OctoInformBeta

OpenAI preprint, September 2026 · §3. A joining cost with one entropy budget

§3 (setting): the interior two-point cost: Center and half-difference coordinates for the joining cost and the entropy deficit representation

Entropy and dissipation of a two-point field.

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

A two-point field on an edge of the discrete cube takes the values m+δm+\delta and m−δm-\delta with equal probability, where mm is its center and δ\delta is its half-difference. An interior pair with center mm and half-difference δ\delta is a two-point field whose values m+δm+\delta and m−δm-\delta both lie in (−1,1)(-1,1).

The binary entropy deficit ψ\psi is the function given by (1), with natural logarithms. ψ(v)=1+v2log⁡(1+v)+1−v2log⁡(1−v).\psi(v)=\frac{1+v}{2}\log(1+v)+\frac{1-v}{2}\log(1-v).(1)✦ The binary entropy HH in nats is the function given by (2), where L=log⁡2L=\log 2. H(v)=L−ψ(v).H(v)=L-\psi(v).(2)✦ Write atanh\operatorname{atanh} for the inverse hyperbolic tangent.

The Claim

Definition 1. For real mm and δ\delta, the two-point entropy gap G(m,δ)G(m,\delta) is the entropy difference in (3). G(m,δ)=H(m)−H(m+δ)+H(m−δ)2.G(m,\delta)=H(m)-\frac{H(m+\delta)+H(m-\delta)}{2}.(3)✦ The joining cost D(m,δ)D(m,\delta) of a two-point field with center mm and half-difference δ\delta, also called its two-point dissipation, is the real number in (4). D(m,δ)=δ2(atanh⁡(m+δ)−atanh⁡(m−δ)).D(m,\delta)=\frac{\delta}{2} \bigl(\operatorname{atanh}(m+\delta)-\operatorname{atanh}(m-\delta)\bigr).(4)✦

Theorem 1. For real mm and δ\delta, the two-point entropy gap has the entropy deficit expression (5). G(m,δ)=ψ(m+δ)+ψ(m−δ)2−ψ(m).G(m,\delta)=\frac{\psi(m+\delta)+\psi(m-\delta)}{2}-\psi(m).(5)✦

What Is Checked

The definitions (3) and (4) and the identity (5) are stated over real centers and half-differences, with HH and ψ\psi as in (2) and (1).