Pi & Planck logoPi & Planck

Geometry and probability · November 3, 2026

The KLS Conjecture

Infographic titled The KLS Conjecture. Left: a disc cut in two equal halves by a flat line of length 2.00 or by a wavy dashed curve of length 2.26, with the statement Var f is at most C times the mean squared gradient, C independent of dimension. Centre: a line chart over dimensions 1 to 1000 of exact Poincare constants for isotropic shapes: Gaussian 1, cube 12 over pi squared, about 1.216, ball falling from 1.216 toward 1 (1.0010 at n=1000). Right: the best known upper bounds by year, from square root of n in 1995 through n to the 1/4 (2017/2024), subpolynomial (2021), polylogarithmic (2022 to 2026), to a constant claimed by three 2026 preprints. Bottom cards: posed in 1995; three preprints claim a proof; verification is ongoing.
Download high resolutionImage 4096×2304 (JPG)PDF for print

In short

A 1995 question: in any dimension, is a flat cut always nearly the cheapest way to split a balanced convex shape? Three proofs were posted in 2026; verification is ongoing. Here are the question and the bounds that led here.

Sources

  1. R. Kannan, L. Lovász, M. Simonovits, Isoperimetric problems for convex bodies and a localization lemma, Discrete & Computational Geometry 13, 541-559 (1995), doi:10.1007/BF02574061
  2. Z. Song, X. Zhang, An O(1) Bound for the KLS Constant, arXiv:2610.01447 (v1 1 Oct 2026, v2 4 Oct 2026; preprint, with Table 1 of previous bounds)
  3. P. Bizeul, B. Klartag, J. Lehec, Presenting a proof of the Kannan-Lovasz-Simonovits conjecture, arXiv:2610.05474 (preprint)
  4. K. Balasubramanian, S. Kasiviswanathan, A Dimension-Free Bound on the Poincare Constant of Isotropic Log-Concave Measures, arXiv:2610.07728 (preprint)
  5. Poincare constants of the cube and ball computed in this project (cube: interval of length 2*sqrt(3), C_P = 12/pi^2; ball: first Neumann eigenvalue via Bessel functions)
  6. Photos: Ken Larsen, cut poudretteite, Smithsonian NMNH (CC0); Pseudopanax, amethyst cluster (public domain)

حدس KLS

اینفوگرافیک «حدس KLS». سمت راست: یک قرص که با خط تخت به طول ۲٫۰۰ یا با منحنی موجی خط‌چین به طول ۲٫۲۶ به دو نیمهٔ هم‌مساحت تقسیم شده، با گزارهٔ واریانس f حداکثر C برابر میانگین مجذور گرادیان است و C به بُعد بستگی ندارد. وسط: نمودار ثابت پوانکارهٔ دقیق شکل‌های ایزوتروپ در بُعد ۱ تا ۱۰۰۰: گاوسی ۱، مکعب ۱۲ بر π² یعنی حدود ۱٫۲۱۶، و گوی که از ۱٫۲۱۶ به سمت ۱ می‌آید (۱٫۰۰۱۰ در n=۱۰۰۰). سمت چپ: بهترین کران‌های شناخته‌شده به تفکیک سال، از ریشهٔ n در ۱۹۹۵ تا n به توان ۱/۴، زیرچندجمله‌ای (۲۰۲۱)، چندلگاریتمی (۲۰۲۲ تا ۲۰۲۶) و کران ثابتِ ادعاشده در سه پیش‌چاپ ۲۰۲۶. کارت‌های پایین: پرسش در ۱۹۹۵؛ سه پیش‌چاپ ادعای اثبات دارند؛ راستی‌آزمایی ادامه دارد.
دانلود با کیفیت بالاتصویر ۴۰۹۶×۲۳۰۴ (JPG)PDF چاپی

خلاصه

پرسشی از ۱۹۹۵: در هر بُعدی، آیا یک برش تخت همیشه تقریباً ارزان‌ترین راه برای دوپاره‌کردن یک شکل محدب متعادل است؟ سه اثبات در ۲۰۲۶ منتشر شده و راستی‌آزمایی ادامه دارد.

منابع

  1. R. Kannan, L. Lovász, M. Simonovits, Isoperimetric problems for convex bodies and a localization lemma, Discrete & Computational Geometry 13, 541-559 (1995), doi:10.1007/BF02574061
  2. Z. Song, X. Zhang, An O(1) Bound for the KLS Constant, arXiv:2610.01447 (v1 1 Oct 2026, v2 4 Oct 2026; preprint, with Table 1 of previous bounds)
  3. P. Bizeul, B. Klartag, J. Lehec, Presenting a proof of the Kannan-Lovasz-Simonovits conjecture, arXiv:2610.05474 (preprint)
  4. K. Balasubramanian, S. Kasiviswanathan, A Dimension-Free Bound on the Poincare Constant of Isotropic Log-Concave Measures, arXiv:2610.07728 (preprint)
  5. Poincare constants of the cube and ball computed in this project (cube: interval of length 2*sqrt(3), C_P = 12/pi^2; ball: first Neumann eigenvalue via Bessel functions)
  6. Photos: Ken Larsen, cut poudretteite, Smithsonian NMNH (CC0); Pseudopanax, amethyst cluster (public domain)

← All posts