{"id":"cbdec437-e88f-4df2-a8b0-975c91e567da","arxiv_id":"2411.14830","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors establish Lp × Lq to Lr estimates for bilinear fractional integrals along general curves with suitable curvature, plus an almost sharp Hardy-Littlewood-Sobolev inequality for the related fractional integral operator.","lead":"This mathematics paper proves boundedness estimates for bilinear fractional integral operators that integrate two functions along a curve, for a broad class of curves satisfying curvature and growth conditions. The result generalizes earlier work on monomial curves and includes an almost sharp Hardy-Littlewood-Sobolev inequality for the associated fractional integral operator.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"When ω0,2=ω∞,1 the HLS interval in Lemma 2.2 is empty, so Proposition 2.4 supplies no restricted weak-type estimates at the C/D vertices; the stated theorems therefore have a genuine proof gap in this equality subcase.","rationale":"The reader's weakest-assumption analysis identifies the same load-bearing gap: Lemma 2.2 proves the HLS inequality on an open exponent interval, while the theorem's proof needs the equality case ω0,2=ω∞,1. The strict-inequality case is likely correct because the open convex hull can be interpolated from interior M-values, but when the interval is empty no restricted weak-type estimate is available at the C/D vertices. The equality case is not exotic: monomials and slowly deformed powers satisfy (H) with all exponents equal, and Remark 2.3 only handles exact power-like behavior. The example list in Example 1.4(3) is also inaccurate, since t^β log(1+t) and t^β arctan t have ω0,2=β+1 and ω∞,1=β, violating ω∞,1≥ω0,2, which reinforces the class/example mismatch but is secondary. No counterexample to the theorem itself has been established here, so the conditional verdict is appropriate: the paper's main line is credible for strictly separated exponents, but the equality subcase needs either a new endpoint HLS proof or an explicit additional hypothesis.","tokens_in":33854,"tokens_out":25277,"duration_ms":255995,"concrete_test":"Derive the change-of-variables estimate in Lemma 2.2 for the admissible curve γ(t)=t^M/(1+log(1+1/t)) at M=ω0,2=ω∞,1. The proof needs t^α≲γ(t)^{α/M}; in fact t^α/γ(t)^{α/M}=(1+log(1+1/t))^{α/M}→∞ as t→0, so the displayed reduction to the HLS kernel fails. Then test the resulting log-weighted kernel by a weighted Hardy-Littlewood-Sobolev criterion or by truncation test functions f_N(y)=y^{-1/p}χ_{(2^{-N},1)}. If the log-weighted HLS estimate fails, Theorem 1.1 is false for this admissible curve; if it holds, Lemma 2.2 still needs a new closed-interval argument. Either way, the manuscript's current proof of the equality case is incomplete.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorems 1.1 and 1.2 depends on Proposition 2.4 for restricted weak-type estimates at the vertices C1, C2, D1, D2, and these estimates are derived from Lemma 2.2. Lemma 2.2 establishes the Hardy-Littlewood-Sobolev inequality for I_{α,γ} only when 1/p−1/q=α/M with M in the open interval (ω0,2,ω∞,1). When ω0,2<ω∞,1 this is sufficient because the theorems claim only the open convex hull, so one can interpolate from M-values arbitrarily close to the endpoints. The load-bearing case is ω0,2=ω∞,1: the open interval is then empty, Proposition 2.4 supplies no estimate for the C/D vertices at all, and the interpolation argument cannot begin. This is not a corner case: it includes every monomial γ(t)=t^M and admissible non-power-like curves such as γ(t)=t^M/(1+log(1+1/t)), which satisfy (H) with all four exponents equal to M. Remark 2.3 covers the closed interval only for curves that are exactly comparable to powers near zero and infinity, so it does not repair the gap. Unless Lemma 2.2 is extended to M=ω0,2=ω∞,1 for the full class, or the theorems are restricted by an explicit strict-separation hypothesis, the stated range is not proved.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper establishes L^p(R) x L^q(R) -> L^r(R) estimates for the bilinear fractional integral operator H_{α,γ} along a general plane curve (t, γ(t)) under a set of curvature hypotheses (H), and an 'almost sharp' Hardy-Littlewood-Sobolev inequality for the associated fractional integral I_{α,γ}. The main theorems describe the boundedness region as the open convex hull of explicit points determined by the logarithmic growth exponents ω_{0,1}, ω_{0,2}, ω_{∞,1}, ω_{∞,2}. The proof combines restricted weak-type estimates at p=∞ or q=∞, uniform estimates for dyadic pieces H_{α,γ,j}, and interpolation, with the most technical parts being the endpoint p=1 and q=1 cases.","tokens_in":34133,"tokens_out":4822,"duration_ms":44265,"significance":"If correct, the results unify and extend prior work on monomials and finite-type curves to a broad class of curves satisfying curvature and growth conditions. The explicit description of the boundedness range in terms of four exponents is a useful contribution. The paper is largely self-contained in its dyadic and interpolation arguments, and it does not rely on circular reasoning; cited prior results are used as tools. The main theorems are accompanied by concrete examples showing the scope of the hypotheses. The gap identified in this report concerns a substantial equality subcase, so the central framework is plausible but not fully established as stated.","major_comments":[{"comment":"Lemma 2.2 establishes the Hardy-Littlewood-Sobolev inequality for I_{α,γ} only for exponents M in the open interval (ω_{0,2}, ω_{∞,1}) (Eq. (2.4)). Proposition 2.4 then derives restricted weak-type estimates at the C/D vertices from this lemma, with M also in (ω_{0,2}, ω_{∞,1}). When ω_{0,2}=ω_{∞,1}, this interval is empty, so Proposition 2.4 supplies no restricted weak-type estimates for the vertices C1, C2, D1, D2. Since Hypothesis (H) only requires ω_{∞,1}≥ω_{0,2}, equality is allowed; it occurs for monomials γ(t)=t^β and for curves such as γ(t)=t^β/(1+log(1+t)) that satisfy (H) with all four exponents equal to β but are not comparable to a power near infinity. Remark 2.3 covers the closed interval only for curves that are exactly comparable to powers near both zero and infinity, so it does not fill this gap. Consequently, the proofs of Theorems 1.1 and 1.2 do not establish the stated boundedness ranges in the equality case ω_{0,2}=ω_{∞,1}.","section":"§2.1, Lemma 2.2 and §2.2, Proposition 2.4"},{"comment":"Because the restricted weak-type estimates at the vertices C1, C2, D1, D2 are load-bearing for the interpolation argument in the proofs of Theorems 1.1 and 1.2, the theorems as stated are broader than the proof supports. The authors should either (a) extend Lemma 2.2 to include the endpoint M=ω_{0,2}=ω_{∞,1} for all γ satisfying (H), or (b) add an explicit strict-separation hypothesis such as ω_{0,2}<ω_{∞,1} to Theorems 1.1 and 1.2, or (c) provide an alternative derivation of the C/D restricted weak-type estimates in the equality case. Without one of these, the claimed open-convex-hull regions are not proved for a substantial family of admissible curves.","section":"Theorems 1.1 and 1.2"}],"minor_comments":[{"comment":"In the interpolation for the critical cases, the displayed estimate uses θ1 instead of θ2; this appears to be a typo since the surrounding notation uses θ2 throughout Section 5.","section":"Section 5, after Eq. (5.2)"},{"comment":"There are several typographical errors, e.g., 'founded' for 'found' in the Introduction, 'Cauchy-Schwartz' for 'Cauchy-Schwarz' in the proof of Proposition 3.2, and inconsistent rendering of displayed fractions in the text. These do not affect the mathematics but should be corrected.","section":"Throughout"},{"comment":"Some references list page ranges without article numbers (e.g., [26], [27]) and the bibliography style is inconsistent; a uniform format would improve readability.","section":"Reference formatting"}],"recommendation":"major_revision","confidential_remarks":"The equality case ω_{0,2}=ω_{∞,1} is not a negligible corner: it includes every monomial curve and also non-power-like curves satisfying (H). The authors may be able to fix this by observing that for monomials the required endpoint estimate is already in Remark 2.3, but the gap remains for general equality-case curves. This is a load-bearing issue, so the paper should not be accepted without either extending Lemma 2.2 or restricting the theorems. I recommend major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you the short version: this is a serious, technically heavy piece that genuinely extends the Li–Liu monomial result to a broader class of curves with distinct logarithmic growth exponents, and it also proves an almost sharp HLS inequality for the associated fractional integral. But the theorems as stated are not fully proved. When ω0,2 = ω∞,1, Lemma 2.2's HLS estimate only holds for M in the open interval (ω0,2, ω∞,1), which is empty; Proposition 2.4 then gives no restricted weak-type estimates at the C/D vertices. Remark 2.3 only covers curves exactly comparable to powers near 0 and ∞, which excludes natural examples like t^β log(1+t). This is not a corner case: monomials and many finite-type curves fall into it.\n\nWhat is genuinely new: the boundedness region for H_{α,γ} under the curvature hypothesis (H) when the exponents are separated, and the HLS inequality with explicit dependence on the ω's. The dyadic decomposition and interpolation arguments are careful; the stationary-phase treatment of the critical scale is a real piece of work. The paper is honest about its tools, and the self-citations are prior results used as machinery, not circular dependencies.\n\nThe soft spot is exactly where the reader flagged. The proof simply does not establish the vertex estimates for general curves with ω0,2 = ω∞,1. This can be fixed by adding an explicit strict-separation hypothesis, or by extending the endpoint HLS to the full class; the latter might require a genuinely new idea. Also, Example 1.4(3) lists curves that do not satisfy (H), so the claim that their regions follow is at least misleading.\n\nOverall, the main body for the generic case is probably right and is a worthwhile contribution to the bilinear fractional integral literature, but the theorem statements overreach. I would send it to a competent referee with the expectation of major revision. The reader's CONDITIONAL verdict is fair; I lean slightly toward 'needs work but likely sound in the strict case.'","headline":"Solid technical extension to general curves, but the stated theorems overclaim: the proof does not cover the equality case ω0,2 = ω∞,1 for non-power-like curves.","tokens_in":34689,"tokens_out":3940,"would_cite":true,"duration_ms":38632,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B15","42A85"],"pacs":[],"model":"deepseek-v4-flash","headline":"For curves satisfying a curvature condition, the bilinear fractional integral $H_{\\alpha,\\gamma}$ maps $L^p\\times L^q$ to $L^r$ on a convex hull set by the curve's growth exponents.","keywords":["bilinear fractional integral","fractional integrals along curves","Hardy–Littlewood–Sobolev inequality","restricted weak type estimate","bilinear interpolation","curvature conditions","logarithmic growth exponents"],"falsifier":"Take a curve satisfying (H) with $\\omega_{0,2}=\\omega_{\\infty,1}$ but with logarithmic oscillation, such as $\\gamma(t)=t^\\beta(1+\\varepsilon\\sin(\\log\\log(1/t)))$ near $0$ and a matched oscillation near $\\infty$, and compute $\\sup_{A,B}|B|^{-\\alpha/M}\\|H_{\\alpha,\\gamma}(\\chi_A,\\chi_B)\\|_\\infty$ at $M=\\omega_{0,2}=\\omega_{\\infty,1}$; an infinite supremum would show that the restricted weak-type vertex feeding the interpolation argument fails for that curve.","tokens_in":33568,"feed_emoji":"📐","tokens_out":11689,"duration_ms":103033,"temperature":0.7,"pith_summary":"The paper aims to pin down the boundedness of bilinear fractional integrals taken along a curved path: for a broad class of plane curves $(t,\\gamma(t))$ with controlled curvature, it claims that $H_{\\alpha,\\gamma}(f,g)(x)=\\int_0^\\infty f(x-t)g(x-\\gamma(t))\\,dt/t^{1-\\alpha}$ sends $L^p(\\mathbb R)\\times L^q(\\mathbb R)$ into $L^r(\\mathbb R)$ for a whole polyhedral region of exponent triples. The region is described as the open convex hull of explicit vertices whose coordinates are determined by four logarithmic growth rates of $\\gamma$ near zero and infinity. If correct, this unifies and extends the monomial case $\\gamma(t)=t^\\beta$ to polynomials without linear or constant terms and to finite-type curves, and it also yields an almost sharp Hardy–Littlewood–Sobolev inequality for the one-dimensional fractional integral along the same curve.","feed_headline":"Curved-path fractional integrals get explicit Lp bounds","feed_subtitle":"For monomial, polynomial, and finite-type curves, the bounded range is set by growth at zero and infinity.","key_machinery":"The load-bearing mechanism is the one-dimensional curve-adapted fractional integral $I_{\\alpha,\\gamma}f(x)=\\int_0^\\infty |f(x-\\gamma(t))|\\,dt/t^{1-\\alpha}$: its Hardy–Littlewood–Sobolev estimates supply the restricted weak-type bounds at the $p=\\infty$ and $q=\\infty$ vertices of the claimed region. The bilinear operator is decomposed into dyadic pieces $H_{\\alpha,\\gamma,j}$ on the annulus $[2^{j-1},2^j]$, and each piece is controlled by a change of variables whose Jacobian is $1-\\gamma'(t)$, separating the non-critical scales from the at-most-three critical scales near a stationary point $\\gamma'(t_0)=1$. The four logarithmic growth exponents $\\omega_{0,1},\\omega_{0,2},\\omega_{\\infty,1},\\omega_{\\infty,2}$ enter through Lemma 2.1, which confines $|\\gamma(t)|$ between nearby powers of $t$ near zero and infinity; these exponents determine every vertex in the convex hull. Real interpolation of the endpoint restricted weak-type estimates produces the strong-type bounds in the interior of the hull.","core_discovery":"On the paper's own terms, the central claim is that Hypothesis (H) — monotonicity, $\\gamma(0)=\\gamma'(0)=0$, the bounds on $t\\gamma'(t)/\\gamma(t)$ and $t^2\\gamma''(t)/\\gamma(t)$, and $\\omega_{\\infty,1}\\ge\\omega_{0,2}$, $\\omega_{0,1}>1$ — suffices for $H_{\\alpha,\\gamma}$ to be bounded from $L^p(\\mathbb R)\\times L^q(\\mathbb R)$ to $L^r(\\mathbb R)$ for every reciprocal triple in the open convex hull of explicit points. Theorem 1.1 does this when $\\frac{\\omega_{\\infty,2}-\\omega_{\\infty,1}}{\\omega_{\\infty,2}-1}<\\frac{\\alpha}{1-\\alpha}<\\frac{\\omega_{\\infty,2}}{\\omega_{\\infty,2}-\\omega_{\\infty,1}+1}$, with vertices $A,B,C_1,C_2,D_1,D_2,E,F$; Theorem 1.2 covers $\\frac{\\alpha}{1-\\alpha}\\ge\\omega_{\\infty,1}$ with $G_1,G_1^2,G_2^2,H$ replacing $E,F$ in suitable subranges. The companion one-dimensional fractional integral $I_{\\alpha,\\gamma}$ is shown to satisfy an almost sharp Hardy–Littlewood–Sobolev inequality: $1/p-1/q=\\alpha/M$ with $M\\in(\\omega_{0,2},\\omega_{\\infty,1})$ guarantees $L^p\\to L^q$, and any such estimate forces $M\\in[\\omega_{0,1},\\omega_{\\infty,2}]$.","pith_inferences":["A testable extension would be to prove Lemma 2.2 with the closed interval $M\\in[\\omega_{0,2},\\omega_{\\infty,1}]$ for every curve satisfying (H), not just the power-like curves of Remark 2.3; that would justify the boundary vertices $C_1,C_2,D_1,D_2$ when $\\omega_{0,2}=\\omega_{\\infty,1}$.","The same endpoint issue suggests a concrete experiment: compute the restricted weak-type norm at $M=\\omega_{0,2}=\\omega_{\\infty,1}$ for an oscillatory curve with equal growth rates, such as $\\gamma(t)=t^\\beta(1+\\varepsilon\\sin(\\log\\log(1/t)))$ near $0$ and a matching oscillation near $\\infty$; failure there would narrow the interpolated region.","The dyadic stationary-phase decomposition is one-dimensional, but the same ratio $\\alpha/(1-\\alpha)$ versus $\\omega_{\\infty,1}$ should govern similar operators built from several curves, where the boundedness region would be a higher-dimensional polytope.","A natural next step is to prove necessity for the bilinear region: the dilation argument used in Lemma 2.2, applied to both variables of $H_{\\alpha,\\gamma}$, should show that no exponent triple outside the closure of the hull can be bounded."],"forward_implications":["For $\\gamma(t)=t^\\beta$, the region in Theorems 1.1 and 1.2 reduces to the open convex hull of $A,B,C,D,E,F$ (or $G_1,G_1^2,G_2^2,H$), reproducing the monomial case with the dependence on $\\alpha/\\beta$ made explicit.","For a polynomial $\\gamma(t)=\\sum_{i=1}^k \\beta_i t^{\\alpha_i}$ with $1<\\alpha_1<\\cdots<\\alpha_k$ and no constant or linear term, the exponents are $\\omega_{0,1}=\\omega_{0,2}=\\alpha_1$ and $\\omega_{\\infty,1}=\\omega_{\\infty,2}=\\alpha_k$, so the boundedness region is fully determined by the lowest and highest monomial degrees.","Curves of finite type at zero, such as $t^\\beta\\log(1+t)$, $t^\\beta\\arctan t$, or $e^{\\beta t}-\\beta t-1$, satisfy the hypothesis, so the same boundedness region applies to them.","The Hardy–Littlewood–Sobolev inequality for $I_{\\alpha,\\gamma}$ is almost sharp: the exponent gap $1/p-1/q$ is forced to equal $\\alpha/M$ with $M$ between the near-zero and near-infinity growth rates, and for power-like curves the condition is necessary and sufficient.","Restricted weak-type estimates hold at the vertices $A,B,C_1,C_2,D_1,D_2$ (and the $G,H$ vertices in the complementary case), and interpolation then gives the open convex hull as a strong-type region."],"supporting_citations":[{"why":"Established the monomial case $\\gamma(t)=t^\\beta$, the base example that the present theorems generalize to arbitrary curves satisfying (H).","marker":"[26]"},{"why":"Supplies the classical Hardy–Littlewood–Sobolev inequality used in Lemma 2.2 to bound the fractional integral $I_{\\alpha,\\gamma}$ near zero and infinity.","marker":"[42]"},{"why":"Provides the doubling-type estimate and exponent bounds used to derive Lemma 2.1 and to control the stationary point of $\\gamma$.","marker":"[35]"},{"why":"The source of the classical bilinear fractional integral estimates and endpoint restricted weak-type framework that the paper adapts to curved paths.","marker":"[21]"},{"why":"Supplies the multilinear interpolation and endpoint results used to convert restricted weak-type estimates into the open convex hull.","marker":"[14]"},{"why":"Gives the classical treatment of the bilinear fractional integral whose vertex structure is the model for the regions in Theorems 1.1 and 1.2.","marker":"[12]"}],"fun_headline_variants":["Bilinear fractional integrals along curves get explicit Lp bounds","Curved fractional integrals: boundedness pinned by curve shape","Almost sharp HLS inequality for curved fractional integrals","New Lp bounds for a family of curved fractional integrals","Curve hypotheses yield Lp bounds for fractional integrals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof needs an endpoint Hardy–Littlewood–Sobolev estimate for the one-dimensional operator that the paper supplies only when the curve's near-zero and near-infinity growth rates are strictly ordered; when those rates coincide and the curve is not power-like, that endpoint estimate is missing.","fun_headline_variants_meta":{"raw":{"variants":["Bilinear fractional integrals along curves get explicit Lp bounds","Curved fractional integrals: boundedness pinned by curve shape","Almost sharp HLS inequality for curved fractional integrals","New Lp bounds for a family of curved fractional integrals","Curve hypotheses yield Lp bounds for fractional integrals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000999,"raw_usage":{"total_tokens":4307,"prompt_tokens":1104,"completion_tokens":3203,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":720,"completion_tokens_details":{"reasoning_tokens":3125}},"tokens_in":720,"tokens_out":3203,"duration_ms":25323,"temperature":1.0,"reasoning_tokens":3125,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:51:52.233336+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a curve satisfying (H) with $\\omega_{0,2}=\\omega_{\\infty,1}$ but with logarithmic oscillation, such as $\\gamma(t)=t^\\beta(1+\\varepsilon\\sin(\\log\\log(1/t)))$ near $0$ and a matched oscillation near $\\infty$, and compute $\\sup_{A,B}|B|^{-\\alpha/M}\\|H_{\\alpha,\\gamma}(\\chi_A,\\chi_B)\\|_\\infty$ at $M=\\omega_{0,2}=\\omega_{\\infty,1}$; an infinite supremum would show that the restricted weak-type vertex feeding the interpolation argument fails for that curve.","supporting_citations":[{"cited_title":"Li and P","cited_arxiv_id":null,"evidence_quote":"Established the monomial case $\\gamma(t)=t^\\beta$, the base example that the present theorems generalize to arbitrary curves satisfying (H)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the classical Hardy–Littlewood–Sobolev inequality used in Lemma 2.2 to bound the fractional integral $I_{\\alpha,\\gamma}$ near zero and infinity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the doubling-type estimate and exponent bounds used to derive Lemma 2.1 and to control the stationary point of $\\gamma$."},{"cited_title":"Kenig and E.M","cited_arxiv_id":null,"evidence_quote":"The source of the classical bilinear fractional integral estimates and endpoint restricted weak-type framework that the paper adapts to curved paths."},{"cited_title":"Grafakos and N","cited_arxiv_id":null,"evidence_quote":"Supplies the multilinear interpolation and endpoint results used to convert restricted weak-type estimates into the open convex hull."},{"cited_title":"Grafakos, On multilinear fractional integrals, Studia Math","cited_arxiv_id":null,"evidence_quote":"Gives the classical treatment of the bilinear fractional integral whose vertex structure is the model for the regions in Theorems 1.1 and 1.2."}],"review_version":1}