{"id":"082b7c89-a759-4c30-a1fb-9a45bedb37e6","arxiv_id":"2412.06346","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Fractional generalized Sobolev-Orlicz spaces associated with the Riesz fractional gradient are introduced, with embedding, compactness, and continuity-in-s results, plus an application to a quasilinear Dirichlet problem.","lead":"This paper builds a new family of function spaces, fractional generalized Sobolev-Orlicz spaces, built around the Riesz fractional gradient, and proves embedding, compactness, and parameter-continuity results for them. The work is aimed at giving PDE theorists a nonlocal analogue of the classical Sobolev-Orlicz framework for nonlinear problems.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The framework depends on [HH19] results proved under the original (A2); the paper's switch to the revised [HHS23] (A2) is asserted, not proved, so Theorems 1–5 and 7 are conditional on an unverified compatibility.","rationale":"I read the paper as proposing a complete functional-analytic framework: new spaces, embeddings, continuity in s, compactness, and a PDE application. The results are structured and many proofs are plausible. The weakest point is not an internal error but the unproved transfer of [HH19] theorems to the corrected (A2). This matches the reader's weakest_assumption. I do not see a reason to reject: the definitions are consistent and the examples are standard. The check would settle whether the main theorems hold for the intended class. Since the reader already assigned CONDITIONAL with low confidence, my read does not move the verdict. I credit the paper for explicitly flagging the (A2) revision; that self-identified issue is precisely what needs external verification.","tokens_in":20050,"tokens_out":13979,"duration_ms":137135,"concrete_test":"Locate in [HHS23] the precise statements that replace [HH19, Cor. 5.4.5] and [HH19, Thm. 4.4.3], and verify that their hypotheses match the revised (A2) exactly as stated in Definition 2. Then take a Φ-function A satisfying (A0), (A1) and the revised (A2) but not the original HH19 (A2) (e.g., one constructed in HHS23 if available), choose f supported in a ball with modular bounded, and compute the two estimates: ‖I_s f‖_{L^B} for the pair in Theorem 1, and sup_k ‖T_k f‖_{L^A(B_{3R})}. If either estimate fails on such an example, Theorems 1 and 2 lose their stated hypotheses and the central framework would need a revised proof.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The single most load-bearing assumption is that the corrected (A2) from [HHS23] is compatible with every [HH19] theorem invoked. Definition 2 explicitly adopts the revised (A2), and Remark 2 states that the [HH19] results used 'hold true' for this definition, but no proof or precise [HHS23] theorem is supplied. In particular: Theorem 1 (Sobolev embedding) applies [HH19, Cor. 5.4.5] for the Riesz potential I_s on L^A; Theorem 2 (Poincaré) applies [HH19, Thm. 4.4.3] for uniform L^A-boundedness of dyadic operators T_k; Theorem 3, Proposition 2 and Theorem 5 reuse these bounds; Theorem 4 uses the same harmonic-analysis toolkit; and Theorem 7 depends on Theorems 2, 3, 5. If the revised (A2) does not restore these results, the embedding and compactness theorems have no proof, and the PDE application collapses. This is not an internal contradiction: the definitions are coherent. It is an unverified external dependency at the base of the proof chain, and no example in the paper is checked against the revised condition beyond citing [HH19] results that were formulated for the old (A2).","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a new family of function spaces, Λ^{s,A}_0(Ω), defined as the completion of C_c^∞(Ω) with respect to the norm ‖u‖_{L^A(R^d)} + ‖D^s u‖_{L^A(R^d;R^d)}, where A is a generalized Φ-function satisfying growth conditions (Inc)_p and (Dec)_q and the technical hypotheses (A0), (A1), (A2). The main results are: a Sobolev embedding (Theorem 1), a fractional Poincaré inequality (Theorem 2), continuity of D^s in s (Theorem 3), an interpolation inequality (Proposition 2), compact embeddings (Theorems 4 and 5), and an application to existence, uniqueness and continuous dependence of solutions to a quasilinear Dirichlet problem driven by the Riesz fractional gradient (Theorems 6 and 7). The proofs rely on harmonic-analysis results for generalized Orlicz spaces from [HH19] and [HHS23].","tokens_in":20323,"tokens_out":8106,"duration_ms":79460,"significance":"If the main results are correct, the paper provides a useful functional-analytic framework that extends Lions–Calderón spaces Λ^{s,p}_0 to a generalized Orlicz setting with the Riesz fractional gradient. The structure is clear and the intended application to nonlocal problems with nonstandard growth is natural. Among the strengths: the use of the fractional fundamental theorem and the Riesz transform to reduce fractional-gradient estimates to classical Riesz-potential estimates; an explicit Poincaré constant depending on s in a transparent way; the construction of an interpolation inequality; and a complete variational existence and uniqueness argument for the PDE application. However, the proof chain is heavily dependent on an unverified compatibility between the revised condition (A2) taken from [HHS23] and the theorems from [HH19] that are invoked throughout. Several estimates, especially the far-field term in Theorem 3 and the interpolation step in Proposition 2, are asserted rather than proved in detail. These are load-bearing issues, not mere presentation problems.","major_comments":[{"comment":"The paper explicitly adopts the revised condition (A2) from [HHS23], while most harmonic-analysis results are quoted from [HH19], which were formulated with the original (A2). Remark 2 states that the [HH19] results used 'hold true' for the revised definition, but no proof or precise statement from [HHS23] is supplied. This is load-bearing: Theorem 1 invokes [HH19, Corollary 5.4.5], Theorem 2 invokes [HH19, Theorem 4.4.3], Theorem 4 invokes [HH19, Corollary 5.4.3], and Theorems 5 and 7 inherit these dependencies. Moreover, Examples 1–3 are justified by citations to [HH19], not by a check against the revised (A2). The authors must either prove that every [HH19] result used remains valid under the revised (A2), or replace each invocation by the corresponding theorem from [HHS23], and verify the examples under the revised condition.","section":"Definition 2, Remark 2"},{"comment":"The proof of the far-field bound in Theorem 3 is incomplete. The passage from ‖∫_Ω |u(y)|/|x−y|^{d+σ} dy‖_{L^A((Ω')^c)} to ∫_Ω |u(y)| ‖ |x−y|^{-(d+σ)} ‖_{L^A((Ω')^c)} dy is asserted with the phrase 'Jensen's inequality, Lemma 2, and [HH19, Lemma 3.7.7]', but no justification is given for moving the Luxemburg norm inside the integral or for replacing the L^A norm of the kernel by the maximum of its L^p and L^q norms. The displayed estimate max{(C/((d+σ)p−d))^{1/p}, (C/((d+σ)q−d))^{1/q}} ≤ max{(C/σ)^{1/p}, (C/σ)^{1/q}} ≤ max{1, C/σ} also omits the computation of the relevant norms. Since Theorem 3 is used in Theorem 5 and Theorem 7, this gap needs to be closed with a detailed derivation.","section":"Theorem 3, equations (13)–(16)"},{"comment":"The proof of the interpolation inequality is not complete. The step ‖u + (−Δ)^{t/2}u‖_{L^A} ≤ C‖u‖_{L^A}^{(t−s)/t}‖(−Δ)^{t/2}u‖_{L^A}^{s/t} is justified only by 'an argument where we optimize the dilation of u', which is not carried out. In the following displayed inequality the exponent on ‖D^s u‖ appears as s/σ, which should presumably be σ/s for interpolation between 0 and s; no derivation is given. The relation D^σ u = R(−Δ)^{σ/2}u is also used without stating the relevant boundedness result for R. As (21) is a key tool for Propositions 3–4 and for Theorems 5 and 7, the proof must be supplied in full.","section":"Proposition 2, equation (21)"},{"comment":"The statement of Theorem 5 allows σ to be an arbitrary limit point in [0,1], but the proof requires choosing 0 < s_* < 1 with s_* < s_n for every n; this is impossible when s_n → 0. If the theorem is intended only for σ > 0, that assumption must be stated, since the later application in Theorem 7 is for σ ∈ (0,1]. In the identification step, after taking limits one obtains ∫ v·φ = −∫ u div^σ φ, not 'div^{s_n} φ' as written; the displayed formula appears to contain a typo that obscures the argument.","section":"Theorem 5 and proof"}],"minor_comments":[{"comment":"The conclusion states 'D^{s_n}u_n ⇀ Du in L^A(R^d)', but the limiting object should be D^σ u. Please correct this typo.","section":"Theorem 5, statement"},{"comment":"Equation (24) has an unbalanced parenthesis in 'a(x,D^{s_n}u_n)D^{s_n}u_n · D^{s_n}w) dx'. In the limiting equation, 'D^σu_n' should read 'D^σu' and 'D^sσ w' should read 'D^σ w'.","section":"Theorem 7, equation (24) and proof"},{"comment":"The notation 'L^A(R^d; R^s)' appears in the sentence 'D^{s_n}g_n → D^σu in L^A(R^d; R^s)'; the target space should be R^d. Also, the construction of the sequence {u_n} would benefit from a clearer definition of the indices q(k).","section":"Proposition 4, proof"},{"comment":"In the logarithmic perturbation example, the function A is written as ℓ^{p(x)} log(e+r), but the variable r is not defined; it should be ℓ. This makes the example hard to parse.","section":"Example 2"},{"comment":"In the reference [HHS23], the author name appears corrupted as 'Artur S/suppress labuszewski'; it should be corrected (presumably Artur Słabuszewski).","section":"References"},{"comment":"The equality 'A(x,r)=Cr^p' when p=q requires also the normalization A(x,1) constant in x; as stated, the conclusion is immediate only with the additional condition (A0). This may be worth clarifying.","section":"Remark 3"}],"recommendation":"major_revision","confidential_remarks":"The main risk is external dependency: the paper's central results rest on the assertion that the revised (A2) from [HHS23] is compatible with the [HH19] theorems used. This is not an internal contradiction, and the program is credible, but the compatibility must be verified explicitly. I would not reject the paper outright; a careful revision that supplies the missing harmonic-analysis verification and completes the proofs of Theorem 3 and Proposition 2 could make the contribution solid. The paper also has several typographical issues that should be corrected."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, quick take on arXiv:2412.06346. The paper introduces Lambda^{s,A}_0(Omega), fractional generalized Sobolev-Orlicz spaces based on the Riesz fractional gradient. That is genuinely new and it puts together a useful package: Sobolev and Poincare inequalities, interpolation, continuity in s, compact embeddings, plus existence/uniqueness/continuous dependence for a quasilinear PDE. I think the reader's assessment is about right: conditional, not because of a discovered counterexample, but because the proof chain has an external dependency that is asserted rather than verified.\n\nThe main soft spot is the (A2) condition. Definition 2 adopts the revised (A2) from HHS23, while several of the cited harmonic-analysis results, notably [HH19, Cor 5.4.5] for the Riesz potential and [HH19, Thm 4.4.3] for dyadic operators, were proved under the old (A2). Remark 2 says the results 'hold true' for the new definition, but no proof or precise reference is given. That is load-bearing: Theorems 1, 2, 3, 5, and 7 all sit on those bounds. The paper's examples (variable exponent, double phase, log perturbation) all come from HH19/HHS23 and are likely fine, but the general compatibility claim needs to be settled. This is fixable, but it has to be fixed before the framework is fully dependable.\n\nOne thing I'd push back on: the reader flags Theorem 3's far-field constant as asserted without computation. I actually think the computation is there in the proof, via Lemma 2 bounding the L^A norm by the Lp/Lq norms of 1/|x|^{d+sigma} and then the explicit integral bound. It is a bit quick, but it is not missing. So that specific worry is lighter than it looks.\n\nThe PDE section is standard monotonicity plus the compactness machinery, and it seems fine. I don't see a hidden contradiction in the paper; it is coherent on its own terms. The heavy reliance on the author's prior work and the same small community is not by itself a problem, since the results are genuinely building on that work.\n\nWho this is for: anyone working on nonlocal equations with nonstandard growth, especially in peridynamics-type models. It gives a ready-made functional-analytic toolbox. A serious referee should be engaged, mainly to force the (A2) compatibility issue into the open and to ask for the missing verifications. I would accept it for review, with an expectation of revision. Not a desk reject.","headline":"A genuinely new fractional Sobolev-Orlicz framework worth refereeing, but the proof chain rests on an unverified compatibility between the revised (A2) condition and the [HH19] results it invokes.","tokens_in":20942,"tokens_out":2670,"would_cite":true,"duration_ms":26295,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46E35","46E30","35J60","35R11"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper builds fractional Sobolev-Orlicz spaces around the Riesz fractional gradient and proves they support a full PDE theory: embeddings, Poincaré inequality, compactness, and existence, uniqueness, and continuous dependence of…","keywords":["fractional Sobolev-Orlicz spaces","Riesz fractional gradient","generalized Orlicz spaces","Sobolev embedding","Poincaré inequality","compact embedding","quasilinear PDE","continuous dependence"],"falsifier":"Take a double-phase $\\Phi$-function $A(x,\\ell)=\\ell^p+\\alpha(x)\\ell^q$ with $\\alpha\\in L^\\infty$ but $\\alpha\\notin C^{d/(q-p)}$, and check numerically or analytically whether the compact embedding $\\Lambda^{s,A}_0(\\Omega)\\hookrightarrow L^A(\\Omega)$ still holds for some $s\\in(0,1)$; a failure would show that $(A1)$ is load-bearing. Equivalently, look for a $\\Phi$-function satisfying $(\\mathrm{Inc})_p$ and $(\\mathrm{Dec})_q$ but violating $(A2)$ on an unbounded domain for which the Poincaré inequality of Theorem 2 breaks down.","tokens_in":19759,"feed_emoji":"🧮","tokens_out":7978,"duration_ms":63808,"temperature":0.7,"pith_summary":"This paper constructs a new family of function spaces, the fractional generalized Sobolev-Orlicz spaces $\\Lambda^{s,A}_0(\\Omega)$, obtained by completing test functions with respect to a norm that measures a function and its Riesz fractional gradient $D^s u$ in a generalized Orlicz space. The goal is to give nonlinear PDEs that involve $D^s$—natural models for nonlocal elasticity—the same functional-analytic toolbox that classical Sobolev spaces give local problems. The paper proves the fractional Sobolev embedding and Poincaré inequality, continuity of $D^s$ in the fractional parameter $s$, an interpolation inequality, and two compact embedding theorems. These tools yield existence and uniqueness of weak solutions to the quasilinear Dirichlet problem $-D^s\\cdot(a(x,D^s u)D^s u)=F$, and continuous dependence of solutions as $s\\to\\sigma$. If the framework holds, it extends the Lions–Calderón spaces $\\Lambda^{s,p}_0$ to problems with non-uniform, position-dependent growth.","feed_headline":"New fractional spaces give PDEs with Riesz gradients a full toolbox","feed_subtitle":"From Sobolev and Poincaré inequalities to compactness and continuous dependence in s, all in one Banach-space framework.","key_machinery":"The carrying mechanism is the Riesz fractional gradient $D^s u=D(I^{1-s}u)$, which is, up to a constant, the unique rotationally and translationally invariant operator of order $s$ mapping scalars to vectors and generating the fractional Laplacian through $-D^s\\cdot D^s u=(-\\Delta)^s u$. The argument combines the fractional fundamental theorem $u=I_s(R\\cdot D^s u)$ with boundedness of the Riesz potential, the Riesz transform $R$, and Calderón–Zygmund singular integrals on generalized Orlicz spaces $L^A$. The generalized $\\Phi$-function $A(x,\\ell)$ encodes position-dependent, non-power growth; the conditions $(\\mathrm{Inc})_p$ and $(\\mathrm{Dec})_q$ control its growth, while $(A0)$, $(A1)$, $(A2)$ make the harmonic-analysis estimates available.","core_discovery":"The central claim is that $\\Lambda^{s,A}_0(\\Omega)$ is a complete functional-analytic setting for PDEs built on the Riesz fractional gradient, in the same way that $W^{1,A}_0$ serves local problems. Under growth conditions $(\\mathrm{Inc})_p$ and $(\\mathrm{Dec})_q$ on the generalized $\\Phi$-function $A$, together with three technical assumptions $(A0)$, $(A1)$ and $(A2)$, the paper establishes a Sobolev embedding into a companion Orlicz space, a Poincaré inequality with constant $C/(1-2^{-s})$, continuity of $D^s$ as $s\\to\\sigma$, a Gagliardo–Nirenberg-type interpolation inequality, and two compactness theorems. On this basis it proves that the quasilinear problem (23) has a unique weak solution for every $F\\in\\Lambda^{-s,A'}(\\Omega)$, and that solutions depend continuously on $s$ as $s\\to\\sigma\\in(0,1]$.","pith_inferences":["The same machinery should extend to obstacle problems and variational inequalities in the Orlicz setting, as the author notes in Remark 6; the monotonicity and compactness tools already point in that direction.","The explicit $s$-dependence of the constants (such as $1/(1-2^{-s})$) suggests quantitative bounds on how fast solutions change as $s\\to 0$ or $s\\to 1$, which could be tested computationally.","Relaxing $(A1)$ and $(A2)$ would likely require a weighted generalized Orlicz theory, analogous to the role of Muckenhoupt weights in the variable-exponent case."],"forward_implications":["The Sobolev embedding $\\Lambda^{s,A}_0(\\Omega)\\subset L^B(\\Omega)$ controls lower-integrability Orlicz norms by the fractional gradient, exactly as in the classical case.","The Poincaré inequality makes $\\|D^s u\\|_{L^A}$ an equivalent norm on $\\Lambda^{s,A}_0$, with a constant that stays uniform in $s$ apart from the factor $1/(1-2^{-s})$.","The compact embeddings give the precompactness needed for direct-method existence proofs and for extracting convergent subsequences when $s$ varies.","The interpolation inequality and continuity of $D^s$ in $s$ permit passing to limits $s_n\\to\\sigma$, which is what makes continuous dependence of PDE solutions possible.","The quasilinear problem (23) has a unique weak solution, and the solution map is continuous in the fractional parameter $s$."],"supporting_citations":[{"why":"introduces the Riesz fractional gradient and the fractional fundamental theorem that underlies every proof.","marker":"[SS15]"},{"why":"supplies the generalized Orlicz-space harmonic analysis (Riesz potential and Calderón–Zygmund boundedness, dyadic operator estimates).","marker":"[HH19]"},{"why":"gives the corrected (A2) condition used to apply harmonic analysis on unbounded domains.","marker":"[HHS23]"},{"why":"the previous existence and continuous-dependence results in Λ^{s,p} that the PDE section extends.","marker":"[CR23]"},{"why":"provides the p-case interpolation and continuity in s that Propositions 2 and 3 generalize.","marker":"[BCCS22]"},{"why":"supplies the Riesz-potential splitting used in the Poincaré and continuity proofs.","marker":"[BCMC21]"},{"why":"provides the strict-monotonicity argument used for uniqueness in Theorem 6.","marker":"[LR24]"},{"why":"states the fractional fundamental theorem of calculus, identity (9).","marker":"[Pon16]"},{"why":"gives the dyadic decomposition lemma used in the Poincaré inequality proof.","marker":"[DHHR11]"},{"why":"proves the Young-conjugate estimate used to bound a(x,|ξ|)ξ in L^{A'}.","marker":"[FBS19]"}],"fun_headline_variants":["New spaces tame Riesz fractional gradient PDEs","Fractional gradient PDEs get unique solutions in new spaces","Riesz fractional gradient spaces solve quasilinear PDEs","Continuous dependence on fractional order for Riesz-gradient PDEs","New Orlicz spaces for Riesz fractional gradients yield well-posed PDEs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the generalized $\\Phi$-function $A$ satisfies the abstract technical conditions $(A1)$ and $(A2)$, which guarantee that $A$ and its inverse vary in a controlled pointwise way across the domain; the paper assumes these conditions rather than deriving them from simpler growth assumptions, and they are needed for the Riesz-potential and Calderón–Zygmund bounds that power the Sobolev, Poincaré, and compactness theorems.","fun_headline_variants_meta":{"raw":{"variants":["New spaces tame Riesz fractional gradient PDEs","Fractional gradient PDEs get unique solutions in new spaces","Riesz fractional gradient spaces solve quasilinear PDEs","Continuous dependence on fractional order for Riesz-gradient PDEs","New Orlicz spaces for Riesz fractional gradients yield well-posed PDEs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001414,"raw_usage":{"total_tokens":5709,"prompt_tokens":943,"completion_tokens":4766,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":559,"completion_tokens_details":{"reasoning_tokens":4683}},"tokens_in":559,"tokens_out":4766,"duration_ms":35847,"temperature":1.0,"reasoning_tokens":4683,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:44:23.469926+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a double-phase $\\Phi$-function $A(x,\\ell)=\\ell^p+\\alpha(x)\\ell^q$ with $\\alpha\\in L^\\infty$ but $\\alpha\\notin C^{d/(q-p)}$, and check numerically or analytically whether the compact embedding $\\Lambda^{s,A}_0(\\Omega)\\hookrightarrow L^A(\\Omega)$ still holds for some $s\\in(0,1)$; a failure would show that $(A1)$ is load-bearing. Equivalently, look for a $\\Phi$-function satisfying $(\\mathrm{Inc})_p$ and $(\\mathrm{Dec})_q$ but violating $(A2)$ on an unbounded domain for which the Poincaré inequality of Theorem 2 breaks down.","supporting_citations":[],"review_version":1}