OctoInformBeta

OpenAI preprint, September 2026 · Section 2. An extremal weight and its cross-mass bound

The growth function: A scalar function of a real argument, with an arbitrary real linear coefficient

A logarithmic correction to linear growth.

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 growth function combines linear dependence on a mass variable with a logarithmic correction truncated below at zero.

The Definition

Definition 1. Let xx be a real parameter. The growth function gx:ℝ→ℝg_x\colon\mathbb{R}\to\mathbb{R} is the function defined by gx(s)={s(x+max{0,log1|s|}),s≠0,0,s=0.g_x(s)= \begin{cases} s\left(x+\max\left\{0,\log\dfrac{1}{|s|}\right\}\right), & s\ne0,\\ 0, & s=0. \end{cases}(1)✦

In the paper, the definition reads: for x≥1x\ge1, gx(s)=s(x+log⁡+(1/s))g_x(s)=s(x+\log^+(1/s)) when s>0s>0 and gx(0)=0g_x(0)=0, where log⁡+a=max⁡{0,log⁡a}\log^+a=\max\{0,\log a\} for a>0a>0.

What Is Checked

The definition (1) is stated for arbitrary real parameters xx and arguments ss.