Digesting the proof of the sharp thin-shell inequality

Boaz Klartag; Yuansi Chen

This paper proves the optimal constant in the thin-shell variance conjecture: every isotropic log-concave random vector X in R^n satisfies Var(|X|^2) ≤ 8n.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-07-31 23:50 UTC pith:K5UMSPQD

load-bearing objection The sharp constant 8 looks right; the proof is a genuine advance and deserves a referee, though the main theorem leans on an unproved H^{-1} inequality that should be checked carefully.

arxiv 2607.23307 v1 pith:K5UMSPQD submitted 2026-07-25 math.MG math.FAmath.PR

Digesting the proof of the sharp thin-shell inequality

Yuansi Chen , Boaz Klartag This is my paper

classification math.MG math.FAmath.PR MSC 52A4060E1552A2335J96

keywords thin-shell inequalityvariance conjecturelog-concave measuresmoment measuresMonge-Ampère equationHessian boundthird-moment tensorsimplex extremality

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper resolves the variance conjecture for log-concave measures by proving the sharp universal bound Var(|X|^2) ≤ 8n for any isotropic log-concave random vector in R^n, with equality attained when the coordinates are independent standard centered exponentials. This settles the optimal constant in the thin-shell problem, a route to central limit theorems for convex bodies and to slicing-type inequalities. The proof works through the theory of log-concave moment measures and the Monge–Ampère equation, converting the question into a bound on the Hessian H = ∇²ψ of the moment-map potential: ∫ ||H||²_HS dν ≤ 2n. Along the way it gives a sharp bound on the Hilbert–Schmidt norm of the third-moment tensor, ||T||²_HS ≤ 4n, and, for uniform distributions on convex bodies, identifies the regular simplex as the extremal case for Var(|X|²).

Core claim

The paper's central claim is that the variance conjecture holds with the optimal constant: if X is any isotropic log-concave vector in R^n, then E[(|X|²−n)²] ≤ 8n. The constant is sharp, with equality for independent standard centered exponential coordinates. The paper also proves the companion sharp bound ||T||²_HS ≤ 4n for the third-moment tensor, and derives, for uniform measures on convex bodies, the sharp bounds Var(|X|²) ≤ 4n(n+1)²/((n+3)(n+4)) and ||T||²_HS ≤ 4n(n−1)(n+2)/((n+3)²), attaining equality at the regular simplex. All of these rest on the assertion that, for the moment-map potential ψ of an isotropic log-concave measure, ∫ ||∇²ψ||²_HS e^{-ψ} dx ≤ 2n under suitable regularity

What carries the argument

The central object is the Hessian field H = ∇²ψ of the potential ψ that represents the log-concave measure as a moment measure, together with the weighted Riemannian manifold (R^n, ∇²ψ, e^{-ψ}dx). The main new step is an exact pointwise identity (Lemma 3.5): q² − d² = (1/6)Σ ψ²_{aij}/(λ_aλ_iλ_j)[(λ_a−λ_i)²+(λ_i−λ_j)²+(λ_j−λ_a)²] ≥ 0, where the λ_i are the eigenvalues of H and d², q² are trace-type averages of third derivatives. This gives Q² ≥ D², and combining this with the identity N₂ = D² + A² + Q² and the Poincaré-type inequality D² ≥ N₂ − n yields ∫||H||²_HS dν ≤ 2n. A Stein-kernel integration by parts transfers this Hessian bound to the original measure, bounding Σ_i ||x_i||²_{H^{−1}(µ

Load-bearing premise

The argument that removes the smoothness assumptions relies on being able to approximate any log-concave measure by very smooth ones while keeping all moments up to order four under control, and on a certain 'gradient-cost' functional not dropping in the limit — a property the paper cites from earlier work rather than proves.

What would settle it

A single counterexample would settle it: an isotropic log-concave random vector with Var(|X|²) > 8n would refute the main theorem. A more targeted test: find a sequence of isotropic log-concave measures converging weakly for which the functional Σ_i ||x_i||²_{H^{−1}(µ)} fails to be lower-semicontinuous — that would break the limiting step (24) and reduce the claim to smooth measures.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

If the proof holds, the variance conjecture is settled with the optimal constant 8, so every isotropic log-concave vector concentrates in a shell of width of order √n around radius √n with the sharp constant.
The sharp third-moment bound ||T||²_HS ≤ 4n gives an extremal statement for third-order structure: the product of centered exponentials attains it, and the regular simplex attains the uniform-on-convex-body version.
The cone construction transfers the sharp inequality to uniform distributions on convex bodies, identifying the regular simplex as the maximizer of Var(|X|²) among all isotropic convex bodies.
The sharp thin-shell constant gives new evidence for the strong slicing conjecture, since the moment-measure method suggests a route to the functional version h(X) ≥ n, where equality holds for centered exponentials.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the sharp bound is stable, one expects a quantitative stability result: a log-concave isotropic vector with Var(|X|²) close to 8n should be close, in some transport or moment sense, to the product of centered exponentials; the exact identity behind Lemma 3.5 could be the tool to prove such a statement.
The proof's dependence on lower-semicontinuity of the H^{-1} functional (cited to another work) means that a self-contained proof of that semicontinuity would remove the last technical gap and could extend the method beyond the log-concave class.
The same Hessian-trace bootstrap might apply to higher moments: if an analogue of Lemma 3.5 holds for traces of H^k, it could yield sharp bounds on higher cumulants or even a direct proof of the functional slicing inequality h(X) ≥ n.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

The sharp constant 8 looks right; the proof is a genuine advance and deserves a referee, though the main theorem leans on an unproved H^{-1} inequality that should be checked carefully.

read the letter

Yuansi and Boaz have done something real here. The paper determines the optimal constant in the variance conjecture for isotropic log-concave measures: Var(|X|²) ≤ 8n, sharp for iid centered exponentials. The novelty is not a tweaked argument; it's a new identity — Lemma 3.5 — showing q² ≥ d² pointwise, which lets the authors bootstrap N₂ ≥ 2(N₂−n) and get N₂ ≤ 2n. From that they also get a sharp third-moment bound and a clean simplex-extremal corollary. I agree with the reader's assessment: the central chain checks out, the constants are derived rather than fitted, and the circularity burden is low.

The soft spot is exactly what the stress-test calls out. Theorem 1.1 depends on inequality (2), Var(|X|²) ≤ 4∑‖x_i‖²_{H^{-1}(µ)}, which is imported as a black box from [18]/[3]/[24]. The factor 4 is essential: it converts Theorem 1.4's 2n into 8n. The paper cites it, but does not prove it or even state it as a lemma with the constant displayed. If that inequality carries an extra log or a wrong constant, the sharp upper bound collapses. This is load-bearing external input and should be verified by the referee — or better, proved in the paper.

The regularity-removal paragraph is terse, and the lower-semicontinuity claim is cited to [24]. I think the stress-test is right that this is a secondary concern: the approximation by smooth, truncated, convolved measures is standard, and the moment-convergence argument is sketched adequately. But a referee might want a few more lines there.

On the GPT provenance: the disclosure is honest and unusual. The authors say the initial proof came from GPT-5.6 Pro and they verified and revised it. That does not by itself affect correctness, but it raises the bar for human checking. The paper would be stronger with machine-checked proofs, though that's not typical for this area.

This is a paper for anyone working on thin-shell bounds, slicing, or moment measures; the method matters as much as the constant. Overall, it deserves a serious referee and likely publication after the H^{-1} inequality is either proved or precisely sourced. I would bring it to a reading group and cite it.

Referee Report

0 major / 5 minor

Summary. The paper proves the sharp thin-shell inequality for isotropic log-concave random vectors: E[(|X|^2 - n)^2] <= 8n, with equality for centered exponentials. It also proves the companion sharp bound ||T(X)||_HS^2 <= 4n for the third-moment tensor and, for uniform measures on convex bodies, sharp variance and third-moment bounds attained by regular simplices. The proof proceeds through log-concave moment measures: under the regularity assumptions of [20], the Hessian H of the moment-map potential satisfies the exact identity LH+H = A+Q (Lemma 3.1), from which the paper derives a pointwise identity for LTr(H^2) and, via a bootstrap using positivity of Q-D and the Poincare inequality, the key estimate int Tr(H^2) dnu <= 2n (Theorem 1.5). This is transferred to H^{-1} norms of the coordinate functions by the Stein-kernel identity (23), yielding Theorem 1.4, and then to the variance bound through the H^{-1} inequality (2). The extremal examples are computed explicitly.

Significance. If correct, the paper settles a long-standing conjecture with the optimal constant 8, identifies the extremal measures, and gives a new route to sharp moment bounds via moment measures and Monge-Ampere equations. The main chain is built from exact pointwise identities and has no fitted constants; the bootstrap in Section 3 is simple and elegant. The result also has potential implications for the slicing problem and the non-symmetric Mahler conjecture, as the authors explain. The main caveats are the reliance on the external H^{-1} inequality (2) and the compressed regularity-removal argument; both are standard and cited, but the exposition should make them precise.

minor comments (5)

[Section 1, Eq. (2)] The factor 4 in the H^{-1} inequality is load-bearing for the final constant 8, yet the paper only says that the inequality follows from [18] and [3] as explained in [24]. I recommend stating the precise theorem and either giving a proof in an appendix or quoting the exact statement with theorem numbers. This would remove any ambiguity about the sharp constant at the bridge between Theorem 1.4 and Theorem 1.1.
[Section 3, end; Section 4, display (24)] The regularity-removal argument is compressed. Please expand the construction of the approximating measures, the diagonal choice of (delta, epsilon, R), and the convergence in moments of order at most four after recentering and isotropic normalization. For (24), the lower-semicontinuity of the functional is cited to [24]; a one-paragraph proof using fixed compactly supported g would make the paper self-contained at this point.
[Section 2, action coordinates] There is a typo: 'Let Let phi = psi^*' should read 'Let phi = psi^*'.
[Section 3, Lemma 3.5] The cyclic-average trick in Lemma 3.5 is correct, but the sentence explaining why the tensor symmetry allows the replacement could be expanded for readability: the sum is over ordered triples and psi_{aij} is fully symmetric, so averaging the numerator over the three cyclic permutations is valid.
[Appendix A, Lemma A.1] In the proof of (28), the constant C is defined only after the inequality as C = int eta'^2. It would be clearer to define it before the display.

Circularity Check

0 steps flagged

No circularity: the sharp thin-shell and third-moment bounds are derived from exact pointwise identities, with self-citations used only as established prior theorems that do not encode the target constant.

full rationale

The core of the paper is the proof of Theorem 1.5, and it is not circular. Lemma 3.1 derives the pointwise identity LH+H=A+Q from the Monge-Ampère equation; Lemma 3.5 is an exact algebraic identity showing Q2≥D2; Lemmas 3.3 and 3.4 convert these into the quantitative bound N2≤2n using only the Poincaré inequality and justified integrations by parts. No constant is fitted: the exponential example in Example 3.6 is computed after the bound to demonstrate equality. Theorem 1.4 is then obtained from the Stein-kernel transfer (23) and Cauchy-Schwarz, and Theorem 1.1 is the combination of Theorem 1.4 with the previously known H^{-1} inequality (2) cited to [18]/[3]/[24]. That external inequality is parameter-free and does not contain the sharp constant 8n, so its use is an independent input rather than a definitional reduction. The same applies to the moment-measure machinery from [20] and the lower-semicontinuity/regularity-removal step (24): they are prior results or routine weak-continuity facts needed for full generality, but they do not smuggle in the conclusion. The regularity-removal proof is sketched, which is a completeness/verification issue, not circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 0 invented entities

The proof introduces no free parameters: the constants 8, 4, 2 are universal, derived from exact identities, and matched by computed extremal examples (centered exponential, simplex). It rests on published external framework theorems: the moment-measure theorem, the [20] manifold results (Poincaré inequality, boundedness (10)), the H−1 reduction (2), and a cited lower-semicontinuity property. No entities are invented; the weighted manifold and Stein-kernel objects predate the paper.

axioms (5)

domain assumption Moment-measure theorem: for every centered full-dimensional log-concave µ, there is a convex ψ, unique up to translations, with (∇ψ)#(e^{−ψ}dx) = µ (Cordero-Erausquin–Klartag [9]).
Section 2: the whole proof operates on the representation µ = (∇ψ)#ν, ν = e^{−ψ}dx; if this theorem failed or required extra hypotheses, H = ∇²ψ and the manifold M*_µ would not exist.
ad hoc to paper Technical regularity from [20]: µ supported in a bounded open convex K, V = −log ρ smooth with all derivatives bounded, ∇ψ a diffeomorphism, and the bound TrH(x) ≤ 2R(K)² from [20, Thm 1.1], display (10).
Used in Sections 2–3 and Appendix A to justify integrations by parts and boundedness of G = Tr(H²). Later discharged: the proof removes these hypotheses by Gaussian-convolution/truncation approximation (end of Section 3), so the main theorems do not ultimately depend on them.
domain assumption Poincaré inequality on the weighted manifold (8): Var_ν(u) ≤ ∫ ψ^{ij} u_i u_j dν for u ∈ D(E), from [20].
Applied entrywise to H in Lemma 3.4 to get N2 − n ≤ D2, a load-bearing half of the bootstrap N2 ≤ 2n.
domain assumption H−1 inequality (2): Var(|X|²) ≤ 4 Σ_i ||x_i||²_{H−1(µ)}, attributed to Klartag [18] and Barthe–Klartag [3], as explained in [24].
Bridges Theorem 1.4's H−1 bound (2n) to Theorem 1.1's Var ≤ 8n; the factor 4 enters the sharp constant, so the paper depends on its exact form.
domain assumption Lower semi-continuity of µ ↦ Σ_i ||x_i||²_{H−1(µ)} under weak convergence of isotropic log-concave measures (cited to [24], display (24)).
The regularity-removal argument uses a liminf passage; if the functional were not lower-semicontinuous, Theorems 1.1, 1.2 and 1.4 would hold only under smoothness assumptions.

pith-pipeline@v1.3.0-alltime-deepseek · 13689 in / 41602 out tokens · 341665 ms · 2026-07-31T23:50:47.819884+00:00 · methodology

0 comments

read the original abstract

We present a proof that determines the optimal value of the universal constant in the thin-shell theorem for log-concave distributions in high dimensions. We prove that for any log-concave random vector $X = (X_1,\ldots,X_n)$ in $\mathbb{R}^n$ with mean zero and identity covariance, $$ {\rm Var}( |X|^2 ) \leq 8 n. $$ The constant $8$ is optimal: equality is attained when $X_1,\ldots,X_n$ are independent, identically distributed, standard, centered exponential random variables. Moreover, among isotropic random vectors distributed uniformly on convex bodies in $\mathbb{R}^n$, the quantity ${\rm Var}(|X|^2)$ is maximized by the uniform distribution on a regular simplex. We also provide a corresponding sharp bound on the Hilbert-Schmidt norm of the tensor of $3^{rd}$-moments of isotropic, log-concave distributions. The argument relies on the analysis of a weighted Riemannian manifold associated with log-concave moment measures and the Monge-Amp\`ere equation. This manifold was studied in this context in \cite{lc_moment}. The main improvement over \cite{lc_moment} comes from a concise yet effective analysis of the $3^{rd}$-derivatives tensor of the potential. The proof was found by GPT-5.6 Pro in response to prompts supplied by the first-named author, following general discussions between the two authors concerning log-concave moment measures. The prompts referred to the paper ``Logarithmically-concave moment measures I'' and suggested bootstrapping a bound on the second trace moment.

discussion (0)

Reference graph

Works this paper leans on

27 extracted references · 3 linked inside Pith

[1]

and Perissinaki, I.,The central limit problem for convex bodies.Trans

Anttila, M., Ball, K. and Perissinaki, I.,The central limit problem for convex bodies.Trans. Amer. Math. Soc., V ol. 355, no. 12, (2003), 4723–4735

2003
[2]

Springer, 2014

Bakry, D., Gentil, I., Ledoux, M.,Analysis and Geometry of Markov Diffusion Operators. Springer, 2014

2014
[3]

and Klartag, B.,Spectral gaps, symmetries and log-concave perturbations.Bull

Barthe, F. and Klartag, B.,Spectral gaps, symmetries and log-concave perturbations.Bull. Hellenic Math. Soc., V ol. 64, (2020), 1–31

2020
[4]

Berman, R. J. and Berndtsson, B.,Real Monge–Amp `ere equations and K ¨ahler–Ricci soli- tons on toric log Fano varieties.Ann. Fac. Sci. Toulouse Math. (6), V ol. 22, no. 4, (2013), 649–711

2013
[5]

Bobkov, S. G. and Koldobsky, A.,On the central limit property of convex bodies.In: Ge- ometric Aspects of Functional Analysis (2001–02), Lecture Notes in Math., V ol. 1807, Springer, (2003), 44–52

2001
[6]

Brazitikos, S., Giannopoulos, A., Valettas, P., Vritsiou, B.-H.,Geometry of isotropic convex bodies.American Mathematical Society, 2014

2014
[7]

Chen, Y .,An almost constant lower bound of the isoperimetric coefficient in the KLS con- jecture.Geom. Funct. Anal., V ol. 31, no. 1, (2021), 34–61

2021
[8]

Preprint, arXiv:2605.09334

Chen, S., Li, Y ., Xi, D., Xu, Z.,The Mahler conjecture in three dimensions. Preprint, arXiv:2605.09334

Pith/arXiv arXiv
[9]

and Klartag, B.,Moment measures.J

Cordero-Erausquin, D. and Klartag, B.,Moment measures.J. Funct. Anal., V ol. 268, no. 12, (2015), 3834–3866. 21

2015
[10]

Eldan, R.,Thin shell implies spectral gap up to polylog via a stochastic localization scheme. Geom. Funct. Anal., V ol. 23, no. 2, (2013), 532–569

2013
[11]

and Klartag, B.,Approximately Gaussian marginals and the hyperplane conjec- ture.In: Concentration, Functional Inequalities and Isoperimetry, Contemp

Eldan, R. and Klartag, B.,Approximately Gaussian marginals and the hyperplane conjec- ture.In: Concentration, Functional Inequalities and Isoperimetry, Contemp. Math., V ol. 545, Amer. Math. Soc., (2011), 55–68

2011
[12]

Probab.47(2019), no

Fathi, M.,Stein kernels and moment maps, Ann. Probab.47(2019), no. 4, 2172–2185

2019
[13]

Fleury, B.,Concentration in a thin Euclidean shell for log-concave measures.J. Funct. Anal., V ol. 259, no. 4, (2010), 832–841

2010
[14]

Guan, Q.,A note on Bourgain’s slicing problem.Preprint, arXiv:2412.09075, (2024)

Pith/arXiv arXiv 2024
[15]

and Milman, E.,Interpolating thin-shell and sharp large-deviation estimates for isotropic log-concave measures.Geom

Gu ´edon, O. and Milman, E.,Interpolating thin-shell and sharp large-deviation estimates for isotropic log-concave measures.Geom. Funct. Anal., V ol. 21, no. 5, (2011), 1043–1068

2011
[16]

Jambulapati, A., Lee, Y . T. and Vempala, S. S.,A slightly improved bound for the KLS constant.Preprint, arXiv:2208.11644, (2022)

Pith/arXiv arXiv 2022
[17]

Klartag, B.,Power-law estimates for the central limit theorem for convex sets.J. Funct. Anal., V ol. 245, no. 1, (2007), 284–310

2007
[18]

Klartag, B.,A Berry–Esseen type inequality for convex bodies with an unconditional basis. Probab. Theory Related Fields, V ol. 145, no. 1–2, (2009), 1–33

2009
[19]

Klartag, B.,High-dimensional distributions with convexity properties.In: Proceedings of the Fifth European Congress of Mathematics, Amsterdam, July 2008, European Mathemat- ical Society, (2010), 401–417

2008
[20]

2116, Springer, (2014), 231–260

Klartag, B.,Logarithmically-concave moment measures I.Geometric Aspects of Functional Analysis, Lecture Notes in Math., V ol. 2116, Springer, (2014), 231–260

2014
[21]

Math., V ol

Klartag, B.,Isotropic constants and Mahler volumes.Adv. Math., V ol. 330, (2018), 74–108

2018
[22]

4, (2023), 17 pp

Klartag, B.,Logarithmic bounds for isoperimetry and slices of convex sets.Ars Inveniendi Analytica, Paper no. 4, (2023), 17 pp

2023
[23]

and Lehec, J.,Bourgain’s slicing problem and KLS isoperimetry up to polylog

Klartag, B. and Lehec, J.,Bourgain’s slicing problem and KLS isoperimetry up to polylog. Geom. Funct. Anal., V ol. 32, no. 5, (2022), 1134–1159

2022
[24]

and Lehec, J.,Thin-shell bounds via parallel coupling.Preprint, arXiv:2507.15495, (2025)

Klartag, B. and Lehec, J.,Thin-shell bounds via parallel coupling.Preprint, arXiv:2507.15495, (2025)

arXiv 2025
[25]

Lee, Y . T. and Vempala, S.,Eldan’s stochastic localization and the KLS hyperplane con- jecture: an improved lower bound for expansion.In: Proceedings of the 58th Annual IEEE Symposium on Foundations of Computer Science (FOCS 2017), IEEE Computer Society, (2017), 998–1007. 22

2017
[26]

7, (1938/39), 118–127

Mahler, K.,Ein Minimalproblem f ¨ur konvexe Polygone, Mathematica (Zutphen) B, V ol. 7, (1938/39), 118–127

1938
[27]

and Zhu, X.,K ¨ahler–Ricci solitons on toric manifolds with positive first Chern class.Adv

Wang, X.-J. and Zhu, X.,K ¨ahler–Ricci solitons on toric manifolds with positive first Chern class.Adv. Math., V ol. 188, no. 1, (2004), 87–103. Seminar for Statistics, Department of Mathematics, ETH Zurich, 8092 Zurich, Switzerland. e-mail:yuansi.chen@stat.math.ethz.ch School of Mathematical Sciences, Tel Aviv University, Tel Aviv 6997801, Israel; and De...

2004

[1] [1]

and Perissinaki, I.,The central limit problem for convex bodies.Trans

Anttila, M., Ball, K. and Perissinaki, I.,The central limit problem for convex bodies.Trans. Amer. Math. Soc., V ol. 355, no. 12, (2003), 4723–4735

2003

[2] [2]

Springer, 2014

Bakry, D., Gentil, I., Ledoux, M.,Analysis and Geometry of Markov Diffusion Operators. Springer, 2014

2014

[3] [3]

and Klartag, B.,Spectral gaps, symmetries and log-concave perturbations.Bull

Barthe, F. and Klartag, B.,Spectral gaps, symmetries and log-concave perturbations.Bull. Hellenic Math. Soc., V ol. 64, (2020), 1–31

2020

[4] [4]

Berman, R. J. and Berndtsson, B.,Real Monge–Amp `ere equations and K ¨ahler–Ricci soli- tons on toric log Fano varieties.Ann. Fac. Sci. Toulouse Math. (6), V ol. 22, no. 4, (2013), 649–711

2013

[5] [5]

Bobkov, S. G. and Koldobsky, A.,On the central limit property of convex bodies.In: Ge- ometric Aspects of Functional Analysis (2001–02), Lecture Notes in Math., V ol. 1807, Springer, (2003), 44–52

2001

[6] [6]

Brazitikos, S., Giannopoulos, A., Valettas, P., Vritsiou, B.-H.,Geometry of isotropic convex bodies.American Mathematical Society, 2014

2014

[7] [7]

Chen, Y .,An almost constant lower bound of the isoperimetric coefficient in the KLS con- jecture.Geom. Funct. Anal., V ol. 31, no. 1, (2021), 34–61

2021

[8] [8]

Preprint, arXiv:2605.09334

Chen, S., Li, Y ., Xi, D., Xu, Z.,The Mahler conjecture in three dimensions. Preprint, arXiv:2605.09334

Pith/arXiv arXiv

[9] [9]

and Klartag, B.,Moment measures.J

Cordero-Erausquin, D. and Klartag, B.,Moment measures.J. Funct. Anal., V ol. 268, no. 12, (2015), 3834–3866. 21

2015

[10] [10]

Eldan, R.,Thin shell implies spectral gap up to polylog via a stochastic localization scheme. Geom. Funct. Anal., V ol. 23, no. 2, (2013), 532–569

2013

[11] [11]

and Klartag, B.,Approximately Gaussian marginals and the hyperplane conjec- ture.In: Concentration, Functional Inequalities and Isoperimetry, Contemp

Eldan, R. and Klartag, B.,Approximately Gaussian marginals and the hyperplane conjec- ture.In: Concentration, Functional Inequalities and Isoperimetry, Contemp. Math., V ol. 545, Amer. Math. Soc., (2011), 55–68

2011

[12] [12]

Probab.47(2019), no

Fathi, M.,Stein kernels and moment maps, Ann. Probab.47(2019), no. 4, 2172–2185

2019

[13] [13]

Fleury, B.,Concentration in a thin Euclidean shell for log-concave measures.J. Funct. Anal., V ol. 259, no. 4, (2010), 832–841

2010

[14] [14]

Guan, Q.,A note on Bourgain’s slicing problem.Preprint, arXiv:2412.09075, (2024)

Pith/arXiv arXiv 2024

[15] [15]

and Milman, E.,Interpolating thin-shell and sharp large-deviation estimates for isotropic log-concave measures.Geom

Gu ´edon, O. and Milman, E.,Interpolating thin-shell and sharp large-deviation estimates for isotropic log-concave measures.Geom. Funct. Anal., V ol. 21, no. 5, (2011), 1043–1068

2011

[16] [16]

Jambulapati, A., Lee, Y . T. and Vempala, S. S.,A slightly improved bound for the KLS constant.Preprint, arXiv:2208.11644, (2022)

Pith/arXiv arXiv 2022

[17] [17]

Klartag, B.,Power-law estimates for the central limit theorem for convex sets.J. Funct. Anal., V ol. 245, no. 1, (2007), 284–310

2007

[18] [18]

Klartag, B.,A Berry–Esseen type inequality for convex bodies with an unconditional basis. Probab. Theory Related Fields, V ol. 145, no. 1–2, (2009), 1–33

2009

[19] [19]

Klartag, B.,High-dimensional distributions with convexity properties.In: Proceedings of the Fifth European Congress of Mathematics, Amsterdam, July 2008, European Mathemat- ical Society, (2010), 401–417

2008

[20] [20]

2116, Springer, (2014), 231–260

Klartag, B.,Logarithmically-concave moment measures I.Geometric Aspects of Functional Analysis, Lecture Notes in Math., V ol. 2116, Springer, (2014), 231–260

2014

[21] [21]

Math., V ol

Klartag, B.,Isotropic constants and Mahler volumes.Adv. Math., V ol. 330, (2018), 74–108

2018

[22] [22]

4, (2023), 17 pp

Klartag, B.,Logarithmic bounds for isoperimetry and slices of convex sets.Ars Inveniendi Analytica, Paper no. 4, (2023), 17 pp

2023

[23] [23]

and Lehec, J.,Bourgain’s slicing problem and KLS isoperimetry up to polylog

Klartag, B. and Lehec, J.,Bourgain’s slicing problem and KLS isoperimetry up to polylog. Geom. Funct. Anal., V ol. 32, no. 5, (2022), 1134–1159

2022

[24] [24]

and Lehec, J.,Thin-shell bounds via parallel coupling.Preprint, arXiv:2507.15495, (2025)

Klartag, B. and Lehec, J.,Thin-shell bounds via parallel coupling.Preprint, arXiv:2507.15495, (2025)

arXiv 2025

[25] [25]

Lee, Y . T. and Vempala, S.,Eldan’s stochastic localization and the KLS hyperplane con- jecture: an improved lower bound for expansion.In: Proceedings of the 58th Annual IEEE Symposium on Foundations of Computer Science (FOCS 2017), IEEE Computer Society, (2017), 998–1007. 22

2017

[26] [26]

7, (1938/39), 118–127

Mahler, K.,Ein Minimalproblem f ¨ur konvexe Polygone, Mathematica (Zutphen) B, V ol. 7, (1938/39), 118–127

1938

[27] [27]

and Zhu, X.,K ¨ahler–Ricci solitons on toric manifolds with positive first Chern class.Adv

Wang, X.-J. and Zhu, X.,K ¨ahler–Ricci solitons on toric manifolds with positive first Chern class.Adv. Math., V ol. 188, no. 1, (2004), 87–103. Seminar for Statistics, Department of Mathematics, ETH Zurich, 8092 Zurich, Switzerland. e-mail:yuansi.chen@stat.math.ethz.ch School of Mathematical Sciences, Tel Aviv University, Tel Aviv 6997801, Israel; and De...

2004