A Dimension-Free Bound on the Poincar'e Constant of Isotropic Log-Concave Measures
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.
Ada: I'm Ada.
Emmy: And I'm Emmy, here with Kolya.
Ada: We're looking at the paper: "A Dimension-Free Bound on the Poincar'e Constant of Isotropic Log-Concave Measures".
Emmy: Here's the gist of the paper. The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of dimension.
Ada: This paper provides a deterministic variational proof establishing an explicit bound on the Poincaré constant CP (µ) ≤ 25 for any isotropic log-concave probability measure.
Emmy: The Core Result Theorem 1.1 states that for every n ≥ 1, the supremum of the Poincaré constant CP (µ) over all isotropic log-concave probability measures is bounded by 25.
Ada: This bound is achieved through a deterministic proof starting from elliptic moment estimates and a quadratic variance inequality.
Paper summary: Ada: So we're looking at this paper today called "A Dimension-Free Bound on the Poincaré Constant of Isotropic Log-Concave Measures". It tackles the Kannan–Lovász–Simonovits conjecture.
Emmy: That conjecture essentially asks if isotropic log-concave probability measures have a Poincaré constant bounded by some universal number, regardless of how high up in dimension you go.
Kolya: What's the main claim here? Is it that they finally proved that universal bound?
Ada: They give a deterministic variational proof and establish an explicit bound on the Poincaré constant CP (mu) less than or equal to twenty-five for any isotropic log-concave probability measure <ref:2610.07728#pg1,an explicit bound on the Poincaré constant CP>.
Emmy: That is a very concrete number, twenty-five which they claim holds for every single dimension <ref:2610.07728#pg1>. It’s a universal constant that doesn't change based on the dimension of the space we are working in.
Kolya: Twenty-five seems like a solid number to aim for in these kinds of geometric concentration problems. What kind of machinery did they use to get there?
Ada: They started with elliptic moment estimates and a quadratic variance inequality, then they establish geometric bounds on normalized Appell coefficient norms through joint maximization over both the measure and the test function.
Emmy: It’s interesting how they set up this joint maximization, where varying the measure gives you one kind of bound and varying the test function controls those highest-order cumulant terms.
Kolya: And what about those two concave barriers mentioned? Are those just tools to close the induction? What structure are they imposing here?
Ada: They use two concave barriers built from quadratic and cubic polynomials, and these are what close the induction step without needing separate estimates for higher-order cumulant terms.
Emmy: Then they bring in a weighted Helmholtz–Hodge decomposition to control the curl correction of weighted divergence, which gives them a curvature estimate for compatible symmetric tensor fields that is uniform in rank.
Kolya: That sounds like they are trying to generalize from simple constant tensors to more complex, arbitrary compatible tensor fields. How does that affect the final Poincaré constant bound?
Ada: Lemma six point one shows that the set of projected divergences is dense in Gr, which establishes the required curvature estimate for general compatible fields without needing any auxiliary positive curvature.
Emmy: So they manage to get this control over arbitrary tensor fields using a weighted Helmholtz–Hodge decomposition, which makes it dimension-free.
Kolya: And finally, how does that translate from these coefficient estimates to the actual Poincaré bound? I need to know how the spectral gap connects in words.
Ada: They use Lemma seven point one and Lemma seven point two to control the spectral radius of the integration operator, which leads them to say that the norm bound is at most twenty-four resulting in a final Poincaré constant CP (mu) less than or equal to twenty-five.
Emmy: That recurrence derived from that inequality proves the norm bound is at most twenty-four which gives us that final result of twenty-five <ref:2610.07728#pg1>. It’s a tight control over the spectral radius of the operator LV equals minus delta plus grad V dot grad on L squared mu.
Kolya: So, what does this mean for someone who just listens to the show? Is it just a number now, or does it change how we think about these random measures?
Ada: It means that for any log-concave probability measure you pick in any dimension, you can guarantee its Poincaré constant will be no larger than twenty-five <ref:2610.07728#pg1>.
Emmy: That universal bound is what makes the KLS conjecture true deterministically, and it’s independent of the dimension of the space.
Kolya: The paper also mentions that they show regular isotropic approximants converge weakly to the original measure and their curvature constants satisfy this bound of twenty-five <ref:2610.07728#pg1>. What's that implication for practical applications?
Ada: This convergence means you can use these approximations to extend the Poincaré inequality to every locally Lipschitz function with finite gradient energy, and it also proves its square integrability.
Emmy: That square integrability is important because it means the Poincaré inequality holds in a broader sense across functions, not just specific test functions like the one they used initially.
Kolya: So, to wrap up this paper, what's the big picture regarding the title "A Dimension-Free Bound on the Poincaré Constant of Isotropic Log-Concave Measures"?
Ada: It means they found a way to prove that these measures have a universal Poincaré constant bound of twenty-five regardless of dimension <ref:2610.07728#pg1>.
Emmy: It’s a very strong statement because it settles the KLS conjecture for this class of measures by providing an explicit and dimension-free bound.
Kolya: It also suggests that we can use these tools to control the conductance and mixing of geometric random walks used to sample log-concave distributions, which is relevant for computation.
Ada: And it enters the complexity of rounding and volume algorithms for convex bodies, which are areas where these concentration properties matter a lot.
Conclusion: Ada: So we've seen how they got to the number twenty five for this Poincaré constant in isotropic log-concave measures, regardless of dimension.
Emmy: That paper, "A Dimension-Free Bound on the Poincar'e Constant of Isotropic Log-Concave Measures," it’s actually quite dense. It focuses on proving that this bound holds universally.
Kolya: I mean, twenty five is a specific number for a universal constant. What does that really imply about how these probability measures behave in high dimensions?
Ada: It implies there's a fundamental geometric limit on the concentration of these measures, no matter how much space you have to work in. They used elliptic moment estimates and some joint variational argument to show this bound is tight.
Emmy: The structure they found involves reducing things down to bounds on these coefficient energies, which then connect back to the spectral gap of an operator on the measure itself. It’s a very clever way to link geometry and probability theory.
Kolya: So, for someone listening who isn't a number theorist, what does it mean that this bound is independent of dimension? Does that make these measures behave more predictably than we thought?
Ada: It means the underlying geometric constraints are strong enough to enforce this limit everywhere. They even showed that regular approximations of these measures still stick to that twenty five limit.
Emmy: And they extended the control from just constant tensors to arbitrary compatible tensor fields, which is a big step for general geometry. It shows the mechanism works more broadly than just simple shapes.
Kolya: That's interesting because if it holds for all compatible fields, it suggests a really robust structure underlying these log-concave distributions. It makes me wonder what else this kind of geometric control can be applied to in other areas of mathematics.
Krishnakumar Balasubramanian, Shiva Kasiviswanathan
Amazon · University of California, Davis
math.PR, math.FA, math.MG, math.SP
Submitted: 2026-10-06
Updated: 2026-11-09
Project page: https://pierrebizeul.github.io/optimal-mmstar-bounds.pdf
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 92/100
The gist: The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of dimension.
Key concepts
- Poincaré Constant (CP)
- This constant measures how well a probability measure can be approximated by smooth functions, essentially quantifying the 'smoothness' or regularity of the measure. A smaller Poincaré constant indicates better approximation properties for functions defined on that measure.
- Isotropic Log-Concave Measure
- These are probability distributions where the density function is log-concave (meaning its logarithm is a concave function) and isotropic (meaning it has rotational symmetry). These measures are fundamental objects in geometric functional analysis and often appear in problems related to high-dimensional geometry.
- Variational Framework
- The proof uses a variational approach, maximizing a specific quotient involving the energies of successive coefficients of the measure. This method allows researchers to derive complementary bounds from optimizing both the measure itself and an auxiliary test function.
Terminology
Summary
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of dimension. p. 0 edu kasivisw@gmail. com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. We give a deterministic variational proof with an explicit bound on the Poincaré
constant CP (µ) ≤ 25, where µ is any isotropic log-concave probability measure.Where the paper says it
Where the paper says it
The Core Result
Theorem 1.1 states that for every n ≥ 1, the supremum of the Poincaré constant CP (µ) over all isotropic log-concave probability measures is bounded by 25. p. 0 edu kasivisw@gmail. com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. We give a deterministic variational proof with an explicit bound on the Poincaré
constant CP (µ) ≤ 25, where µ is any isotropic log-concave probability measure. Starting from
elliptic moment estimates and a quadratic variance inequality, we establish geometric bounds
on normalized Appell coefficient norms through joint maximization over the measure and test
function. A spectral-radius estimate and a scalar growth inequality
for adjoint iterates then yield the Poincaré bound. The final estimate is independent of the
auxiliary positive curvature, allowing approximation to complete the proof for general isotropic
log-concave measures.Where the paper says it
Where the paper says it
Where the paper says it
The Variational Framework
The proof employs a joint variational argument over the measure and the test function to yield bounds on tilt coefficients. This structure involves maximizing a covariancenormalized quotient of successive coefficient energies over both the law and the test function. p. 3 Our induction follows this variational structure. At each step, we maximize a covariancenormalized quotient of successive coefficient energies over both the law and the test function. Varying
the measure gives a maximum-principle inequality, while varying the test function gives an identity
for the two highest-order cumulant terms in that inequality. This identity avoids separate estimates
for those cumulants.Where the paper says it
Where the paper says it
The Coefficient Induction
The induction proceeds by maximizing a covariancenormalized quotient of successive coefficient energies over both the measure and the test function. This process reduces the induction step to a scalar condition on a positive concave barrier. p. 3 The remaining induction reduces
to a scalar inequality for a positive concave barrier. A quadratic polynomial certifies the degree-three
step, and a single cubic polynomial handles all higher degrees. The lower-degree bounds control the remaining terms, reducing the induction
step to a scalar condition on the concave function. The next section constructs these denominators
by inverting decreasing branches of explicit quadratic and cubic polynomials.Where the paper says it
Where the paper says it
The Spectral Gap Connection
The transition from coefficient estimates to a Poincaré bound is achieved through the inverse gradient, which relates it to the spectral gap of the operator LV = −∆ + ∇V · ∇ on L2(µ). A weighted Helmholtz–Hodge decomposition controls the curl correction of weighted divergence, yielding a curvature estimate for compatible symmetric tensor fields that is uniform in rank. p. 0 Two concave barriers constructed
from quadratic and cubic polynomials close the induction. A weighted Helmholtz–Hodge decomposition then controls the curl correction of weighted divergence, yielding a curvature estimate
for compatible symmetric tensor fields that is uniform in rank. Together with the density of projected divergences (Lemma 6. 1), the Hodge estimate permits
quadratic duality and yields the operator comparison
J
∗J ⪯ ga(JJ∗ + LL∗
), ga(t) = t
1 + at
.Where the paper says it
Where the paper says it
The Geometric Extension
The analysis extends the control from constant tensors to arbitrary compatible tensor fields using a weighted Helmholtz–Hodge decomposition. Lemma 6.1 proves that the set of projected divergences is dense in Gr, which establishes the required curvature estimate for general compatible fields. p. 4 An approximation argument therefore removes the auxiliary curvature and regularity
assumptions, completing the proof for general isotropic log-concave laws without a localization step. 4 Organization and Conventions
Section 2 proves the required third-moment estimate. A spectral-radius estimate and a scalar growth inequality
for adjoint iterates then yield the Poincaré bound. The final estimate is independent of the
auxiliary positive curvature, allowing approximation to complete the proof for general isotropic
log-concave measures.Where the paper says it
Where the paper says it
The Final Bound
The Poincaré bound is established by applying Lemma 7.1 and Lemma 7. p. 3 1, Theorem 2. Section 2 combines weighted moment-map bound with the third-moment identity by Chen and
Klartag (2026, Lemma 3. Quadratic duality converts this
estimate into an operator comparison for centered integration, linking the coefficient bounds
to control of the inverse gradient. A spectral-radius estimate and a scalar growth inequality
for adjoint iterates then yield the Poincaré bound. Sections 5 and 6 establish the
Hodge estimate and the operator comparison for centered integration on compatible tensor fields. Section 7 proves the operator bound, and Section 8 completes the theorem. A spectral-radius estimate and a scalar growth inequality
for adjoint iterates then yield the Poincaré bound. The final estimate is independent of the
auxiliary positive curvature, allowing approximation to complete the proof for general isotropic
log-concave measures.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Approximation and Convergence
The paper concludes by showing that regular isotropic approximants converge weakly to the original measure, and their curvature constants satisfy the bound 25. This allows the Poincaré inequality to be extended to every locally Lipschitz function with finite gradient energy and also proves its square integrability. p. 0 Here the inner supremum is over locally Lipschitz
f ∈ L
2
(µ) with finite, nonzero gradient energy. For a fixed µ, this inner supremum (over f) is its
Poincaré constant CP (µ). A spectral-radius estimate and a scalar growth inequality
for adjoint iterates then yield the Poincaré bound. The final estimate is independent of the
auxiliary positive curvature, allowing approximation to complete the proof for general isotropic
log-concave measures.Where the paper says it
Where the paper says it
References
M. Anshelevich, Appell polynomials and their relatives, International Mathematics Research Notices 2004, no. p. 4 The Brascamp–Lieb
inequality Carlen, Cordero-Erausquin, and Lieb (2013) gives the preliminary bound CP (µ) ≤ a
−1
and ensures that the inverse gradients are bounded. Repeated centered integration of constant
tensors produces the classical Appell polynomials (Anshelevich, 2004).Where the paper says it
D. N. Arnold, R. S. Falk, and R. Winther, Finite element exterior calculus, homological techniques, and applications Acta Numerica 15 (2006), 1–155.
K. Balasubramanian, Finite-particle convergence rates for conservative and non-conservative drifting models arXiv:2605. p. 5 Compatible integration operators and an operator norm
estimate then turn the coefficient bounds into a spectral gap. Three AI interactions are worth
emphasizing:
• Motivated by the use of Helmholtz–Hodge decomposition in the analysis of drifting models (Balasubramanian, 2026), we asked the AI whether a similar decomposition could help control
curl in our proposed KLS argument. 07728v2 [math. PR] 8 Oct 2026Where the paper says it
Where the paper says it
K. Balasubramanian, S. p. 5 Compatible integration operators and an operator norm
estimate then turn the coefficient bounds into a spectral gap. Three AI interactions are worth
emphasizing:
• Motivated by the use of Helmholtz–Hodge decomposition in the analysis of drifting models (Balasubramanian, 2026), we asked the AI whether a similar decomposition could help control
curl in our proposed KLS argument. Our induction follows this variational structure. At each step, we maximize a covariancenormalized quotient of successive coefficient energies over both the law and the test function. • We also asked the AI whether the Fourier coefficient strategy of Balasubramanian et al. (2025)
could inform the KLS argument.Where the paper says it
Where the paper says it
Where the paper says it
P. Bizeul, Optimal MM∗ bounds for convex bodies https://arxiv. p. 1 KLS bounds have consequences for both probability and computation. They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024). 07728v2 [math. PR] 8 Oct 2026Where the paper says it
Where the paper says it
P. Bizeul, B. Klartag, and J. p. 1 For isotropic X ∼ µ, the test function
f(x) = |x|
2 gives Var(|X|
2
) ≤ 4nCP (µ). This connects KLS to the thin-shell problem and, through
thin-shell concentration, to the central limit phenomenon for convex bodies (Klartag, 2007). com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. We give a deterministic variational proof with an explicit bound on the Poincaré
constant CP (µ) ≤ 25, where µ is any isotropic log-concave probability measure. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610. 07728v2 [math. PR] 8 Oct 2026Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
S. G. Bobkov, On isoperimetric constants for log-concave probability distributions in Geometric Aspects of Functional Analysis, Lecture Notes in Mathematics 1910, Springer, 2007, pp. p. 1 KLS bounds have consequences for both probability and computation. They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024).Where the paper says it
C. Borell, Convex set functions in d-space, Periodica Mathematica Hungarica 6 (1975), 111–136 doi:10. p. 5 The affine support is the affine hull
of the support, and gradients there are intrinsic. The density and set definitions of log-concavity
agree on that space by Borell’s characterization (Borell, 1975).Where the paper says it
E. A. Carlen, D. Cordero-Erausquin, and E. p. 1 1) subsequently established the sharp inequality Var(|X|
2
) ≤ 8n for isotropic log-concave
X by an elliptic method. Their proof analyzes the Monge–Ampère equation for moment measures,
using the construction of Cordero-Erausquin and Klartag (2015) and the associated weighted
geometry developed by Klartag (2014). For these operator constructions, we first assume D2V ⪰ aI for some a > 0. The Brascamp–Lieb
inequality Carlen, Cordero-Erausquin, and Lieb (2013) gives the preliminary bound CP (µ) ≤ a
−1
and ensures that the inverse gradients are bounded.Where the paper says it
Where the paper says it
P. Cattiaux and A. Guillin, A journey with the integrated Γ2 criterion and its weak forms https://arxiv.org/abs/2201.07475v1 (2022). p. 1 Lee and Vempala (2024)
obtained ψn ≲ n
1/4
, and in a landmark work, Chen (2021) established the first subpolynomial
bound, ψn = n
o(1)
. Klartag and Lehec (2022) proved ψn ≲ (log n)
5
, giving the first polylogarithmic bound.Where the paper says it
Y. Chen, An almost constant lower bound of the isoperimetric coefficient in the KLS conjecture Geometric and Functional Analysis 31 (2021), no. p. 0 For a fixed µ, this inner supremum (over f) is its
Poincaré constant CP (µ). The original geometric formulation of KLS asks whether hyperplane cuts approximate the
optimal isoperimetric ratio up to a universal factor. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610. 1 Related Work and Progress
The seminal deterministic localization argument of Kannan, Lovász, and Simonovits (1995) gave
ψn ≲
√
n. Bobkov’s comparison between isoperimetry and fluctuations of the Euclidean norm
Bobkov (2007), combined with the thin-shell estimate of Guédon and Milman (2011), improved this
to ψn ≲ n
5/12
.Where the paper says it
Where the paper says it
Where the paper says it
Y. Chen and B. p. 1 Subsequent general KLS bounds developed this stochastic approach. Lee and Vempala (2024)
obtained ψn ≲ n
1/4
, and in a landmark work, Chen (2021) established the first subpolynomial
bound, ψn = n
o(1)
. 1) subsequently established the sharp inequality Var(|X|
2
) ≤ 8n for isotropic log-concave
X by an elliptic method. Their proof analyzes the Monge–Ampère equation for moment measures,
using the construction of Cordero-Erausquin and Klartag (2015) and the associated weighted
geometry developed by Klartag (2014).Where the paper says it
Where the paper says it
D. Cordero-Erausquin and B. p. 1 1) subsequently established the sharp inequality Var(|X|
2
) ≤ 8n for isotropic log-concave
X by an elliptic method. Their proof analyzes the Monge–Ampère equation for moment measures,
using the construction of Cordero-Erausquin and Klartag (2015) and the associated weighted
geometry developed by Klartag (2014). 1) subsequently established the sharp inequality Var(|X|
2
) ≤ 8n for isotropic log-concave
X by an elliptic method. Their proof analyzes the Monge–Ampère equation for moment measures,
using the construction of Cordero-Erausquin and Klartag (2015) and the associated weighted
geometry developed by Klartag (2014). Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
Where the paper says it
M. Costabel and A. McIntosh, On Bogovski˘ı and regularized Poincaré integral operators for de Rham complexes on Lipschitz domains Mathematische Zeitschrift 265 (2010), no. p. 0 Here the inner supremum is over locally Lipschitz
f ∈ L
2
(µ) with finite, nonzero gradient energy. For a fixed µ, this inner supremum (over f) is its
Poincaré constant CP (µ). Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
R. Eldan, Thin shell implies spectral gap up to polylog via a stochastic localization scheme Geometric and Functional Analysis 23 (2013), no. p. 1 Bobkov’s comparison between isoperimetry and fluctuations of the Euclidean norm
Bobkov (2007), combined with the thin-shell estimate of Guédon and Milman (2011), improved this
to ψn ≲ n
5/12
. Eldan (2013) introduced stochastic localization and related thin-shell estimates to
the spectral gap up to logarithmic factors; together with Guédon and Milman (2011), this yielded
ψn ≲ n
1/3√
log n. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
Q. Guan, A note on Bourgain’s slicing problem https://arxiv. p. 1 Progress on the thin-shell problem supplied stronger moment estimates. Building on the
covariance estimates by (Guan, 2024), Klartag and Lehec (2026) proved a dimension-free thin-shell
bound using a parallel coupling associated with stochastic localization. KLS bounds have consequences for both probability and computation. They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024).Where the paper says it
Where the paper says it
O. Guédon and E. p. 1 1 Related Work and Progress
The seminal deterministic localization argument of Kannan, Lovász, and Simonovits (1995) gave
ψn ≲
√
n. Bobkov’s comparison between isoperimetry and fluctuations of the Euclidean norm
Bobkov (2007), combined with the thin-shell estimate of Guédon and Milman (2011), improved this
to ψn ≲ n
5/12
. Bobkov’s comparison between isoperimetry and fluctuations of the Euclidean norm
Bobkov (2007), combined with the thin-shell estimate of Guédon and Milman (2011), improved this
to ψn ≲ n
5/12
. Eldan (2013) introduced stochastic localization and related thin-shell estimates to
the spectral gap up to logarithmic factors; together with Guédon and Milman (2011), this yielded
ψn ≲ n
1/3√
log n.Where the paper says it
Where the paper says it
A. Jambulapati, Y. p. 1 Klartag and Lehec (2022) proved ψn ≲ (log n)
5
, giving the first polylogarithmic bound. Jambulapati, Lee, and Vempala (2022) improved the exponent to 3. KLS bounds have consequences for both probability and computation. They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024). KLS bounds have consequences for both probability and computation. They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024). Klartag and Lehec (2022) proved ψn ≲ (log n)
5
, giving the first polylogarithmic bound. Jambulapati, Lee, and Vempala (2022) improved the exponent to 3. Lee and Vempala (2024)
obtained ψn ≲ n
1/4
, and in a landmark work, Chen (2021) established the first subpolynomial
bound, ψn = n
o(1)
. Klartag and Lehec (2022) proved ψn ≲ (log n)
5
, giving the first polylogarithmic bound.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
H. Jia, A. Laddha, Y. T. Lee, and S. S. Vempala, Reducing isotropy and volume to KLS: faster rounding and volume algorithms https://arxiv. p. 1 KLS bounds have consequences for both probability and computation. They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024). KLS bounds have consequences for both probability and computation. They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024).Where the paper says it
Where the paper says it
J. Johnsen, On the spectral properties of Witten-Laplacians, their range projections and Brascamp–Lieb’s inequality Integral Equations and Operator Theory 36 (2000), 288–324 doi:10. p. 4 For these operator constructions, we first assume D2V ⪰ aI for some a > 0. The Brascamp–Lieb
inequality Carlen, Cordero-Erausquin, and Lieb (2013) gives the preliminary bound CP (µ) ≤ a
−1
and ensures that the inverse gradients are bounded.Where the paper says it
R. Kannan, L. Lovász, and M. p. 0 edu kasivisw@gmail. com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. This connects KLS to the thin-shell problem and, through
thin-shell concentration, to the central limit phenomenon for convex bodies (Klartag, 2007). 1 Related Work and Progress
The seminal deterministic localization argument of Kannan, Lovász, and Simonovits (1995) gave
ψn ≲
√
n.Where the paper says it
Where the paper says it
T. Kato, Perturbation Theory for Linear Operators, corrected printing of the second edition, Springer, 1980; reprinted in Classics in Mathematics, 1995 doi:10. p. 9 6) annihilates the
expectations of each fixed coordinate and each fixed quadratic monomial, while the product rule
contributes P
i
(∂im)(∂im)
T = I to D(mmT
). The second identity follows by differentiating the
covariance in a linear-tilt direction.Where the paper says it
B. Klartag, A central limit theorem for convex sets Inventiones Mathematicae 168 (2007), no. p. 1 For isotropic X ∼ µ, the test function
f(x) = |x|
2 gives Var(|X|
2
) ≤ 4nCP (µ). This connects KLS to the thin-shell problem and, through
thin-shell concentration, to the central limit phenomenon for convex bodies (Klartag, 2007).Where the paper says it
B. Klartag, Logarithmically-concave moment measures I, in Geometric Aspects of Functional Analysis, Lecture Notes in Mathematics 2116, Springer, 2014 pp. p. 1 1) subsequently established the sharp inequality Var(|X|
2
) ≤ 8n for isotropic log-concave
X by an elliptic method. Their proof analyzes the Monge–Ampère equation for moment measures,
using the construction of Cordero-Erausquin and Klartag (2015) and the associated weighted
geometry developed by Klartag (2014). Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
B. Klartag, Logarithmic bounds for isoperimetry and slices of convex sets Ars Inveniendi Analytica (2023), Paper No. p. 1 Bobkov’s comparison between isoperimetry and fluctuations of the Euclidean norm
Bobkov (2007), combined with the thin-shell estimate of Guédon and Milman (2011), improved this
to ψn ≲ n
5/12
. Eldan (2013) introduced stochastic localization and related thin-shell estimates to
the spectral gap up to logarithmic factors; together with Guédon and Milman (2011), this yielded
ψn ≲ n
1/3√
log n. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
B. Klartag and J. p. 1 For isotropic X ∼ µ, the test function
f(x) = |x|
2 gives Var(|X|
2
) ≤ 4nCP (µ). This connects KLS to the thin-shell problem and, through
thin-shell concentration, to the central limit phenomenon for convex bodies (Klartag, 2007). (2025)
could inform the KLS argument. We suggested organizing the functional-analytic problem
around a tractable hierarchy of coefficients controlled by recursive bounds. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
Where the paper says it
B. Klartag and E. p. 1 For isotropic X ∼ µ, the test function
f(x) = |x|
2 gives Var(|X|
2
) ≤ 4nCP (µ). This connects KLS to the thin-shell problem and, through
thin-shell concentration, to the central limit phenomenon for convex bodies (Klartag, 2007). There, this relation compares variances along a stochastic tilt, and
covariance estimates along the process lead to a spectral-gap bound. Related deterministic heat-flow
identities appear in Klartag and Putterman (2023, Sections 2 and 4.Where the paper says it
Where the paper says it
Y. T. Lee and S. S. Vempala, Eldan’s stochastic localization and the KLS conjecture: isoperimetry, concentration and mixing Annals of Mathematics 199 (2024), no. p. 1 They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024). A uniform spectral gap also yields uniform exponential
concentration of Lipschitz test functions (Milman, 2009). Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
B. Letwin, The KLS constant is O(log 1/4 n) https://arxiv. p. 0 edu kasivisw@gmail. com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. 07728v2 [math. PR] 8 Oct 2026Where the paper says it
Where the paper says it
E. S. Meckes and M. W. Meckes, On the equivalence of modes of convergence for log-concave measures in Geometric Aspects of Functional Analysis, Springer, 2014 pp. p. 2 1)
where the inner supremum is over locally Lipschitz f ∈ L
2
(µ) with nonzero finite gradient energy. The Poincaré–Cheeger equivalence for log-concave measures (Milman, 2009, Theorem 1. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
E. Milman, On the role of convexity in isoperimetry, spectral gap and concentration Inventiones Mathematicae 177 (2009), no. p. 1 They control the
conductance and mixing of geometric random walks used to sample log-concave distributions (Lee
and Vempala, 2024), and enter the complexity of rounding and volume algorithms for convex bodies
(Jia, Laddha, Lee, and Vempala, 2024). A uniform spectral gap also yields uniform exponential
concentration of Lipschitz test functions (Milman, 2009). Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
M. H. Protter and H. C. Weinberger, Maximum Principles in Differential Equations, Springer, New York, 1984 doi:10. p. 3 Quantitative Coefficient Bounds. Maximum principles also enter the elliptic estimates underlying the induction.Where the paper says it
O. Roustant, F. Barthe, and B. Iooss, Poincaré inequalities on intervals—application to sensitivity analysis Electronic Journal of Statistics 11 (2017), no. p. 0 edu kasivisw@gmail. com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. Its connection to the formulation above is
∗This work does not relate to the authors’ positions at Amazon. 1
arXiv:2610.Where the paper says it
Where the paper says it
Z. Song and X. p. 2 1
and its proof). During the final stages of preparation of this manuscript, a recent work by Song and Zhang
(2026, Version 1, Theorem 7. 1
and its proof). During the final stages of preparation of this manuscript, a recent work by Song and Zhang
(2026, Version 1, Theorem 7. com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. We give a deterministic variational proof with an explicit bound on the Poincaré
constant CP (µ) ≤ 25, where µ is any isotropic log-concave probability measure.Where the paper says it
Where the paper says it
Where the paper says it
Z. Song and X. Zhang, An O(1) bound for the KLS constant https://arxiv. p. 0 com
1Amazon∗
2Department of Statistics, University of California, Davis
Abstract
The Kannan–Lovász–Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of
dimension. We give a deterministic variational proof with an explicit bound on the Poincaré
constant CP (µ) ≤ 25, where µ is any isotropic log-concave probability measure.Where the paper says it
A Compactness of Isotropic Log-concave Laws. p. 4 The final bound is independent of a, and no constant from fixed-dimensional compactness
enters it. An approximation argument therefore removes the auxiliary curvature and regularity
assumptions, completing the proof for general isotropic log-concave laws without a localization step.Where the paper says it
B Primitives and Operator Domains. p. 0 A weighted Helmholtz–Hodge decomposition then controls the curl correction of weighted divergence, yielding a curvature estimate
for compatible symmetric tensor fields that is uniform in rank. Quadratic duality converts this
estimate into an operator comparison for centered integration, linking the coefficient bounds
to control of the inverse gradient.Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Automated proof construction for KLS bounds: The system can construct deterministic variational proofs by iteratively maximizing a covariance-normalized quotient over both the measure and a test function, using maximum principles to close induction steps as described in Section 3.
-
Dimensionality-free Poincaré constant estimation: The improved system can estimate the Poincaré constant CP (mu) for isotropic log-concave measures by utilizing the final operator comparison derived from quadratic duality and spectral radius estimates, yielding a bound of 25 independent of dimension.
-
Geometric bound verification via tensor field analysis: The system can verify curvature estimates for compatible symmetric tensor fields using weighted Helmholtz–Hodge decompositions to control the curl correction term, ensuring uniformity in tensor rank as shown in Section 4.
-
Coefficient induction and barrier construction: The AI can automatically construct the required concave barriers by inverting explicit polynomials at each step of the induction, allowing it to verify if a proposed bound satisfies the necessary criterion for all degrees d greater than 3.
-
Spectral gap estimation from coefficient bounds: The system can translate derived coefficient bounds into a spectral gap estimate by applying the inverse gradient and using a curl-correction argument to establish an operator comparison for centered integration on compatible tensor fields.
-
Approximation of general log-concave laws: The system can approximate the Poincaré inequality for any locally Lipschitz function with finite gradient energy by utilizing regular isotropic approximants that converge weakly, ensuring the bound holds for all isotropic log-concave laws.
Abstract
The Kannan--Lovász--Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincaré constants bounded by a universal constant, independently of dimension. In this work, we provide a deterministic variational proof that the Poincaré constant C P(μ) at most25, where μ is any isotropic log-concave probability measure. The proof centers on mean-zero integration of compatible symmetric tensor fields. Starting from constant tensors, we obtain factorially normalized Appell polynomials by iterating this procedure. Using moment estimates and a quadratic variance inequality, we establish geometric bounds on these iterated primitives through a joint variational argument over the measure and test function. At a maximizing pair, stationarity in the test function controls the weighted sum of the two highest-order cumulant terms arising from the measure variation. To extend these bounds to arbitrary compatible fields, we use a weighted Helmholtz--Hodge decomposition of weighted divergence, the formal adjoint of differentiation. Controlling the curl in the decomposition yields a curvature comparison for the integration operators, uniformly in tensor order. Combining this comparison with the Appell bounds gives a bound on the full integration operator and hence the stated Poincaré inequality. The estimate is independent of the auxiliary positive curvature, allowing approximation to general isotropic log-concave measures.
Sources
- Appell polynomials and their relatives
- Finite-Particle Convergence Rates for Conservative and Non-Conservative Drifting Models
- On the Structure of Stationary Solutions to McKean-Vlasov Equations with Applications to Noisy Transformers
- Optimal MM* bounds for convex bodies
- A journey with the integrated $\gamma$2 criterion and its weak forms
- Moment Measures
- On Bogovski\u{\i} and regularized Poincar'e integral operators for de Rham complexes on Lipschitz domains
- Thin shell implies spectral gap up to polylog via a stochastic localization scheme
- A note on Bourgain's slicing problem
- A Slightly Improved Bound for the KLS Constant
- Reducing Isotropy and Volume to KLS: Faster Rounding and Volume Algorithms
- On the spectral properties of Witten-Laplacians, their range projections and Brascamp-Lieb's inequality
- A Central Limit Theorem for Convex Sets
- Logarithmically-concave moment measures I
- Logarithmic bounds for isoperimetry and slices of convex sets
- Bourgain's slicing problem and KLS isoperimetry up to polylog
- The KLS constant is $O(\log^{1/4} n)$
- On the equivalence of modes of convergence for log-concave measures
- On the role of Convexity in Isoperimetry, Spectral-Gap and Concentration
Related papers
- Rank-One HCIZ Integrals and Multiplicative Convolutions
- Repeated averaging on random d-regular graphs
- Precise cover times for branching random walks on Hamming graphs: (iterated) logarithmic corrections
- Markov Chain CLTs: Resolving Open Problems
- Marked stable L'evy processes in Brownian local time
- A Sharp Refinement of Samuels' Inequality via Cap Moves