{"id":"b9023c6f-7c34-48be-a804-3e20a41a833f","arxiv_id":"1908.01513","paper_version":5,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A quasi-convex relaxation of the curvature-dimension condition gives dimension-independent Poincaré and log-Sobolev constants on Heisenberg groups and other sub-Riemannian manifolds, up to a universal factor.","lead":"Mathematicians have developed a new condition, the Quasi Curvature-Dimension condition, that transfers sharp analytic inequalities from curved spaces to sub-Riemannian spaces such as the Heisenberg group. This yields the first explicit, dimension-independent bounds for Poincaré and log-Sobolev inequalities on these spaces.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the main risk is the external interpolation inequality (Thm 2.1/2.2), already flagged by the reader.","rationale":"I attempted to stress-test the central claim by checking the internal transitions: the Jensen step in Proposition 2.4, the localization theorem's preservation of QCD coefficients, the one-dimensional envelope construction in Proposition 5.7, and the stability argument in Theorem 6.3. Each step survived scrutiny. The localization proof is technically involved but the structure follows the Cavalletti-Mondino paradigm, adapted to general coefficients, and the measurability issues are addressed in Lemma 4.5. The one-dimensional equivalence is justified by the model-density envelope argument and the connectedness/slope parametrization in Lemma 5.9. The only genuine risk to the headline result is the external Jacobian interpolation inequality, exactly as the reader's weakest_assumption states. Since that inequality is published and the paper uses it only under the stated hypotheses (ideal sub-Riemannian manifolds and corank 1 Carnot groups), I do not see a reason to alter the ACCEPT verdict. The recommended check is a concrete numerical validation of the external inequality in the simplest nontrivial case, H^1, which would directly test the exponent and constants that drive the Q factor.","tokens_in":42082,"tokens_out":27633,"duration_ms":265719,"concrete_test":"Specialize to the Heisenberg group H^1: explicitly parametrize the W2 geodesic between two smooth compactly supported densities (e.g., Gaussians), compute the densities rho_t and the distortion coefficients beta_t from the known sub-Riemannian geodesic structure, and numerically verify the pointwise inequality (2.1) in Theorem 2.1 on a fine grid of t in (0,1). If the inequality fails or requires an additional factor depending on the dimension, the Q=4 dimension-independent claim would need revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's internal chain is coherent: MCP plus the Jacobian interpolation inequality yields QCD via Jensen's inequality (Proposition 2.4); the general localization theorem (Theorem 4.1) transfers the QCD inequality to one-dimensional needles; Proposition 5.7 shows a QCD density is trapped between h and Qh by a CD density; and the stability lemma then transfers functional inequalities with a Q factor. I found no algebraic inconsistency or circular step in this argument. The single load-bearing external dependency is the exact Jacobian interpolation inequality (2.1), due to Barilari-Rizzi in the ideal case and Balogh-Kristaly-Sipos in the corank-1 non-ideal case. In particular, the proof uses the dimension n appearing in the exponents of rho_0 and rho_1 and the exact form with no additional multiplicative constant. If that inequality carried a hidden dimension-dependent factor or used an effective dimension different from the topological dimension n, the constant Q=2^{N-n} in Corollaries 2.5 and 2.6, and hence the factor 4 for Heisenberg groups, would change. The paper does not reprove these external theorems but cites them transparently; they are published peer-reviewed results. Thus the dependency is real but does not constitute an internal flaw.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces the Quasi Curvature-Dimension condition QCD(Q,K,N), a quasi-convex relaxation of the Lott-Sturm-Villani CD(K,N) condition. On an n-dimensional ideal sub-Riemannian manifold, it shows that MCP(K,N) implies QCD(2^{N-n},K,N) by combining the Barilari-Rizzi Jacobian interpolation inequality with Jensen's inequality; a result of Balogh-Kristaly-Sipos covers non-ideal corank-1 Carnot groups. The paper then proves a general localization theorem for Monge spaces satisfying arbitrary interpolation inequalities, establishes that one-dimensional QCD densities are equivalent up to the factor Q to CD densities via a CD upper envelope, and combines these to show that L^p-Poincaré and log-Sobolev constants on QCD spaces lie within a factor Q of sharp one-dimensional CD constants. Applications include Heisenberg groups (factor 4 in all dimensions), generalized H-type groups, corank-1 Carnot groups, the Grushin plane, and Sasakian/3-Sasakian manifolds.","tokens_in":42214,"tokens_out":17616,"duration_ms":184368,"significance":"If correct, the result gives the best known quantitative L^p-Poincaré and log-Sobolev estimates on these sub-Riemannian spaces and the first dimension-independent spectral-gap estimate up to factor 4 on all Heisenberg groups. The paper's main strengths are the explicit, parameter-free constant Q=2^{N-n}, the fully detailed localization argument for general interpolation coefficients, the clean one-dimensional equivalence (Proposition 5.7), and the optimality discussion via Juillet's construction. The central derivation is transparent: the only non-elementary input is the exact Jacobian interpolation inequality (2.1) from [17] and [14], which is cited rather than reproved. This is a genuine external dependency, but it is clearly identified and does not create an internal gap.","major_comments":[],"minor_comments":[{"comment":"The denominator appears to be 4k^2 C D^2, whereas Theorem 2.9 together with Q=4k yields the constant pi^2/(4k C D^2); please correct the displayed formula or explain the extra factor k.","section":"Theorem 1.1, log-Sobolev display"},{"comment":"The first author's name is corrupted as 'Haj/suppress lasz'; it should read Hajłasz.","section":"References, [44]"},{"comment":"The entry N in [n, infinity) is ambiguous without a sentence clarifying that Q=2^{N-n} and that dimension-independent estimates require a uniform bound on N-n.","section":"Section 2 table, ideal Carnot groups row"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is within the journal's scope and I see no grounds for rejection. The requested changes are typographical or local; the external interpolation inequality dependency is substantial but transparently attributed to published work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Milman's QCD paper is a real advance. The new definition QCD(Q,K,N) is a quasi-convex relaxation of CD that is weak enough to hold on ideal sub-Riemannian manifolds and strong enough to deliver quantitative functional inequalities. The main Theorem 2.9 delivers what the abstract promises: L^p-Poincaré and log-Sobolev constants on QCD spaces are within factor Q of the sharp CD constants. For Heisenberg groups Q=4, independent of dimension, which is the first estimate of that kind. The chain is transparent: MCP plus the Barilari–Rizzi interpolation inequality yields QCD via Jensen with Q=2^{N-n}, the generalized localization Theorem 4.1 reduces the problem to geodesics, Proposition 5.7 shows one-dimensional QCD densities are trapped between h and Qh by a CD density, and the stability lemma transfers the functional inequalities with factor Q. I read the proofs of Theorem 4.1 and Section 6 and found no gaps. The Q factor is not fitted; it is derived from explicit exponents. The paper is also honest about what is external and gives credit where it is due.\n\nThe soft spots are real but not fatal. The load-bearing external input is the Jacobian interpolation inequality (2.1), cited to Barilari–Rizzi and Balogh–Kristály–Sipos, in the exact form with exponent 1/n and no extra constant. If that inequality carried a hidden dimension-dependent factor or used an effective dimension different from the topological dimension, the factor Q=2^{N-n} would change. This is a dependency on published peer-reviewed results, not an internal flaw. Relatedly, the non-ideal case is covered only for corank 1 Carnot groups, and the ideal assumption excludes many sub-Riemannian structures; the paper says so explicitly. The long analytic proofs are not machine-checked, so the reader's MODERATE confidence is fair. The citation pattern is fine: self-citations to Han–Milman and the author's earlier localization work are appropriate, and the central claim does not reduce to those results.\n\nThis paper is for researchers in optimal transport, metric measure spaces, sub-Riemannian geometry, and functional inequalities. It deserves a serious referee. I would send it out and expect it to be accepted after normal revision; the main theorem is important and the internal structure is coherent.","headline":"A coherent and important paper: the QCD relaxation plus localization yields first dimension-independent Poincaré/log-Sobolev constants on Heisenberg and ideal sub-Riemannian spaces; the external interpolation inequality is the only real caveat.","tokens_in":752,"tokens_out":1008,"would_cite":true,"duration_ms":30480,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C17","49Q22","35P15"],"pacs":[],"model":"deepseek-v4-flash","headline":"A quasi-curvature-dimension condition extends sharp spectral estimates to Heisenberg groups.","keywords":["Quasi Curvature-Dimension condition","sub-Riemannian manifolds","Heisenberg groups","Lp-Poincaré inequality","log-Sobolev inequality","localization","measure contraction property","spectral gap"],"falsifier":"Compute the Neumann spectral gap of the sub-Laplacian on a geodesic ball of diameter $D$ in the Heisenberg group $\\mathbb{H}^d$. The paper predicts $D^2\\lambda \\geq \\pi^2/4$ for every $d$; if any dimension yields a normalized gap below this threshold, the factor-$4$ conclusion fails.","tokens_in":41781,"feed_emoji":"📐","tokens_out":8407,"duration_ms":82523,"temperature":0.7,"pith_summary":"This paper argues that many strictly sub-Riemannian spaces, long known to fail the curvature-dimension condition, still satisfy a quasi-convex version of it, and that this is enough to transfer the standard analytic consequences of curvature-dimension theory. The new condition, $\\mathsf{QCD}(Q,K,N)$, relaxes the $\\mathsf{CD}(K,N)$ interpolation inequality by a slack factor $Q$; on an ideal sub-Riemannian manifold of topological dimension $n$, the measure contraction property $\\mathsf{MCP}(K,N)$ implies $\\mathsf{QCD}(Q,K,N)$ with $Q = 2^{N-n}$. A general localization theorem then reduces functional inequalities on QCD spaces to the one-dimensional line, where every QCD density is squeezed between a CD density and $Q$ times it. The payoff is that on every Heisenberg group, regardless of dimension, the Li-Yau / Zhong-Yang spectral-gap estimate holds up to the universal factor $Q = 4$.","feed_headline":"Heisenberg groups obey sharp spectral bounds up to factor 4","feed_subtitle":"The new QCD condition transfers one-dimensional sharp constants to sub-Riemannian manifolds in every dimension.","key_machinery":"The driving object is the new condition $\\mathsf{QCD}(Q,K,N)$, a quasi-convex relaxation in which the $\\mathsf{CD}(K,N)$ density interpolation inequality holds with an extra factor $1/Q^{1/N}$. It is powered by the Jacobian interpolation inequality (2.1), which bounds the density along a Wasserstein geodesic by the interpolated endpoint densities with sharp exponents $1/n$; combining that with $\\mathsf{MCP}(K,N)$ and Jensen's inequality yields the QCD inequality with $Q = 2^{N-n}$. The second mechanism is the general localization theorem, which disintegrates a QCD space into one-dimensional geodesics carrying the same interpolation coefficients, and the one-dimensional equivalence—any QCD density lies between a CD density and $Q$ times it—turns the geodesic problem into a known one-dimensional extremal problem.","core_discovery":"The central claim is that a broad family of strictly sub-Riemannian spaces, which are excluded from the usual curvature-dimension theory, nevertheless carries a quasi-convex version of it, and that this version is strong enough to reproduce the standard analytic consequences. The author proves that on an ideal sub-Riemannian manifold of topological dimension $n$, the measure contraction property $\\mathsf{MCP}(K,N)$ together with the Jacobian interpolation inequality along Wasserstein geodesics (Theorem 2.1) implies $\\mathsf{QCD}(Q,K,N)$ with $Q = 2^{N-n}$; for non-ideal corank-1 Carnot groups the same conclusion, with $Q = 4$, follows from the companion interpolation inequality (Theorem 2.2). A localization theorem for arbitrary interpolation coefficients reduces any $\\mathsf{QCD}(Q,K,N)$ space to one-dimensional geodesics, and on the line every QCD density $h$ is squeezed between a CD density $f$ and $Qh$. Hence the best $L^p$-Poincaré and log-Sobolev constants on the QCD space are bounded below by $1/Q$ times the sharp one-dimensional CD constants, which gives the first dimension-independent spectral-gap and log-Sobolev estimates on Heisenberg groups in every dimension.","pith_inferences":["If the Jacobian interpolation inequality is established beyond the ideal and corank-1 settings, the same proof should produce QCD and hence dimension-independent functional inequalities on broader classes of sub-Riemannian manifolds.","The one-dimensional density characterization suggests that a QCD space is, in a measure-theoretic sense, sandwiched between a genuine CD space and a bounded perturbation of it; a natural testable extension is to search for explicit CD densities dominating the Heisenberg measure along transport geodesics.","The quasi Brunn-Minkowski inequality implied by QCD points toward a possible stability and tensorization theory for QCD, although the paper does not develop such a theory, and this remains an open direction.","Because the constants are explicit, one could numerically test the sharpness of the factor $4$ by computing spectral gaps of the sub-Laplacian on geodesic balls in the Heisenberg group and comparing with the predicted universal constant."],"forward_implications":["On every Heisenberg group $\\mathbb{H}^d$, the sharp Li-Yau / Zhong-Yang spectral-gap bound holds up to the universal factor $4$, independently of the dimension $d$.","Ideal generalized H-type groups, the Grushin plane, Sasakian and 3-Sasakian manifolds, and corank-1 Carnot groups all obtain $L^p$-Poincaré and log-Sobolev constants within the factor $Q = 2^{N-n}$ of the best one-dimensional $\\mathsf{CD}(K,N)$ constants.","Because $Q$ depends only on the gap $N-n$ between geodesic and topological dimension, any family of these spaces with uniformly bounded $N-n$ admits estimates that do not degrade with the ambient dimension.","The results constitute the first quantitative functional inequalities on these sub-Riemannian spaces that are truly independent of dimension.","The same mechanism applies to any property of $\\mathsf{CD}(K,N)$ spaces that is amenable to localization and stable under the one-dimensional density perturbation $h \\leq f \\leq Qh$."],"supporting_citations":[{"why":"Supplies the Jacobian interpolation inequality along Wasserstein geodesics in the ideal sub-Riemannian setting, the key input behind Proposition 2.4.","marker":"[17]"},{"why":"Supplies the analogous Jacobian interpolation inequality for non-ideal corank-1 Carnot groups, used to obtain QCD with Q = 4.","marker":"[14]"},{"why":"Provides the original Riemannian interpolation inequality that the QCD condition relaxes and whose one-dimensional constants serve as the benchmark.","marker":"[34]"},{"why":"The localization theorem for CD spaces whose proof is extended here to completely general interpolation coefficients.","marker":"[30]"},{"why":"Provides localization for locally finite measures on MCP spaces, needed in the proof of the general localization theorem.","marker":"[32]"},{"why":"Yields the sharp one-dimensional spectral-gap constant for p = 2 that the QCD comparison is measured against.","marker":"[12]"},{"why":"Establishes the sharp one-dimensional $L^p$-Poincaré constants for general p, used to evaluate the model constants in Theorem 2.9.","marker":"[85]"},{"why":"Establishes MCP(0, n+2) and the measure distortion coefficients for corank-1 Carnot groups, underlying Corollary 2.6.","marker":"[80]"}],"fun_headline_variants":["QCD condition wins dimension-free bounds on Heisenberg","Quasi Curvature-Dimension: sharp sub-Riemannian inequalities","MCP to QCD: sharp spectral gap on all Heisenberg groups","First dim-independent Poincare and log-Sobolev on Heisenberg","Sub-Riemannian spaces get CD-type bounds via QCD"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is an external, deep Jacobian interpolation inequality along optimal geodesics, holding with exact exponents and no error term; if that inequality carries any correction or extra dimension factor, the final constants degrade by at least that amount, and outside the ideal and corank-1 settings it is not covered.","fun_headline_variants_meta":{"raw":{"variants":["QCD condition wins dimension-free bounds on Heisenberg","Quasi Curvature-Dimension: sharp sub-Riemannian inequalities","MCP to QCD: sharp spectral gap on all Heisenberg groups","First dim-independent Poincare and log-Sobolev on Heisenberg","Sub-Riemannian spaces get CD-type bounds via QCD"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000213,"raw_usage":{"total_tokens":1553,"prompt_tokens":1205,"completion_tokens":348,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":821,"completion_tokens_details":{"reasoning_tokens":258}},"tokens_in":821,"tokens_out":348,"duration_ms":4149,"temperature":1.0,"reasoning_tokens":258,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:10:31.617173+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the Neumann spectral gap of the sub-Laplacian on a geodesic ball of diameter $D$ in the Heisenberg group $\\mathbb{H}^d$. The paper predicts $D^2\\lambda \\geq \\pi^2/4$ for every $d$; if any dimension yields a normalized gap below this threshold, the factor-$4$ conclusion fails.","supporting_citations":[{"cited_title":"Barilari and L","cited_arxiv_id":null,"evidence_quote":"Supplies the Jacobian interpolation inequality along Wasserstein geodesics in the ideal sub-Riemannian setting, the key input behind Proposition 2.4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the analogous Jacobian interpolation inequality for non-ideal corank-1 Carnot groups, used to obtain QCD with Q = 4."},{"cited_title":"Cordero-Erausquin, R","cited_arxiv_id":null,"evidence_quote":"Provides the original Riemannian interpolation inequality that the QCD condition relaxes and whose one-dimensional constants serve as the benchmark."},{"cited_title":"Cavalletti and A","cited_arxiv_id":null,"evidence_quote":"The localization theorem for CD spaces whose proof is extended here to completely general interpolation coefficients."},{"cited_title":"Valtorta","cited_arxiv_id":null,"evidence_quote":"Establishes the sharp one-dimensional $L^p$-Poincaré constants for general p, used to evaluate the model constants in Theorem 2.9."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes MCP(0, n+2) and the measure distortion coefficients for corank-1 Carnot groups, underlying Corollary 2.6."}],"review_version":1}