{"id":"bed92f70-de83-4800-ae45-d22a554bc626","arxiv_id":"2507.03338","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For amenable group actions, IE-tuples of the induced measure system equal the closed convex hull of IE-tuples of the original system, with partial analogues for IN and IT tuples and a sharp n ≤ C log m embedding bound for ℓ_q^n inside ℓ_∞^m.","lead":"Each continuous action of a group on a compact space induces an action on its space of probability measures. This paper proves that the local entropy invariants of the induced action are convex hulls of those of the original action, and uses a new combinatorial lemma to get a sharp logarithmic bound for embedding ℓ_q^n into ℓ_∞^m.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the intricate proof of Lemma 1.8 appears internally consistent, and the main theorems follow without a discovered gap.","rationale":"The paper's central claim is Theorem 1.1(1), which is proved by reducing any failure to Lemma 3.7 and then using Lemma 3.24, whose combinatorial engine is Lemma 1.8. I focused my review on the proof of Lemma 1.8, since that is the most intricate and least independently verifiable part. I checked the main estimates in Lemma 3.19: the construction of the threshold sets Z_{ψ,i}, the use of Lemma 3.16 to control the poor equidistribution set, the covering argument using Lemma 3.18, and the final integral bound. The integral bound uses Hölder's inequality on a possibly signed sum; as an upper bound this is legitimate because replacing f_ψ by its positive part can only increase the sum. The derivation of the lower bound on N_S from the fiber counting argument is also correct: it shows every cover W has size at least e^{δ|Z|/|I|}|I|, hence the minimal cover does too. The cardinality inequality |R(Z,k)| ≥ |R(Z',k)| in Lemma 3.20 is true as a count (the ratio of the explicit formulas is >1), even though a naive extension of every ψ' may fail. The applications in Lemma 3.24 and Theorem 1.9 align the parameters correctly, including the p=q/(q−1) duality in Theorem 1.9. I found no circular reasoning or unsupported citations; Lemma 3.18 is a published result. Consequently, I cannot identify a load-bearing concern. The residual risk is the high complexity of the combinatorial proof, which the reader already noted, but without a concrete flaw this does not justify changing the verdict.","tokens_in":1053,"tokens_out":786,"duration_ms":378152,"concrete_test":"Independently re-derive the contradiction chain at the end of Lemma 3.19 for a small explicit case (k=2, p=∞, C=1, R=1/2, r=1/4), checking that the final inequality gives R|Z|^{1/q} ≤ Σ f_ψ(z) < R|Z|^{1/q}. If the strict inequality reproduces, the main combinatorial lemma is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I traced the central argument to Lemma 1.8 and its proof via Lemma 3.19. The proof is long, but I found no concrete error. Specifically: the Hölder upper bound on the signed sum over the complement is valid because negative terms only lower the sum; the lower bound on the covering number N_S is obtained by observing that every cover W must have |W| large, so the minimum cover is also large; and the inequality |R(Z,k)| ≥ |R(Z',k)| used in Lemmas 3.14 and 3.20 is a true cardinality statement, even though individual ψ' in R(Z',k) need not extend to R(Z,k) when |Z\\Z'|>1. The applications in Lemma 3.24 and Theorem 1.9 match the hypotheses of Lemma 1.8 with the correct normalization. The only residual risk is the general complexity of the combinatorial estimates, not a specific identified flaw.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the induced action of a countably infinite discrete group Γ on the compact convex space M(X) of Borel probability measures on X. For amenable Γ, Theorem 1.1 characterizes IE_k(M(X)) as the closed convex hull of IE_k(X), identifies IE_k(X) as the set of extreme points, and gives the explicit barycentric description IE_k(M(X)) = {(μ(1),...,μ(k)) : μ ∈ M(IE_k(X))}. Analogous convex-hull results are proved for measure-IE tuples (Theorem 1.3), IN-tuples and measure-IN tuples (Theorems 1.4 and 1.6), and IT-tuples (Theorem 1.7), with the necessary modifications coming from the non-convexity examples in Section 8. The main technical tool is Lemma 1.8, a new combinatorial lemma proved in Section 3.3; it is also used in Section 9 to prove Theorem 1.9, a sharp logarithmic upper bound for embedding ℓ_q^n into ℓ_∞^m. The paper includes explicit obstructions and sharpness examples for its key lemmas.","tokens_in":64073,"tokens_out":6205,"duration_ms":76012,"significance":"If correct, these results give a complete and unified description of how local entropy-theoretic invariants behave under the quasi-factor map from X to M(X). They recover and extend the Glasner-Weiss and Glasner-Thouvenot-Weiss theorems, and the new combinatorial lemma in Section 3.3 has potential for further applications. The Banach-space application in Theorem 1.9 is concrete and yields a sharp bound. The paper is commendably explicit about where the k≥3 case differs from k=2 (Example 3.9) and about the sharp threshold in Lemma 3.14 (Example 3.15). All main proofs are present, and I found no concrete gap; the main residual risk is the complexity of the proof of Lemma 1.8, but the written argument appears internally consistent.","major_comments":[],"minor_comments":[{"comment":"The symbol R is overloaded: it is a real parameter in the hypothesis 0<r<R≤C and also a subset R⊆[k]^Z in the condition '|R| ≥ |[k]^Z| e^{-δ|Z|}'. Renaming one of them, for instance calling the subset S, would remove an unnecessary source of confusion for the reader.","section":"Section 3.3 (Lemma 3.19)"},{"comment":"The final estimate uses the inequality log(2m) ≤ 2 log m, which is valid for m≥2, but this step is implicit. Spelling it out would make the end of the proof easier to follow.","section":"Section 9 (proof of Theorem 1.9)"},{"comment":"The definition of IN-measures quantifies over n-tuples of points in supp(μ), and Remark 5.2 explains that repetitions do not change the definition. The explanation is correct, but it would be clearer if the closedness of IN_k(X^n) were explicitly invoked at the point where the embedding argument is used.","section":"Section 5 (Definition 5.1 and Remark 5.2)"}],"recommendation":"accept","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a strong paper and deserves a serious referee. Theorem 1.1 settles the natural open question — for every k, IE_k(M(X)) is the closed convex hull of IE_k(X), the points of IE_k(X) are exactly the extreme points, and the explicit barycentric description holds. That completes and generalizes the Glasner-Weiss k=2 result and the recent special cases. The companion results for measure-IE, IN, measure-IN and IT tuples are also new and are proved with the same machinery, with the non-convexity of IN/IT handled honestly by introducing the more subtle MIN/MIT measure classes. The Section 8 Toeplitz example is a useful counterpoint, not decoration.\n\nThe real novelty is Lemma 1.8, a combinatorial lemma that is genuinely stronger than the Sauer-Perles-Shelah/Karpovsky-Milman line. It is not an incremental variation: the k≥3 case of Theorem 1.1 fails for the natural k=2 argument (Example 3.9), and the proof of Lemma 1.8 is the load-bearing part. The authors also use the p>1 range for the Banach space application, Theorem 1.9, which is a nice payoff and a nontrivial bonus.\n\nWhere are the soft spots? The proof of Lemma 1.8 is long and technically dense. I went through the stress-test concerns — the Hölder estimate, the covering-number lower bound, the cardinality monotonicity in R(Z,k) — and they do not land; the written proof is internally consistent. But this is exactly the kind of argument where a subtle normalization error can hide, and I did not verify every line. The authors do not provide an independent check (no code, no short certificate), so the residual risk is the usual one for a hard combinatorial proof, not a discovered flaw. I would not hold that against the paper.\n\nThe citation pattern is clean. The paper leans on Kerr-Li's framework, including one of that pair's lemmas, and carefully credits Glasner-Weiss, Liu-Wei, Vermersch and others. The reliance is legitimate: the cited results are real and the new theorems do not reduce to them.\n\nReadership: topological dynamics people working on local entropy theory, and Banach space geometers interested in the embedding bound. It is a good reading-group paper, but the payoff comes in the introduction and the applications; Section 3.3 is heavy. Send it to a careful referee who is willing to spend time on the combinatorics. I would accept it.","headline":"A strong paper that gives the complete description of IE-tuples for the induced measure action and introduces a genuinely new combinatorial tool; the only substantial risk is the density of the Lemma 1.8 proof.","tokens_in":64690,"tokens_out":2592,"would_cite":true,"duration_ms":31744,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37B40","37A15","05D05","46B07"],"pacs":[],"model":"deepseek-v4-flash","headline":"For amenable group actions, the IE-tuples of the induced action on probability measures form exactly the closed convex hull of the original IE-tuples.","keywords":["local entropy theory","IE-tuple","IN-tuple","IT-tuple","induced action on probability measures","combinatorial independence","amenable group action","ℓ_p embedding"],"falsifier":"A concrete counterexample to Lemma 1.8 would settle the matter: produce k, thresholds 0<r<R≤C, exponents 1<p≤∞ and 1≤q<∞, and functions f_ψ on finite sets Z with ∥f_ψ∥_p≤C and normalized average at least R, yet with no positive-density subset J and no threshold t∈T satisfying the extension property. Equivalently, an amenable group action with IE_2(M(X)) containing a measure pair outside the closed convex hull of IE_2(X) would contradict the main structural theorem.","tokens_in":63703,"feed_emoji":"","tokens_out":7508,"duration_ms":81112,"temperature":0.7,"pith_summary":"The paper studies the action a discrete amenable group induces on the compact convex space of Borel probability measures over a system, comparing the local entropy theory of the induced action with that of the original. Its first main result is a complete structural description: for every length k, the set of IE-tuples of the induced action is the closed convex hull of the IE-tuples of the original action, its extreme points are exactly the original IE-tuples, and every induced IE-tuple is the tuple of coordinate marginals of a measure supported on the original IE-tuples. This turns the earlier zero-entropy result for induced actions into a direct corollary and yields finite-support versions of the description. The same framework is extended to measure IE-tuples, IN-tuples, measure IN-tuples, and IT-tuples, with explicit, though more delicate, descriptions for the latter classes. A new combinatorial lemma drives the proof and also produces a sharp logarithmic bound for embeddings of ℓ_q^n into ℓ_∞^m.","feed_headline":"Induced entropy tuples are exactly convex hulls of the original ones","feed_subtitle":"The induced action on probability measures adds no new independence tuples: only convex mixtures of the original ones.","key_machinery":"The load-bearing object is Lemma 1.8, a new finite combinatorial lemma. It fixes k, thresholds 0<r<R≤C, and conjugate exponents 1<p≤∞ and 1≤q<∞, and it produces a small threshold set T plus a positive-density subset J of any sufficiently large finite set Z. Whenever a family of functions f_ψ, indexed by nearly balanced colorings ψ in R(Z,k), has ℓ_p norm at most C and normalized average at least R, the lemma guarantees J and a threshold t∈T such that every coloring of J extends to one of the ψ with f_ψ(z)≥t_j|Z|^{-1/p} on the j-th color class. Its proof combines a concentration estimate with a covering lemma from the independence literature, and the p=∞ case drives the entropy-tuple theorems while the full range 1<p≤∞ powers the Banach-space application through Lemma 9.1.","core_discovery":"Theorem 1.1 asserts that IE_k(M(X)) equals the closed convex hull of IE_k(X) in M(X)^k, that IE_k(X) is exactly IE_k(M(X)) intersected with X^k, and that IE_k(M(X)) is the set of tuples (μ^(1),…,μ^(k)) obtained as coordinate marginals of a Borel probability measure μ supported on IE_k(X). Since X is identified with its Dirac measures inside M(X), the statement says concretely that no new independence combinatorics appear when passing to probability measures: every IE-tuple of measures is an averaging of point IE-tuples, and the only such tuples that live entirely in X are the original point IE-tuples. The paper proves analogous containments and intersection identities for measure IE-tuples, IN-tuples, measure IN-tuples, and IT-tuples, with the IN and IT cases described through special subclasses of measures that encode independence across all Cartesian powers of X. It also proves a Banach-space consequence: if ℓ_q^n is C-isomorphic to a subspace of ℓ_∞^m, then n is at most c log m, where c depends only on C and q.","pith_inferences":["One can read the convex-hull conclusion as a general principle: for tuple notions defined by combinatorial independence, passing to a compact convex envelope such as M(X) should produce at most the closed convex hull of the original tuple set, with the IN and IT cases showing that the precise formulation requires additional measure-theoretic bookkeeping.","The p>1 range of Lemma 1.8 is likely to be usable beyond ℓ_∞; discretizing coordinate thresholds in the manner of Lemma 9.1 may yield a unified proof of the known bounds for embedding ℓ_q^n into other ℓ_{q'}^m spaces and may extend to non-Euclidean target norms.","A testable extension is whether the equality IE_k(M(X)) equals the closed convex hull of IE_k(X) survives when Γ is only sofic rather than amenable; the amenability hypothesis in Theorem 1.1 is used through the independence-density technology, while the IN and IT analogues hold for arbitrary countable groups.","The explicit description via M(IE_k(X)) suggests a finite algorithm for membership in IE_k(M(X)): given a finite approximation of IE_k(X), project the measure polytope to M(X)^k and check whether the target tuple lies in that projection."],"forward_implications":["Zero topological entropy of X forces zero topological entropy of M(X) for every amenable group Γ, with the k=2 case of Theorem 1.1 recovering the earlier zero-entropy theorem as a direct consequence.","Uniform positive entropy of order k, nullness, and tameness pass between the original action and the induced action on probability measures, and also pass through the weighted simplex spaces M_λ(X).","Finite-support measures cannot create new IE, IN, or IT tuples beyond the originals; the product capacity N=∏N_j bounds the support size in Corollaries 1.2, 1.5, and 7.12.","Any C-isomorphic copy of ℓ_q^n inside ℓ_∞^m has dimension at most c log m, with c depending only on C and q, and this logarithmic bound is sharp.","The descriptions of IN and IT tuples use measures that encode independence across all Cartesian powers X^n, replacing the simpler measure set M(IE_k(X)) by the more delicate IN-measures and IT-measures."],"supporting_citations":[{"why":"Defines IE, IN, and IT tuples for general group actions and supplies the covering lemma that the new combinatorial lemma builds on.","marker":"[53]"},{"why":"Defines measure IE-tuples and measure IN-tuples and proves their coincidence with measure entropy tuples, providing the setting for Theorems 1.3 and 1.6.","marker":"[54]"},{"why":"Provides the standard local entropy theory facts about closures, factor maps, and independence density used throughout the paper.","marker":"[56]"},{"why":"Established the zero-entropy correspondence for induced actions on probability measures that the k=2 IE result recovers.","marker":"[35]"},{"why":"Supplies the coordinate-density dichotomy that underlies the growth estimates in the combinatorial lemmas.","marker":"[50]"},{"why":"Gives the optimal concentration-based embedding bounds for ℓ_q into ℓ_{q'} that Theorem 1.9 sharpens in the case q'=∞.","marker":"[20]"},{"why":"Provides the dichotomy used for IT-tuples and the tame-versus-untame consequences.","marker":"[74]"},{"why":"Initiated the study of the induced action on probability measures and showed that positive entropy becomes infinite on the measure space.","marker":"[5]"}],"fun_headline_variants":["Measure entropy tuples are just convex hulls of point ones","No new independence tuples from probability measures","Induced IE-tuples on measures: convex hulls of originals","Measure action entropy: convex mixtures of point tuples only"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument stands or falls with Lemma 1.8, a new finite combinatorial lemma whose long proof leans on a concentration estimate and a cited covering lemma; if that chain has a gap, every main theorem that depends on it loses its support.","fun_headline_variants_meta":{"raw":{"variants":["Measure entropy tuples are just convex hulls of point ones","No new independence tuples from probability measures","Induced IE-tuples on measures: convex hulls of originals","Measure action entropy: convex mixtures of point tuples only"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000844,"raw_usage":{"total_tokens":3636,"prompt_tokens":867,"completion_tokens":2769,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":483,"completion_tokens_details":{"reasoning_tokens":2703}},"tokens_in":483,"tokens_out":2769,"duration_ms":22103,"temperature":1.0,"reasoning_tokens":2703,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:13:13.532265+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete counterexample to Lemma 1.8 would settle the matter: produce k, thresholds 0<r<R≤C, exponents 1<p≤∞ and 1≤q<∞, and functions f_ψ on finite sets Z with ∥f_ψ∥_p≤C and normalized average at least R, yet with no positive-density subset J and no threshold t∈T satisfying the extension property. Equivalently, an amenable group action with IE_2(M(X)) containing a measure pair outside the closed convex hull of IE_2(X) would contradict the main structural theorem.","supporting_citations":[{"cited_title":"Kerr and H","cited_arxiv_id":null,"evidence_quote":"Defines IE, IN, and IT tuples for general group actions and supplies the covering lemma that the new combinatorial lemma builds on."},{"cited_title":"Kerr and H","cited_arxiv_id":null,"evidence_quote":"Defines measure IE-tuples and measure IN-tuples and proves their coincidence with measure entropy tuples, providing the setting for Theorems 1.3 and 1.6."},{"cited_title":"Kerr and H","cited_arxiv_id":null,"evidence_quote":"Provides the standard local entropy theory facts about closures, factor maps, and independence density used throughout the paper."},{"cited_title":"Glasner and B","cited_arxiv_id":null,"evidence_quote":"Established the zero-entropy correspondence for induced actions on probability measures that the k=2 IE result recovers."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the coordinate-density dichotomy that underlies the growth estimates in the combinatorial lemmas."},{"cited_title":"Figiel, J","cited_arxiv_id":null,"evidence_quote":"Gives the optimal concentration-based embedding bounds for ℓ_q into ℓ_{q'} that Theorem 1.9 sharpens in the case q'=∞."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the dichotomy used for IT-tuples and the tame-versus-untame consequences."},{"cited_title":"Bauer and K","cited_arxiv_id":null,"evidence_quote":"Initiated the study of the induced action on probability measures and showed that positive entropy becomes infinite on the measure space."}],"review_version":1}