith_inferences":["If the continuous-selection gap in the proof of Theorem 2.4 is repaired, the same comparison scheme would likely prove Bowen-type inequalities for observables with values in Banach spaces, since only the Riesz-representation step would need adjustment.","Theorem A suggests that any dynamically defined quantity that is monotone under taking subsets and additive over covers can be rescaled across flow times, which could simplify numerical entropy estimation: a single time-1 calculation determines entropy at every sampling rate.","The almost-specification transfer indicates that suspension flows over maps with specification-like properties are natural test beds for saturatedness, since Theorem C reduces the check for such flows to a discrete check on the base map.","A full converse of Theorem C is unlikely in general: a flow whose time-one map is non-saturated but whose continuous averages repair the missing generic points would require a delicate balance between the discrete and continuous orbit structures."],"forward_implications":["Any continuous flow whose time-one map is saturated is itself saturated, so saturation is inherited by continuous time from discrete time.","Geodesic flows on closed manifolds of negative curvature are saturated: for every invariant measure, the set of generic points carries exactly the measure's entropy.","A continuous flow with the almost specification property is saturated, and if an irregular set for a continuous observable is nonempty, it carries the full topological entropy of the flow.","The equality \\(h(\\Phi,Y)=\\frac1{|t|}h(\\varphi^t,Y)\\) means entropy results for noncompact sets can be translated between a flow and any of its time-\\(t\\) maps with only a constant rescaling.","For ergodic measures, the flow generic set carries full metric entropy even when the measure is not ergodic for the time-one map."],"supporting_citations":[{"why":"Provides the discrete entropy inequality for generic points that the paper extends to flows.","marker":"[3]"},{"why":"Supplies the standard construction of the invariant measure \\(\\bar\\mu=\\int_0^1\\varphi^s_*\\mu\\,ds\\) used to relate flow and time-one generic points.","marker":"[17, p. 968]"},{"why":"Gives the isomorphism result used to compare metric entropies of \\(\\mu\\) and its time-\\(t\\) pushforwards.","marker":"[18, Theorem 4.11]"},{"why":"Supplies the discrete theorem that almost specification implies saturatedness, which the flow version imports via Theorem C.","marker":"[11]"},{"why":"Supplies the discrete dichotomy for irregular sets under almost specification, extended to flows in Theorem 5.8.","marker":"[16]"},{"why":"Provides the specification property for geodesic flow time-one maps used to conclude geodesic flows are saturated.","marker":"[9, Theorem 18.3.13]"}],"fun_headline_variants":["Flow entropy: generic-point bound proven","Generic-point entropy ≤ measure entropy for flows","Ergodic flows: entropy equality on generic points","Flow entropy ties to time-one map entropy","Almost specification implies saturation in flows"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that flow quasi-regular points coincide with time-one quasi-regular points assumes one can choose, continuously in the observable, a time at which two limit averages agree; no proof of that continuous selection is supplied, and the fixed-point step in Theorem 2.4 collapses if it fails.","fun_headline_variants_meta":{"raw":{"variants":["Flow entropy: generic-point bound proven","Generic-point entropy ≤ measure entropy for flows","Ergodic flows: entropy equality on generic points","Flow entropy ties to time-one map entropy","Almost specification implies saturation in flows"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000237,"raw_usage":{"total_tokens":1496,"prompt_tokens":925,"completion_tokens":571,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":541,"completion_tokens_details":{"reasoning_tokens":506}},"tokens_in":541,"tokens_out":571,"duration_ms":484966,"temperature":1.0,"reasoning_tokens":506,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:52:32.263275+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct a continuous flow, an observable \\(\\varphi\\), and a point \\(x\\) for which the set of \\(t\\in[0,1]\\) with \\(C_1(\\varphi\\circ\\varphi^t)=C_2(\\varphi\\circ\\varphi^t)\\) has no continuous branch \\(t(\\varphi)\\); if such a construction also makes \\(G_\\mu(\\Phi)\\not\\subseteq QR(h_\\mu(\\Phi))\\) for some invariant measure \\(\\mu\\), then Theorem B's reduction to the discrete inequality fails. A direct computation of \\(h(\\Phi,G_\\mu(\\Phi))>h_\\mu(\\Phi)\\) for any continuous flow would refute the theorem itself.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the discrete entropy inequality for generic points that the paper extends to flows."},{"cited_title":"and Vericat, F., Saturatedness of dynamical systems under the almost specifi cation property, J","cited_arxiv_id":null,"evidence_quote":"Supplies the discrete theorem that almost specification implies saturatedness, which the flow version imports via Theorem C."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the discrete dichotomy for irregular sets under almost specification, extended to flows in Theorem 5.8."}],"review_version":1}