OctoInformBeta

OpenAI preprint, September 2026 · §3. Pair mass below a fixed scale

Definition 3.1: Finite laws admit optimal feasible pair weights and a cutoff beyond which their capacities vanish

Maximum pair mass at a dyadic scale.

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 pairing problem at a fixed scale concerns the mass that a finite probability law can place on pairs of atoms in a dyadic distance band.

The Claim

Definition 1. A finite probability law ν\nu on ℝ\mathbb R consists of a finite support F⊂ℝF\subset\mathbb R and atom masses (px)x∈F(p_x)_{x\in F} satisfying (1). px>0(x∈F),∑x∈Fpx=1.p_x>0\quad(x\in F),\qquad \sum_{x\in F}p_x=1.(1)✦ For a finite probability law ν\nu with support FF and an integer ℓ\ell, the set Eℓ(ν)E_\ell(\nu) of unordered pairs of points of FF at scale ℓ\ell is given by (2). Eℓ(ν)={{x,y}:x,y∈F,2−ℓ≤|x−y|<21−ℓ}.E_\ell(\nu)= \bigl\{\{x,y\}:x,y\in F,\ 2^{-\ell}\leq |x-y|<2^{1-\ell}\bigr\}.(2)✦ For a finite probability law ν\nu, an integer ℓ\ell and pair weights w:Eℓ(ν)→ℝw:E_\ell(\nu)\to\mathbb R, the incident pair mass Iw(x)I_w(x) at a point x∈ℝx\in\mathbb R is the real number given by (3). Iw(x)=12∑e∈Eℓ(ν)x∈ewe.I_w(x)=\frac12\sum_{\substack{e\in E_\ell(\nu)\\x\in e}}w_e.(3)✦ The set 𝒲ℓ(ν)\mathcal W_\ell(\nu) of feasible pair weights for a finite probability law ν\nu at integer scale ℓ\ell is the set of functions specified in (4). 𝒲ℓ(ν)={w:Eℓ(ν)→ℝ:we≥0for every e∈Eℓ(ν),Iw(x)≤pxfor every x∈F}.\mathcal W_\ell(\nu)= \bigl\{w:E_\ell(\nu)\to\mathbb R: w_e\geq0\ \text{for every }e\in E_\ell(\nu),\quad I_w(x)\leq p_x\ \text{for every }x\in F\bigr\}.(4)✦ For a finite probability law ν\nu and an integer ℓ\ell, the pair mass capacity cℓ(ν)c_\ell(\nu) is the supremum in (5). cℓ(ν)=supw∈𝒲ℓ(ν)∑e∈Eℓ(ν)we.c_\ell(\nu)=\sup_{w\in\mathcal W_\ell(\nu)} \sum_{e\in E_\ell(\nu)}w_e.(5)✦ For every integer ℓ\ell, there are feasible weights w∈𝒲ℓ(ν)w\in\mathcal W_\ell(\nu) satisfying (6). ∑e∈Eℓ(ν)we=cℓ(ν).\sum_{e\in E_\ell(\nu)}w_e=c_\ell(\nu).(6)✦ Given any integer aa, there is an integer B≥aB\geq a for which (7) holds. cℓ(ν)=0for every integer ℓ>B.c_\ell(\nu)=0\qquad\text{for every integer }\ell>B.(7)✦

In the paper, the result reads: for a probability law ν\nu with finite support F⊂ℝF\subset\mathbb R and ℓ∈ℤ\ell\in\mathbb Z, let ℰℓ(F)\mathcal E_\ell(F) be the unordered pairs {x,y}\{x,y\} satisfying 2−ℓ≤|x−y|<21−ℓ2^{-\ell}\leq|x-y|<2^{1-\ell}; then cℓ(ν)c_\ell(\nu) is the maximum of ∑e∈ℰℓ(F)we\sum_{e\in\mathcal E_\ell(F)}w_e over weights we≥0w_e\geq0 satisfying 12∑e∈ℰℓ(F):x∈ewe≤ν({x})\frac12\sum_{e\in\mathcal E_\ell(F):\,x\in e}w_e\leq\nu(\{x\}) for each x∈Fx\in F, and is zero when ℰℓ(F)=⌀\mathcal E_\ell(F)=\varnothing.

What Is Checked

The statements concern finite probability laws on ℝ\mathbb R and their feasible pair weights at integer dyadic scales.