{"id":"61e03701-d8a1-408e-ad7b-ea3ace365e55","arxiv_id":"2502.03267","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A measurable function is exactly one whose dyadic Hausdorff-content Choquet averages recover it at almost every point, and continuous functions are not dense in the corresponding spaces for content dimension delta < n.","lead":"This mathematics paper characterizes Lebesgue measurability through limits of Choquet averages with respect to dyadic Hausdorff content. It also proves that continuous functions fail to be dense in the associated normed spaces when the content dimension is smaller than the ambient dimension.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Measurability of the uncountable limsup in Proposition 5.2(2) is the real gap; the reader's translation-invariance objection is not load-bearing for δ=n.","rationale":"The reader's conditional verdict is reasonable in spirit: Proposition 5.2 as stated for arbitrary δ has an unjustified equality, and the proof of Corollary 1.1 relies on the measurability conclusion of that proposition. However, the specific failure of exact translation invariance only matters when δ<n. For the n-dimensional content used in Corollary 1.1, H^n∞ is exactly Lebesgue measure on balls, so H^n∞(B(x,r))=H^n∞(B(y,r)) is true. The genuine remaining issue in the central claim is that the limsup over an uncountable family is not automatically a pointwise limit of measurable functions; the paper needs the standard inf/sup argument using lower semicontinuity. This is a fixable proof gap rather than a falsified statement, so keeping the verdict at CONDITIONAL with a request for a clarified proof is appropriate. The core mathematical characterization appears correct.","tokens_in":17168,"tokens_out":25208,"duration_ms":241794,"concrete_test":"For δ=n, show directly from the dyadic-cube definition that H^n∞(B(x,r))=ω_n r^n for every x and r, which makes the equality in Proposition 5.2(1) valid. Then verify that for every locally Choquet-integrable f the map r↦∫_{B(x,r)} f dH^n∞ is continuous in r, so the limsup over all real r equals the limsup over rational r and the function L(x)=inf_k sup_{0<r<1/k} (1/H^n∞(B(x,r)))∫_{B(x,r)} f dH^n∞ is Borel measurable. If both checks pass, Corollary 1.1's reverse direction is sound after a minor rewrite.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reverse direction of Corollary 1.1 requires the function L(x)=limsup_{r→0+} f^n_{B(x,r)} to be Lebesgue measurable. Proposition 5.2(2) asserts this is a pointwise limit of the measurable functions x↦f^n_{B(x,r)}, but the limsup is over the uncountable set r>0 and is not a sequential pointwise limit. This is a genuine gap in the written proof of the central claim. The reader's separate objection that H^δ∞(B(x,r))=H^δ∞(B(y,r)) is false does not affect Corollary 1.1 because for δ=n the dyadic Hausdorff content of an open Euclidean ball equals its Lebesgue measure, hence is translation invariant; for δ<n Proposition 5.2(1) is indeed unproved and needs a different argument. The uncountable-limsup measurability can be repaired: part (1) of Proposition 5.2 gives lower semicontinuity, so each G_k=sup_{0<r<1/k} g_r is lower semicontinuous and L=inf_k G_k is Borel measurable. As written, however, the stated justification is incorrect.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops Lebesgue point theory for Choquet integrals with respect to dyadic Hausdorff content, working with functions that are not assumed Lebesgue measurable. It proves auxiliary monotone convergence and Fatou-type results, maximal-function estimates, a norm convergence theorem, density results for continuous functions in the spaces nL1(R^n, H^δ∞), and a central characterization (Corollary 1.1): for nonnegative f with finite Choquet integrals on every open ball relative to H^n∞, f is Lebesgue measurable if and only if f equals the limsup of its Choquet averages on balls, H^n∞-almost everywhere. The paper also gives two examples for δ<n showing that Lebesgue measurable functions need not be nL1-limits of continuous functions.","tokens_in":17217,"tokens_out":15499,"duration_ms":143731,"significance":"If Corollary 1.1 is correct, it is a genuinely interesting result: it characterizes Lebesgue measurability purely through the pointwise behavior of Choquet averages with respect to dyadic Hausdorff content. The companion density and non-density results for continuous functions in nL1(R^n, H^δ∞), especially the counterexamples for δ<n, are also valuable and are supported by explicit constructions. The paper builds on published work [7] for the weak-type maximal estimate, and it gives self-contained convergence and approximation arguments for much of the development. However, the proof of the key measurability assertion in Proposition 5.2 contains a real gap, and several auxiliary results are imported from the authors' own unpublished submitted manuscript [21], which makes the current version difficult to verify as submitted.","major_comments":[{"comment":"The proof of claim (2) says that limsup_{r→0+} f^δ_{B(x,r)} is Lebesgue measurable 'as a pointwise limit of Lebesgue measurable functions.' This is not justified: the limsup is over the uncountable set r>0 and is not a sequential pointwise limit. The measurability of this limsup is exactly what Proposition 5.3 and the reverse direction of Corollary 1.1 require, so the gap is load-bearing. The claim is repairable: if part (1) is available, then G_k(x)=sup_{0<r<1/k} f^δ_{B(x,r)} is lower semicontinuous as a supremum of lower semicontinuous functions, and L(x)=inf_k G_k(x) is Borel measurable. The authors should replace the 'clear' argument with this argument or an equivalent one.","section":"Section 5, Proposition 5.2(2)"},{"comment":"The proof of lower semicontinuity uses the equality H^δ∞(B(x,r)) = H^δ∞(B(y,r)) for balls with different centers. Dyadic Hausdorff content with respect to a fixed dyadic lattice is not translation invariant when δ<n, so this equality is not available in the stated generality. For δ=n the equality is valid because the dyadic n-dimensional content of an open Euclidean ball agrees with its Lebesgue measure, hence is translation invariant, but for δ<n the proof of Proposition 5.2(1) is incomplete as written. If Proposition 5.2 is needed only for δ=n in the proof of Corollary 1.1, the authors should say so and justify the δ=n equality separately; otherwise a different argument is needed for δ<n.","section":"Section 5, Proposition 5.2(1)"},{"comment":"The paper repeatedly relies on the authors' unpublished submitted manuscript [21] for load-bearing results: Remark 3.3 (comparability of the Choquet integral with the Lebesgue integral), Theorem 4.11 (strong-type estimate for Mn, proved in one line from [21, Theorem 4.6]), the ball-content comparability H^δ∞(B(x,r))≈r^δ used in Examples 6.1 and 6.2, and the definition and properties of nL1(R^n, H^δ∞). Since [21] is not available to referees or readers, these results cannot be checked from the submitted manuscript. The authors should include the relevant statements and proofs from [21], or replace them with published references, so that the central claims of the paper are self-contained in the respects on which they depend.","section":"Sections 3 and 4, Theorem 4.11, Remark 3.3"},{"comment":"The covering condition 'E ⊂ int(∪ Qi)' is problematic for the half-open dyadic cubes defined immediately before it. For example, the origin is never contained in the interior of a dyadic cube of the form [m2^k,(m+1)2^k)^n in the fixed lattice, so under the literal definition H^δ∞ of any set containing the origin would be infinite. This would make the assumption in Corollary 1.1 fail for very simple functions such as f≡1 on a ball containing the origin. The authors should clarify the intended convention, presumably E⊂∪ Qi or E⊂∪ int(Qi) as in [42], and ensure that the properties (H1)-(H5) are stated for that convention.","section":"Section 2, definition of H^δ∞"}],"minor_comments":[{"comment":"In the sentence 'χΩ(x) /nequallim sup', the symbol '/nequal' should be '≠'.","section":"Example 6.2"},{"comment":"There is a typo: 'resent results' should be 'recent results'.","section":"Remark 4.10"},{"comment":"There are several typographical inconsistencies, including 'continuos' in the abstract, 'difference' and 'different' with nonstandard ligature breaks, and the garbled phrase 'f (x)] d ˜H δ∞' in the proof of Theorem 4.11. These should be corrected in a final revision.","section":"Throughout"},{"comment":"In the chain of inequalities after the choice of η, the ratio H^δ∞(B(x,r))/H^δ∞(B(y,r)) is written as a multiplier; once the equality of contents is removed (as it must be for δ<n), the displayed calculation should be rewritten in terms of a comparability constant, and the openness of the superlevel set should be verified with that constant.","section":"Section 5, proof of Proposition 5.2(1)"}],"recommendation":"major_revision","confidential_remarks":"The main characterization in Corollary 1.1 is potentially novel and attractive, but the current manuscript is not verifiable as submitted because of the uncountable-limsup gap in Proposition 5.2(2), the incorrect translation-invariance step for δ<n, and the heavy reliance on the unpublished manuscript [21]. The first two issues are repairable, and the paper would be suitable for publication after the authors supply a self-contained proof of the measurability claim and either include or replace the cited results from [21]. Editors may wish to consider whether [21] is expected to appear before this paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the headline: this is a worthwhile paper. The measurability characterization (Corollary 1.1) is new and elegant, and the Section 6 examples showing that continuous functions are not dense in nL1(R^n, H^delta_infty) for delta<n are a nice sharpness result. The paper is clearly written and gives a fair accounting of what is new versus what reduces to prior work: Theorem 5.12(1) is explicitly credited to [7], and Theorem 5.11 to Spector.\n\nThe soft spots are real but reparable. The proof of Proposition 5.2(1) uses the equality H^delta_infty(B(x,r)) = H^delta_infty(B(y,r)) for dyadic Hausdorff content of balls with different centers. For delta=n this equality is true, since H^n_infty(B(x,r)) equals the n-dimensional Lebesgue measure of the ball. For delta<n it is generally false—the dyadic lattice breaks translation invariance—and the proof needs a comparability argument instead. That affects the proposition as stated, but not Corollary 1.1, which only needs delta=n. More importantly, the justification of (2) is wrong as written: a limsup over all r>0 is not a pointwise limit of the functions x -> f^delta_{B(x,r)} over the uncountable index set. The claim is still true—one can use (1) to get lower semicontinuity, then take sup over r<1/k and inf over k—but the stated proof does not work.\n\nThe bigger verification issue is the dependence on the authors' own submitted manuscript [21] for the strong-type maximal estimate (Theorem 4.11) and for comparability of ball contents. Since [21] is not publicly available, a referee cannot check these steps without a preprint or detailed statements. Overall, the central results are likely correct, the gap is local and fixable, and the examples are valuable. This deserves peer review, not desk rejection. I would send it to a referee with a request to verify Proposition 5.2 and to ask the authors for the needed [21] results.","headline":"Worthwhile paper with a new measurability characterization and sharp counterexamples; the proof of Proposition 5.2 has a reparable gap and the paper leans on an unpublished manuscript, but the core results stand up to a referee.","tokens_in":17933,"tokens_out":6213,"would_cite":true,"duration_ms":54922,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A25","28A20","42B25"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that a nonnegative function with locally finite Choquet integrals is Lebesgue measurable exactly when it is the almost-everywhere limsup of its dyadic Hausdorff ball averages.","keywords":["Choquet integral","dyadic Hausdorff content","dyadic Hausdorff capacity","Lebesgue point","maximal operator","non-measurable function","quasicontinuity","Lebesgue measurability"],"falsifier":"Compute $\\widetilde H^n_\\infty(B(x,r))$ and $\\widetilde H^n_\\infty(B(y,r))$ for two balls of equal radius whose centers differ by a non-dyadic vector, using the dyadic cube cover definition; unequal values would disprove the equality used in Proposition 5.2(1), while equal values for every such pair would validate the lower semicontinuity step and with it the reverse direction of Corollary 1.1.","tokens_in":16801,"feed_emoji":"📏","tokens_out":12364,"duration_ms":95290,"temperature":0.7,"pith_summary":"This paper studies Lebesgue points for Choquet integrals taken with respect to the dyadic Hausdorff content, a set function that assigns a size to every set, measurable or not. Its central result characterizes Lebesgue measurability: for a non-negative function with finite Choquet integrals over balls, the function is Lebesgue measurable if and only if $f(x)=\\limsup_{r\\to 0+}\\frac{1}{\\widetilde H^n_\\infty(B(x,r))}\\int_{B(x,r)} f\\,d\\widetilde H^n_\\infty$ holds for $\\widetilde H^n_\\infty$-almost every $x$. The forward direction is a Lebesgue-point theorem for measurable functions, while the reverse direction shows that the limsup of Choquet ball averages is itself measurable, so a function recovered from its averages cannot conceal non-measurability. The paper also proves convergence theorems, weak-type maximal estimates, and a density result showing that continuous functions are dense in the $n$-dimensional space but not in the spaces of dimension $\\delta<n$.","feed_headline":"Choquet ball averages characterize Lebesgue measurability","feed_subtitle":"A function is Lebesgue measurable exactly when dyadic Hausdorff ball averages recover it almost everywhere.","key_machinery":"The central objects are the $\\delta$-dimensional dyadic Hausdorff content $\\widetilde H^\\delta_\\infty$, defined by covering sets with countably many dyadic cubes and minimizing the sum of side-length powers, and the Choquet integral $\\int_\\Omega f\\,d\\widetilde H^\\delta_\\infty=\\int_0^\\infty \\widetilde H^\\delta_\\infty(\\{f>t\\})\\,dt$. The operative mechanism is the ball-average map $f^\\delta_{B(x,r)}$ and its limsup as $r\\to0+$. The argument hinges on showing that $x\\mapsto f^\\delta_{B(x,r)}$ is lower semicontinuous, that the resulting limsup is Lebesgue measurable, that the $n$-dimensional maximal operator satisfies a weak-type estimate, and that being an $nL^1$-limit of continuous functions is equivalent to $\\widetilde H^\\delta_\\infty$-quasicontinuity.","core_discovery":"The paper's central discovery is the equivalence in Corollary 1.1. Given $f:\\mathbb R^n\\to[0,\\infty]$ with $\\int_B f\\,d\\widetilde H^n_\\infty<\\infty$ for every open ball $B$, the function $f$ is Lebesgue measurable exactly when $f(x)=\\limsup_{r\\to0+} f^n_{B(x,r)}$ for $\\widetilde H^n_\\infty$-almost every $x$, where $f^n_{B(x,r)}$ denotes the Choquet average of $f$ over $B(x,r)$ with respect to the $n$-dimensional dyadic Hausdorff content. The forward direction runs through the Lebesgue-point theorem for functions that are $nL^1$-limits of continuous functions, using the fact that at $\\delta=n$ the Choquet integral is comparable to the Lebesgue integral and that continuous functions are dense in $L^1$. The reverse direction uses lower semicontinuity of the ball-average map and measurability of its limsup to conclude that any function satisfying the recovery formula agrees almost everywhere with a Lebesgue measurable function.","pith_inferences":["Editorial inference: if the reverse-direction proof can be completed without exact translation invariance of ball content, the characterization may extend to other capacities whose ball masses are merely comparable, since only lower semicontinuity of the average map and a weak-type maximal estimate are needed.","Editorial inference: the examples suggest a dimension-threshold phenomenon, namely that failure of Lebesgue-point recovery is tied to boundary sets of positive $\\delta$-dimensional content; one could test whether centering dyadic cubes at $x$ instead of balls restores recovery for all $0<\\delta\\le n$.","Editorial inference: the characterization offers a route to detect non-measurability numerically by approximating the limsup of Choquet averages on a grid and comparing it with the function on the complement of a small exceptional set."],"forward_implications":["A Lebesgue measurable non-negative function with locally finite Choquet integral is recovered, up to an $\\widetilde H^n_\\infty$-null set, by the limsup of its dyadic-Hausdorff ball averages.","The same recovery formula forces measurability: functions that are not Lebesgue measurable cannot satisfy the average-recovery identity almost everywhere.","For $\\delta=n$, every Lebesgue measurable $f\\in nL^1(\\mathbb R^n,\\widetilde H^n_\\infty)$ is an $nL^1$-limit of continuous functions, so continuous functions are dense in that space.","For $0<\\delta<n$, Lebesgue measurable functions such as characteristic functions of balls or snowflake-type domains fail the recovery formula at boundary points, showing that continuous functions are not dense in $nL^1(\\mathbb R^n,\\widetilde H^\\delta_\\infty)$.","Being an $nL^1$-limit of continuous functions is equivalent to $\\widetilde H^\\delta_\\infty$-quasicontinuity, and such functions have Choquet Lebesgue points almost everywhere."],"supporting_citations":[{"why":"It supplies the weak-type estimate for the Hausdorff-content maximal operator and the Lebesgue-point theorem for quasicontinuous functions that Theorem 5.12 extends.","marker":"[7]"},{"why":"It provides the Choquet integral foundations, including sublinearity and the monotone-convergence framework used throughout the paper.","marker":"[13]"},{"why":"It defines the $nL^1$ spaces and supplies the comparability $\\widetilde H^\\delta_\\infty(B(x,r))\\approx r^\\delta$ and strong-type estimates for the maximal operator.","marker":"[21]"},{"why":"It establishes the capacity properties of monotonicity, subadditivity, and strong subadditivity, as well as dyadic-content comparability, that the arguments rely on.","marker":"[42]"},{"why":"It gives the equivalence between quasicontinuity and being an $nL^1$-limit of continuous functions used in Theorem 5.11.","marker":"[33]"},{"why":"It supplies the classical density of compactly supported continuous functions in $L^1$ used in the $\\delta=n$ case of Theorem 5.9.","marker":"[6]"},{"why":"It constructs the modified von Koch snowflake with prescribed Hausdorff dimension used in Example 6.2 to show sharpness.","marker":"[22]"}],"fun_headline_variants":["Measurable iff Hausdorff ball averages recover","Choquet ball averages decide measurability","Dyadic Hausdorff averages characterize measurability","Hausdorff content recovery equals measurability","Choquet limsup equals f iff measurable"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument that the limsup of ball averages is measurable depends on the claim that the dyadic Hausdorff content of a ball does not change when the ball's center is moved, so that the map $x\\mapsto f^\\delta_{B(x,r)}$ is lower semicontinuous; if that translation invariance fails, the measurability of the limsup needs a different proof.","fun_headline_variants_meta":{"raw":{"variants":["Measurable iff Hausdorff ball averages recover","Choquet ball averages decide measurability","Dyadic Hausdorff averages characterize measurability","Hausdorff content recovery equals measurability","Choquet limsup equals f iff measurable"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000934,"raw_usage":{"total_tokens":3943,"prompt_tokens":838,"completion_tokens":3105,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":454,"completion_tokens_details":{"reasoning_tokens":3032}},"tokens_in":454,"tokens_out":3105,"duration_ms":23237,"temperature":1.0,"reasoning_tokens":3032,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T05:20:00.133698+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $\\widetilde H^n_\\infty(B(x,r))$ and $\\widetilde H^n_\\infty(B(y,r))$ for two balls of equal radius whose centers differ by a non-dyadic vector, using the dyadic cube cover definition; unequal values would disprove the equality used in Proposition 5.2(1), while equal values for every such pair would validate the lower semicontinuity step and with it the reverse direction of Corollary 1.1.","supporting_citations":[{"cited_title":"27, Klu wer Academic Publish- ers Group, Dordrecht, 1994","cited_arxiv_id":null,"evidence_quote":"It provides the Choquet integral foundations, including sublinearity and the monotone-convergence framework used throughout the paper."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It defines the $nL^1$ spaces and supplies the comparability $\\widetilde H^\\delta_\\infty(B(x,r))\\approx r^\\delta$ and strong-type estimates for the maximal operator."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It establishes the capacity properties of monotonicity, subadditivity, and strong subadditivity, as well as dyadic-content comparability, that the arguments rely on."},{"cited_title":"In: Lenhart, S., Xiao, J","cited_arxiv_id":null,"evidence_quote":"It gives the equivalence between quasicontinuity and being an $nL^1$-limit of continuous functions used in Theorem 5.11."},{"cited_title":"A.: Sobolev Spaces, Academic Press, Inc., Lond on, 1975","cited_arxiv_id":null,"evidence_quote":"It supplies the classical density of compactly supported continuous functions in $L^1$ used in the $\\delta=n$ case of Theorem 5.9."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It constructs the modified von Koch snowflake with prescribed Hausdorff dimension used in Example 6.2 to show sharpness."}],"review_version":1}