OctoInformBeta

OpenAI preprint, September 2026 · Section 4. An interval-pair inequality

Section 4 restricted kernel and endpoint averages: Averaging in each variable for bounded and unbounded open intervals

Restricted Newton kernels and endpoint averages.

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 geometry of an interval pair uses a kernel on the real line together with endpoint averages in the first and second variables.

The Definition

Definition 1. Let n∈ℕn\in\mathbb N and λ,c∈ℝ\lambda,c\in\mathbb R. The restricted Newton kernel kn,λ,c:ℝ→ℝk_{n,\lambda,c}:\mathbb R\to\mathbb R is the function in (1): kn,λ,c(h)=c(λ2+h2)(2−n)/2,h∈ℝ.k_{n,\lambda,c}(h)=c(\lambda^2+h^2)^{(2-n)/2}, \qquad h\in\mathbb R.(1)✦

An open real interval II has one of the endpoint forms in (2), where the finite endpoints range independently over ℝ\mathbb R: I=(s−,s+),I=(−∞,s+),I=(s−,∞),I=ℝ.I=(s_-,s_+),\qquad I=(-\infty,s_+),\qquad I=(s_-,\infty),\qquad I=\mathbb R.(2)✦

Let k:ℝ→ℝk:\mathbb R\to\mathbb R be a profile. For an open real interval II, the endpoint kernel average in the first variable is the function 𝒜ILk:ℝ→ℝ\mathcal A_I^L k:\mathbb R\to\mathbb R defined by (3): (𝒜ILk)(s′)={k(s−−s′)+k(s+−s′)2,I=(s−,s+),0,I is unbounded.(\mathcal A_I^L k)(s')= \begin{cases} \dfrac{k(s_--s')+k(s_+-s')}{2}, & I=(s_-,s_+),\\[6pt] 0, & I\text{ is unbounded}. \end{cases}(3)✦

For an open real interval JJ, the endpoint kernel average in the second variable is the function 𝒜JRk:ℝ→ℝ\mathcal A_J^R k:\mathbb R\to\mathbb R defined by (4): (𝒜JRk)(s)={k(s−t−)+k(s−t+)2,J=(t−,t+),0,J is unbounded.(\mathcal A_J^R k)(s)= \begin{cases} \dfrac{k(s-t_-)+k(s-t_+)}{2}, & J=(t_-,t_+),\\[6pt] 0, & J\text{ is unbounded}. \end{cases}(4)✦

In the paper, the result reads: for a unit vector e∈ℝne\in\mathbb R^n and distinct z,z′∈e⟂z,z'\in e^\perp, write x=z+sex=z+se, y=z′+s′ey=z'+s'e, λ=|z−z′|>0\lambda=|z-z'|>0 and h=s−s′h=s-s'; then 𝒦(s,s′)=(λ2+h2)(2−n)/2/((n−2)|Sn−1|)\mathcal K(s,s')=(\lambda^2+h^2)^{(2-n)/2}/ ((n-2)|S^{n-1}|) and 𝒟=h2/(λ2+h2)\mathcal D=h^2/(\lambda^2+h^2), where |Sn−1||S^{n-1}| is the unnormalized surface area of the unit sphere, while for bounded I=(s−,s+)I=(s_-,s_+) and J=(t−,t+)J=(t_-,t_+) the endpoint averages are 𝒦¯I(s′)=(𝒦(s−,s′)+𝒦(s+,s′))/2\overline{\mathcal K}_I(s')=(\mathcal K(s_-,s')+\mathcal K(s_+,s'))/2 and 𝒦¯J(s)=(𝒦(s,t−)+𝒦(s,t+))/2\overline{\mathcal K}_J(s)=(\mathcal K(s,t_-)+\mathcal K(s,t_+))/2.

What Is Checked

The kernel is defined for n∈ℕn\in\mathbb N and λ,c∈ℝ\lambda,c\in\mathbb R; the endpoint averages are defined for every k:ℝ→ℝk:\mathbb R\to\mathbb R and every pair of open intervals.