{"id":"f044668d-2c11-4597-a6e1-1686b0c546fa","arxiv_id":"2505.16116","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Distance in Lipschitz spaces to many classical subspaces is equivalent, up to constants, to a critical threshold where a measure of large normalized differences becomes finite.","lead":"This paper proves that the distance from a function to a smoother subspace can be read off from how often its finite differences grow too fast, across scales and locations. It provides one framework that covers Sobolev, Besov, and Triebel-Lizorkin spaces, extending earlier BMO and Holder-type results.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 2.4 relies on Theorem 4.10 and Theorem 4.4 with constants that may depend on f, so the asserted f-independent equivalence in (2.6) is not established by the written argument.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing gap: Theorem 4.10 is used in the proof of Theorem 2.4 although its constants may depend on f, while the theorem asserts constants independent of f. My reading confirms this and adds that the same f-dependence enters in the opposite direction through Theorem 4.4 and Lemma 4.13, so the issue affects both sides of the equivalence. The central framework is plausible, the applications are clearly organized, and the gap appears repairable by strengthening the auxiliary inclusions (or by restating the theorem with f-dependent constants), so the reader's CONDITIONAL verdict is appropriate. I do not see an independent counterexample to the theorem, and the paper contains substantial correct-looking material (wavelet characterization Theorem 3.1, concrete lattice examples, and applications), so rejection would not be justified by this review. No change to the reader's verdict is needed.","tokens_in":78396,"tokens_out":11721,"duration_ms":109489,"concrete_test":"Re-derive the proof of Theorem 2.4 with explicit constants: apply Theorem 4.10 exactly as stated with m=m(f), R=R(f), then apply Lemma 4.14 and Definition 2.1(i), and record the dependence on m(f) and R(f). Do the same for the upper bound using δ(f,ε,ε_1) in Lemma 4.13. If the resulting constants are not shown to be bounded for the family f_A = g + A h with h∈Λ_s^X and g fixed outside Λ_s^X, the proof of Theorem 2.4 as written does not prove the asserted f-independence. A sharper check is to try to prove Theorem 4.10 with m and R independent of f; the current choice R_1 ≥ (2C∥f∥_Λs/ε)^{1/(r-s)} depends on f, so a genuinely new argument would be required.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 2.4 claims equivalence constants independent of f, but its proof in §4.3 uses inclusions whose constants are not f-independent. In the lower-bound direction, Theorem 4.10 states explicitly that m and R may depend on f; in its proof, R_1 is chosen so that C∥f∥_Λs(R_1^{-s}+R_1^{-(r-s)}) < ε/2, so R_1 and hence m depend on f and ε. Lemma 4.14 then has a constant depending on R, and iterating the shift bound in Definition 2.1(i) m times introduces an m-dependent factor. The upper-bound direction has the same defect: Theorem 4.4 provides δ depending on f, ε, and ε_1, while Lemma 4.13 has a constant depending on R/δ. Thus both inclusions in the proof of Theorem 2.4 can produce constants that grow with f. This is not a cosmetic issue: e.g., for f_A = g + A h with h ∈ Λ_s^X and dist(g,Λ_s^X) > 0, the true distance stays equal to dist(g,Λ_s^X) while ∥f_A∥_Λs grows, so the proof's f-dependent constants would have to remain bounded for the theorem to follow, and no such bound is shown.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a general framework, based on quasi-normed lattices X of function sequences and Daubechies wavelet systems, for measuring the distance in the Lipschitz space Lambda_s from a function f to the subspace Lambda_s^X. The central result, Theorem 2.4, asserts that, under Assumption I, this distance is comparable, with constants independent of f, to the sum of a difference-based Lipschitz deviation constant epsilon_X f and an infimum over thresholds controlling the level-zero wavelet coefficients. A second theorem, Theorem 2.9, gives an analogous characterization under Assumption II for endpoint cases such as F^s_{infinity,q} and B^s_{infinity,q}. The paper also proves a wavelet characterization (Theorem 3.1), derives closure criteria in terms of bad sets (Theorems 4.2 and 4.3), and applies the framework to Sobolev spaces, J^s(bmo), Besov, Triebel-Lizorkin, Besov-type, and Triebel-Lizorkin-type spaces, including new proper inclusion results.","tokens_in":78671,"tokens_out":5743,"duration_ms":53885,"significance":"If the main theorems are correct, the paper provides a substantial and unified extension of the Garnett-Jones distance formula and of the Saksman-Soler i Gibert results for J^s(bmo), covering all smoothness orders s in (0,infinity) and a wide class of subspaces. The manuscript contains a large amount of detailed technical work, including quantitative hyperbolic-metric lemmas, Whitney-type inequalities, and wavelet-difference comparisons, and it clearly identifies new corollaries for classical and type spaces. The main obstacle is the f-dependence of constants in the difference-to-wavelet inclusions used in the proofs of Theorems 2.4 and 2.9; this is a load-bearing uniformity gap that must be resolved before the stated f-independent equivalences are established.","major_comments":[{"comment":"The proof of the f-independent equivalence (2.6) uses inclusions whose constants are not f-independent. In the lower-bound direction, Theorem 4.4 supplies a positive constant delta depending on f, epsilon, and epsilon_1; Lemma 4.13 then gives a constant depending on delta, and Theorem 4.4 is applied with that delta. Consequently the displayed estimate preceding (2.6) carries an implicit constant C(f,epsilon,epsilon_1), which is incompatible with the statement that the equivalence constants are independent of f. This is not cosmetic: for f_A = g + A h with h in Lambda_s^X and dist(g,Lambda_s^X) > 0, one has dist(f_A,Lambda_s^X) = dist(g,Lambda_s^X) while ||f_A||_{Lambda_s} grows with A; no argument is given showing that the f-dependent constants stay bounded along such a family.","section":"§4.3, proof of Theorem 2.4"},{"comment":"The upper-bound direction uses Theorem 4.10, whose statement explicitly allows m and R to depend on f. In the proof of Theorem 4.10, R_1 is chosen so that C||f||_{Lambda_s}(R_1^{-s}+R_1^{-(r-s)}) < epsilon/2, so R_1, and hence m and R, depend on f and epsilon. The union over i = j-m,...,j+m and the m-fold use of the shift estimate in Definition 2.1(i) introduce an m-dependent constant, while Lemma 4.14 has a constant depending on R. Thus the chain of estimates in the proof of Theorem 2.4 does not establish the asserted f-independent constants.","section":"§4.3, second direction of Theorem 2.4"},{"comment":"The same uniformity defect is inherited by Theorem 2.9. Its proof invokes Theorem 4.10 and then Lemma 4.14 (or, under Assumption II, repeated shifts of the Carleson-type measure via Definition 2.6(iv)) with constants m and R that may depend on f. The first part of the proof also uses the f-dependent delta from Theorem 4.4. Since (2.9) asserts equivalence with constants independent of f, the written proof does not establish the stated uniformity either for the difference term or for the level-zero wavelet term.","section":"§4.4, proof of Theorem 2.9"},{"comment":"Theorem 4.4 is stated only with delta depending on f, epsilon, and epsilon_1. Because this theorem is a key input to both main difference-to-wavelet arguments, the paper should either prove a version in which delta can be chosen independently of f for suitable ratios epsilon/epsilon_1, or explicitly track and control the f-dependence through the rest of the proofs. As written, the dependency prevents the conclusion that the final equivalence constants are independent of f.","section":"Theorem 4.4"}],"minor_comments":[{"comment":"The proof cites 'Corollary 3.2(i)', but no Corollary 3.2 appears in the paper; the intended reference is likely Theorem 3.1 or Proposition 3.2.","section":"Proof of Theorem 2.4"},{"comment":"The proof of Theorem 4.2 concludes with only the wavelet condition ||{sum_{I in W^0_j(s,f,epsilon)} 1_I}_{j in Z_+}||_X < infinity, omitting the difference term that appears in statement (ii) of the theorem; this appears to be a typographical omission.","section":"Proof of Theorem 4.2"},{"comment":"After the general question about a subspace V, the displayed equivalence reads epsilon_{r,s,gamma}(f) ~ dist(f,Lambda_s)_{Lambda_s}; it should presumably be dist(f,V)_{Lambda_s}, since the surrounding discussion concerns the distance to a proper subspace V of Lambda_s.","section":"Introduction, Section 2"},{"comment":"The displayed definition of epsilon^tau_{p,q} f is identical to the L_p(l_q)-type definition used for Triebel-Lizorkin spaces and does not contain the tau-dependent supremum over cubes that appears in Theorem 5.37(ii); this is likely a typo.","section":"Theorem 5.41"},{"comment":"In part (ii), the condition 'sup_{Q in Q} 1/|Q|^tau ||[int_0^{2^{-j_Q vee 0}} 1_{S_r(s,f,epsilon)}(cdot,y) dy/y]^{1/q}||_{L^p(Q)}' is missing the '< infinity' that should conclude the displayed condition.","section":"Theorem 5.50"},{"comment":"The final sentence of the proof of Lemma 4.14 says 'which completes the proof of Lemma 4.13'; this is a copy-paste error and should refer to Lemma 4.14.","section":"Proof of Lemma 4.14"}],"recommendation":"major_revision","confidential_remarks":"The paper is very substantial and, if the uniformity gap is repairable, would be a strong contribution to the field. The central issue is not the conceptual framework but the proof of f-independent constants in Theorems 2.4 and 2.9. I recommend that the authors be asked to make the dependence in Theorems 4.4 and 4.10 explicit and either prove f-independent versions or show how the f-dependent parameters are controlled in the final equivalence. The many applications and inclusion theorems are contingent on this fix."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe main theorems are new and the framework is genuinely useful, but the proof of the central equivalence as written does not deliver the f-independent constants it claims. If the uniformity gap is fixed, this is a significant paper; as is, the central result is conditional.\n\nWhat is new: the Daubechies s-Lipschitz X-based spaces, Theorems 2.4 and 2.9, and the applications to Besov, Triebel–Lizorkin, Besov-type, Triebel–Lizorkin-type spaces, plus J^s(bmo) for all s. The wavelet characterization in Theorem 3.1 is clean and appears correct. The authors have done serious work assembling the hyperbolic metric, Whitney estimates, and wavelet–difference connections. Citations to prior work, including the Saksman–Soler i Gibert result, look fair.\n\nThe soft spot: Theorem 4.4 gives δ depending on f, and Theorem 4.10 gives m and R depending on f. Both feed directly into the proof of Theorem 2.4, whose equivalence constants are supposed to be independent of f. The lower-bound direction uses Lemma 4.13, whose constant depends on R and δ; the upper-bound direction uses Lemma 4.14, whose constant inherits dependence on R. No argument is given to make these uniform. This is not cosmetic: for f_A = g + A h with h in the subspace and dist(g, subspace) > 0, the true distance stays constant while ‖f_A‖ grows, so any f-dependent constant would need to remain bounded in a way that is not shown. The same gap affects Theorem 2.9 and the applications relying on it.\n\nMinor: the abstract states ε(f) ~ dist(...) and omits the level-zero wavelet term that appears in the actual theorems.\n\nBottom line: this deserves a serious referee, but the paper needs major revision. A proof of uniform bad-set inclusions, or a weakening of the claim to f-dependent constants, would clarify what is actually established. I would not cite the main equivalence yet; I would cite Theorem 3.1 and the framework if I needed the wavelet side.","headline":"Substantial new framework and applications, but the proof of the central f-independent distance equivalence has a load-bearing uniformity gap that needs repair before the main theorem can be trusted.","tokens_in":79289,"tokens_out":4148,"would_cite":false,"duration_ms":39215,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B35","26A16","42B25","42C40","46E35"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the distance from a Lipschitz function $f$ to a broad family of smooth subspaces is comparable, uniformly in $f$, to a critical deviation index built from the super-level sets of normalized finite differences.","keywords":["Lipschitz spaces","distance to subspaces","finite differences","Daubechies wavelets","quasi-normed lattices of function sequences","Besov spaces","Triebel–Lizorkin spaces","Carleson measures"],"falsifier":"Construct a family $f_N\\in\\Lambda_s$ whose bad-difference sets concentrate in one hyperbolic ball at scale $N$ while the corresponding wavelet coefficients sit next to that ball, and compute both sides of the claimed equivalence for a lattice $X$ such as $L^p(\\ell^q)$; if $\\varepsilon_X f_N$ and $\\mathrm{dist}(f_N,\\Lambda_s^X)$ fail to be comparable with a uniform constant as $N\\to\\infty$, the central theorem is false.","tokens_in":78154,"feed_emoji":"📐","tokens_out":13520,"duration_ms":112485,"temperature":0.7,"pith_summary":"The paper aims to turn a global approximation problem—how far a continuous function on $\\mathbb{R}^n$ is from a smooth subspace, measured in the Lipschitz norm—into a local condition on super-level sets of finite differences. It claims that for every smoothness order $s>0$ and for any subspace of the Lipschitz space $\\Lambda_s$ that can be built from Daubechies wavelets and a quasi-normed lattice of function sequences $X$, the distance from $f$ to the subspace is comparable, uniformly in $f$, to the smallest threshold $\\varepsilon$ such that the set of points and scales $(x,y)$ with $\\Delta_r f(x,y)/y^s>\\varepsilon$ has finite size in the $X$-norm. The same conclusion holds with a Carleson-type measure in place of the lattice norm for endpoint spaces such as $F^s_{\\infty,q}$ and $B^s_{\\infty,q}$. A sympathetic reader would care because this reduces a hard nonlinear distance problem to a checkable condition on normalized differences, and because it recovers the classical BMO distance formula and extends it from $J^s(bmo)$ with $s\\le 1$ to Sobolev, Besov, Triebel–Lizorkin, and their type variants for all $s>0$. The paper also characterizes, in the same terms, the closures of these subspaces inside $\\Lambda_s$.","feed_headline":"One finite-difference index governs distance in Lipschitz spaces","feed_subtitle":"The same deviation constant governs Sobolev, Besov, and Triebel–Lizorkin distances.","key_machinery":"The load-bearing object is the Lipschitz deviation constant $\\varepsilon_X f$: the infimum of all thresholds $\\varepsilon$ such that the $X$-norm of the sequence $$\\left(\\int_{$2^{{-j-1}}$}^{$2^{{-j}}$}1_{S_{r,j}(s,f,\\varepsilon)}(\\cdot,y)\\,\\frac{dy}{y}\\right)_{j\\in\\mathbb{Z}_+}$$ is finite after $u$-convexification. Its companion is the Daubechies $s$-Lipschitz $X$-based space $\\Lambda_s^X$, defined by requiring the sequence of sums of normalized wavelet coefficients over each dyadic scale to belong to $X$. The proof connects these two objects through a Whitney-type polynomial approximation lemma that bounds a wavelet coefficient by the supremum of $\\Delta_r f(x,y)/y^s$ over a tube above the cube, and through a pair of comparability theorems showing that difference bad sets and wavelet bad sets are contained in fixed hyperbolic neighborhoods of each other up to finite dyadic shifts. The Poincaré hyperbolic metric supplies the quantitative geometry, and the difference operator supplies the smoothness information that wavelets alone do not directly reveal.","core_discovery":"The central claim is stated as a two-sided comparability, $$\\operatorname{dist}(f,\\Lambda_s^X)_{\\Lambda_s}\\sim \\varepsilon_X f+\\inf\\left\\{\\varepsilon>0:\\left\\|\\left\\{\\sum_{I\\in V_0(s,f,\\varepsilon)}1_I\\delta_{0,j}\\right\\}_{j\\in\\mathbb{Z}_+}\\right\\|_X<\\infty\\right\\},$$ for every $f\\in\\Lambda_s$, with constants independent of $f$. Here $\\Lambda_s^X$ is the Daubechies $s$-Lipschitz $X$-based space, whose norm is the $X$-norm of the sequence of normalized wavelet coefficients at each dyadic scale, and $\\varepsilon_X f$ is the critical index at which the $u$-convexified $X$-norm of the sequence of measures of the difference super-level sets becomes finite; $V_0(s,f,\\varepsilon)$ records the unit dyadic cubes whose scaling-function coefficient exceeds $\\varepsilon$. The first term captures the bulk of the approximation error, while the second term is a boundary term coming from the coarsest-scale coefficients. For endpoint lattices where the dilation inequality fails, the same format holds with a Carleson-type measure $\\nu$ replacing the $X$-norm. Applied to concrete spaces, the theorem says, for instance, that the distance to $J^s(bmo)$ is comparable to the critical Carleson index of the difference bad set, and that the distance to the Sobolev space $W^{1,p}$ is comparable to an $L^p$ norm of the weighted measure of second-difference bad sets.","pith_inferences":["A direct testable extension would be to run the same two-sided comparison in a doubling metric measure space with a wavelet basis: the hyperbolic-metric steps generalize, and the only piece needing a new proof is the Whitney lemma.","If the uniformity gap in the bridge lemma can be closed, the approach would immediately imply stability of the equivalence under perturbations of $f$ in a way the current proof does not show.","The endpoint results suggest a natural question the paper does not answer: whether a logarithmic correction appears when the lattice parameter $\\tau$ reaches $1/p$, the boundary of the range covered by the Besov-type and Triebel–Lizorkin-type applications.","One could exploit the equivalence numerically by bisecting on $\\varepsilon$ and computing the $X$-norm of the dyadic super-level indicators, giving an algorithm for Lipschitz distance that never computes a wavelet coefficient."],"forward_implications":["All previously known cases become special cases: the distance to $J^s(bmo)$ is characterized for every $s>0$, not only $0<s\\le 1$, and the one-dimensional case extends to all dimensions.","For Sobolev, Besov, and Triebel–Lizorkin spaces, the distance formulas reduce to checking whether the sequence of measures of difference super-level sets lies in $\\ell^q(L^p)$ or $L^p(\\ell^q)$, which is a directly testable condition.","The same formulas identify the closures: $f$ lies in the $\\Lambda_s$-closure of the subspace exactly when both the deviation constant and the level-zero wavelet boundary term vanish.","For Besov-type and Triebel–Lizorkin-type spaces with $\\tau\\in[0,1/p)$, the same theorem gives distance and closure characterizations that the paper states are new.","Because the equivalence constants are independent of $f$, the critical index $\\varepsilon_X f$ can serve as a computable proxy for the Lipschitz distance in numerical approximation problems."],"supporting_citations":[{"why":"Provides the original distance-in-BMO-to-$L^\\infty$ formula that the new deviation-index characterization generalizes.","marker":"[20]"},{"why":"Gives the previous $J^s(bmo)$ distance theorem for $s\\in(0,1]$ that the paper extends to all $s>0$ and to many other subspaces.","marker":"[50]"},{"why":"States the one-dimensional $n=s=1$ result whose proof does not extend, establishing the prior case the paper covers.","marker":"[48]"},{"why":"Defines the Daubechies wavelet system with specified regularity and vanishing moments on which $\\Lambda_s^X$ is built.","marker":"[44]"},{"why":"Supplies the wavelet characterization of the Lipschitz space and of Besov and Triebel–Lizorkin spaces used to identify the concrete examples.","marker":"[67]"},{"why":"Provides wavelet characterizations of Besov-type and Triebel–Lizorkin-type spaces, making the type-space applications possible.","marker":"[77]"},{"why":"Contains the Fourier-analytic difference estimates used in the lemmas that compare differences at nearby points.","marker":"[23]"},{"why":"Gives the vector-valued maximal inequality used to verify the dilation assumption for the classical $L^p(\\ell^q)$ and related lattices.","marker":"[16]"},{"why":"Supplies the Whitney-type polynomial approximation estimate behind the difference-to-wavelet step.","marker":"[15]"}],"fun_headline_variants":["Critical difference index sets Lipschitz subspace distances","One index measures distance in Lipschitz spaces","Finite-difference index pins down subspace distances","Distance to Sobolev, Besov, Triebel–Lizorkin via one index"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the two-sided comparison between large wavelet coefficients and large normalized finite differences can be made with constants independent of $f$; the proof's bridge lemma only gives such a comparison with constants that may depend on $f$, and no step supplies the required uniformity.","fun_headline_variants_meta":{"raw":{"variants":["Critical difference index sets Lipschitz subspace distances","One index measures distance in Lipschitz spaces","Finite-difference index pins down subspace distances","Distance to Sobolev, Besov, Triebel–Lizorkin via one index"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000238,"raw_usage":{"total_tokens":1713,"prompt_tokens":1353,"completion_tokens":360,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":969,"completion_tokens_details":{"reasoning_tokens":291}},"tokens_in":969,"tokens_out":360,"duration_ms":3773,"temperature":1.0,"reasoning_tokens":291,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:06:58.061515+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct a family $f_N\\in\\Lambda_s$ whose bad-difference sets concentrate in one hyperbolic ball at scale $N$ while the corresponding wavelet coefficients sit next to that ball, and compute both sides of the claimed equivalence for a lattice $X$ such as $L^p(\\ell^q)$; if $\\varepsilon_X f_N$ and $\\mathrm{dist}(f_N,\\Lambda_s^X)$ fail to be comparable with a uniform constant as $N\\to\\infty$, the central theorem is false.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the original distance-in-BMO-to-$L^\\infty$ formula that the new deviation-index characterization generalizes."},{"cited_title":"Saksman and O","cited_arxiv_id":null,"evidence_quote":"Gives the previous $J^s(bmo)$ distance theorem for $s\\in(0,1]$ that the paper extends to all $s>0$ and to many other subspaces."},{"cited_title":"Grafakos, Modern Fourier Analysis, third edition, Grad","cited_arxiv_id":null,"evidence_quote":"Contains the Fourier-analytic difference estimates used in the lemmas that compare differences at nearby points."},{"cited_title":"Fefferman and E","cited_arxiv_id":null,"evidence_quote":"Gives the vector-valued maximal inequality used to verify the dilation assumption for the classical $L^p(\\ell^q)$ and related lattices."},{"cited_title":"Ditzian, Polynomial approximation inL p(S) forp∈(0,∞), Constr","cited_arxiv_id":null,"evidence_quote":"Supplies the Whitney-type polynomial approximation estimate behind the difference-to-wavelet step."}],"review_version":1}