OctoInformBeta

OpenAI preprint, September 2026 · §1. Introduction

Theorem 1.1: Uniform estimates over compact target containers, with three-atom counterexamples to every larger exponent

Brenier maps have sharp one-third stability.

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

The question is how quadratic optimal transport maps from a fixed uniform source change as the target measure varies. In the real Euclidean space ℝd\mathbb R^d of dimension dd, let K⊂ℝdK\subset\mathbb R^d be a compact convex body with nonempty interior, and let Y⊂ℝdY\subset\mathbb R^d be a nonempty compact target container. The uniform measure ρ\rho on KK is the normalized restriction of Lebesgue measure ℒd\mathcal L^d given by (1). ρ=1ℒd(K)ℒd↾K.\rho=\frac{1}{\mathcal L^d(K)}\,\mathcal L^d\!\restriction_K.(1)✦ A measure μ\mu is supported in YY if μ(Yc)=0\mu(Y^c)=0.✦

A coupling π\pi of probability measures μ,ν\mu,\nu on ℝd\mathbb R^d is a probability measure on ℝd×ℝd\mathbb R^d\times\mathbb R^d with first marginal μ\mu and second marginal ν\nu.✦ The quadratic transport cost 𝒞(π)\mathcal C(\pi) of a coupling π\pi is the nonnegative extended real number in (2). 𝒞(π)=∫ℝd×ℝd‖x−y‖2dπ(x,y).\mathcal C(\pi)=\int_{\mathbb R^d\times\mathbb R^d} \lVert x-y\rVert^2\,d\pi(x,y).(2) The quadratic Wasserstein distance W2(μ,ν)W_2(\mu,\nu) between probability measures μ,ν\mu,\nu supported in YY is the nonnegative real number given by (3). W2(μ,ν)=(infπ a coupling of μ,ν𝒞(π))1/2.W_2(\mu,\nu)= \left(\inf_{\pi\text{ a coupling of }\mu,\nu}\mathcal C(\pi)\right)^{1/2}.(3)✦

The graph transport plan πT\pi_T of a measurable map T:ℝd→ℝdT:\mathbb R^d\to\mathbb R^d from ρ\rho is the pushforward in (4). πT=(id⁡,T)#ρ.\pi_T=(\operatorname{id},T)_\#\rho.(4)✦ A quadratic optimal transport map, or Brenier map, TT from ρ\rho to μ\mu is a measurable map with T#ρ=μT_\#\rho=\mu whose graph plan minimizes 𝒞\mathcal C among all couplings of ρ,μ\rho,\mu. A unique quadratic optimal transport map TμT_\mu from ρ\rho to μ\mu is a quadratic optimal transport map whose graph plan is the only minimizing coupling.✦ The L2(ρ)L^2(\rho) distance ‖T−S‖L2(ρ)\lVert T-S\rVert_{L^2(\rho)} between maps T,S:ℝd→ℝdT,S:\mathbb R^d\to\mathbb R^d is the nonnegative extended real number given by (5). ‖T−S‖L2(ρ)=(∫ℝd‖T(x)−S(x)‖2dρ(x))1/2.\lVert T-S\rVert_{L^2(\rho)}= \left(\int_{\mathbb R^d}\lVert T(x)-S(x)\rVert^2\,d\rho(x)\right)^{1/2}.(5)✦

A probability measure μ\mu has exactly three atoms in YY if it has the form (6), where δy\delta_y denotes the Dirac mass at yy, y1,y2,y3∈Yy_1,y_2,y_3\in Y are pairwise distinct sites and m1,m2,m3m_1,m_2,m_3 are positive weights. μ=∑i=13miδyi,∑i=13mi=1.\mu=\sum_{i=1}^3m_i\delta_{y_i},\qquad \sum_{i=1}^3m_i=1.(6) The cube K▫=[−1,1]dK_{\square}=[-1,1]^d consists of the points x∈ℝdx\in\mathbb R^d with |xi|≤1|x_i|\le1 in every coordinate✦; its uniform measure ρ▫\rho_{\square} is defined by (1) with K=K▫K=K_{\square}.✦

The Claim

Theorem 1. Let d≥2d\ge2, let K⊂ℝdK\subset\mathbb R^d be compact and convex with nonempty interior, and let Y⊂ℝdY\subset\mathbb R^d be nonempty and compact. There is a constant C=C(d,K,Y)≥0C=C(d,K,Y)\ge0 such that any probability measures μ,ν\mu,\nu supported in YY admit unique quadratic optimal transport maps Tμ,TνT_\mu,T_\nu from ρ\rho satisfying (7). ‖Tμ−Tν‖L2(ρ)≤CW2(μ,ν)1/3.\lVert T_\mu-T_\nu\rVert_{L^2(\rho)} \le C\,W_2(\mu,\nu)^{1/3}.(7)✦

Let d≥2d\ge2, α>1/3\alpha>1/3 and C>0C>0. There are distinct probability measures μ,ν\mu,\nu, each with exactly three atoms in K▫K_{\square}, that admit unique quadratic optimal transport maps Tμ,TνT_\mu,T_\nu from ρ▫\rho_{\square} and satisfy (8). W2(μ,ν)>0,CW2(μ,ν)α<‖Tμ−Tν‖L2(ρ▫).\begin{gathered} W_2(\mu,\nu)>0,\\ C\,W_2(\mu,\nu)^\alpha <\lVert T_\mu-T_\nu\rVert_{L^2(\rho_{\square})}. \end{gathered}(8)✦

What Is Checked

The statements concern normalized Lebesgue sources on compact convex bodies with nonempty interior in ℝd\mathbb R^d, d≥2d\ge2, and probability targets supported in a nonempty compact container, with the sharpness statement restricted to cube sources and targets with exactly three atoms.