OctoInformBeta

OpenAI preprint, September 2026 · §1. Introduction · §1

Theorem 1.1 [main theorem]: For every a,b≥1a, b \ge 1 there are nn and a symmetric convex body K⊂ℝnK \subset \mathbb{R}^n with log⁡N(K,L)>blog⁡N(L∘,a−1K∘)\log N(K,L) > b \log N(L^\circ, a^{-1}K^\circ), where L=[−1,1]nL = [-1,1]^n

The duality conjecture for metric entropy fails.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

For symmetric convex bodies KK and LL in ℝn\mathbb{R}^n, N(K,L)N(K, L) denotes the covering number of KK by translates of LL, and K∘K^\circ the polar body of KK. The duality conjecture for metric entropy asks for universal constants a,b≥1a, b \ge 1 such that log⁡N(K,L)≤blog⁡N(L∘,a−1K∘)\log N(K, L) \le b \log N(L^\circ, a^{-1} K^\circ)(1) in every dimension nn and for all symmetric convex bodies K,L⊂ℝnK, L \subset \mathbb{R}^n, with log\log the natural logarithm. The theorem concerns (1) when LL is the cube [−1,1]n[-1,1]^n and KK is any symmetric convex body.

Polar Bodies and Covering Numbers

A point xx of ℝn\mathbb{R}^n is a real function on {1,…,n}\{1, \dots, n\}, with coordinates x1,…,xnx_1, \dots, x_n.✦ The pairing of two points x,y∈ℝnx, y \in \mathbb{R}^n is the number ⟨x,y⟩=∑i=1nxiyi.\langle x, y \rangle = \sum_{i=1}^{n} x_i y_i .(2)✦ The polar A∘A^\circ of a set A⊆ℝnA \subseteq \mathbb{R}^n is the set A∘={y∈ℝn:⟨x,y⟩≤1 for every x∈A}.A^\circ = \{ y \in \mathbb{R}^n : \langle x, y \rangle \le 1 \text{ for every } x \in A \}.(3)✦ The cube LL in ℝn\mathbb{R}^n is the unit ball of the maximum norm, L=[−1,1]n={x∈ℝn:|xi|≤1 for i=1,…,n},L = [-1,1]^n = \{ x \in \mathbb{R}^n : |x_i| \le 1 \text{ for } i = 1, \dots, n \},(4) and its polar L∘L^\circ is B1B_1, the unit ball of the ℓ1\ell_1 norm.

A symmetric convex body KK in ℝn\mathbb{R}^n is a compact convex subset of ℝn\mathbb{R}^n with nonempty interior such that −x∈K-x \in K for every x∈Kx \in K.✦

Translates x1+B,…,xM+Bx_1 + B, \dots, x_M + B of a set B⊆ℝnB \subseteq \mathbb{R}^n, with centers x1,…,xMx_1, \dots, x_M anywhere in ℝn\mathbb{R}^n, cover a set A⊆ℝnA \subseteq \mathbb{R}^n when every x∈Ax \in A has x−xj∈Bx - x_j \in B for some jj.✦ A set AA is finitely coverable by BB when some finite family of translates of BB covers AA.✦ The covering number N(A,B)N(A, B) of AA by BB is the natural number N(A,B)=min⁡{M≥0:some M translates of B cover A}N(A, B) = \min \{ M \ge 0 : \text{some } M \text{ translates of } B \text{ cover } A \}(5)✦ when AA is finitely coverable by BB, and N(A,B)=0N(A, B) = 0 when it is not.

The Claim

Theorem 1. Let a,b≥1a, b \ge 1 be real numbers. There are a positive integer nn and a symmetric convex body K⊆ℝnK \subseteq \mathbb{R}^n such that, for the cube L=[−1,1]nL = [-1,1]^n, log⁡N(K,L)>blog⁡N(L∘,a−1K∘).\log N(K, L) > b \log N(L^\circ, a^{-1} K^\circ).(6)✦ Moreover, nn and KK can be chosen so that KK is finitely coverable by LL, L∘L^\circ is finitely coverable by a−1K∘a^{-1} K^\circ, and blog⁡N(L∘,a−1K∘)<log⁡N(K,L).b \log N(L^\circ, a^{-1} K^\circ) < \log N(K, L).(7)✦

In the paper, the result reads: for every a,b≥1a, b \ge 1 there are a positive integer nn and an origin-symmetric convex body K⊂ℝnK \subset \mathbb{R}^n such that, with L=[−1,1]nL = [-1,1]^n, log⁡N(K,L)>blog⁡N(L∘,a−1K∘)\log N(K, L) > b \log N(L^\circ, a^{-1} K^\circ).

What Is Checked

The claim is stated over symmetric convex bodies in ℝn\mathbb{R}^n, the cube LL of (4), the polar (3) taken for the pairing (2), finite coverability by translates, and the covering number (5).