OctoInformBeta

OpenAI preprint, September 2026 · §2. A finite connection comparison

Proposition 2.1: Independent hyperedges on a finite vertex set, with real and extended probability bounds

Uniform connection bounds control joint connections.

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 event that a vertex reaches both of two target sets concerns the same open hyperedges for both connections.

A labelled hypergraph HH on a finite vertex set VV, with a set EE of hyperedge labels, is an assignment e↦H(e)⊆Ve\mapsto H(e)\subseteq V; empty hyperedges are permitted.✦ A configuration ω∈Ω:={0,1}E\omega\in\Omega:=\{0,1\}^E declares a label ee open when ω(e)=1\omega(e)=1.

The Bernoulli hyperedge measure μp\mu_p, for probabilities p:E→[0,1]p:E\to[0,1], is the product law in (1), where δi\delta_i is unit mass at ii: μp=⨂e∈E((1−p(e))δ0+p(e)δ1).\mu_p=\bigotimes_{e\in E} \bigl((1-p(e))\delta_0+p(e)\delta_1\bigr).(1)✦ The real event probability ℙp(F)\mathbb P_p(F) of an event F⊆ΩF\subseteq\Omega is the integral in (2); μp(F)\mu_p(F) denotes its extended measure, with values in [0,∞][0,\infty]. ℙp(F)=∫Ω𝟏F(ω)dμp(ω).\mathbb P_p(F)=\int_{\Omega}\mathbf 1_F(\omega)\,d\mu_p(\omega).(2)

For a forbidden vertex set D⊆VD\subseteq V, open hyperedge adjacency ∼ω,D\sim_{\omega,D} on VV is the relation in (3). x∼ω,Dy⇔∃e∈E:ω(e)=1,H(e)∩D=⌀,x,y∈H(e).x\sim_{\omega,D}y \quad\Longleftrightarrow\quad \exists e\in E:\quad \omega(e)=1,\quad H(e)\cap D=\varnothing,\quad x,y\in H(e).(3)✦ An open connection x↔Dyx\leftrightarrow_D y in HH is a finite chain of open hyperedge adjacencies from xx to yy, with a chain of length zero permitted.✦ For a vertex x∈Vx\in V and a target set S⊆VS\subseteq V, the connection event x↔DSx\leftrightarrow_D S is the event defined by (4). x↔DS⇔∃y∈S:x↔Dy.x\leftrightarrow_D S \quad\Longleftrightarrow\quad \exists y\in S:\quad x\leftrightarrow_D y.(4)✦ When D=⌀D=\varnothing, write ↔\leftrightarrow.

The Claim

Theorem 1. Let HH be a labelled hypergraph on a finite vertex set VV, with independent hyperedge probabilities p:E→[0,1]p:E\to[0,1], a vertex o∈Vo\in V and sets A,T⊆VA,T\subseteq V. Suppose b∈ℝb\in\mathbb R is a lower bound for ℙp(a↔T)\mathbb P_p(a\leftrightarrow T) for every a∈Aa\in A. Then bℙp(o↔A)≤ℙp(o↔A,o↔T).b\,\mathbb P_p(o\leftrightarrow A) \leq \mathbb P_p(o\leftrightarrow A,\ o\leftrightarrow T).(5)✦ If AA is nonempty and b∈[0,∞]b\in[0,\infty] is a lower bound for μp(a↔T)\mu_p(a\leftrightarrow T) for every a∈Aa\in A, then μp(o↔A)b≤μp(o↔A,o↔T).\mu_p(o\leftrightarrow A)\,b \leq \mu_p(o\leftrightarrow A,\ o\leftrightarrow T).(6)✦

In the paper, the result reads: in independent percolation on a finite hypergraph with arbitrary hyperedge probabilities in [0,1][0,1], a vertex oo and vertex sets A,TA,T with A≠⌀A\ne\varnothing satisfy ℙp(o↔A,o↔T)≥ℙp(o↔A)min⁡a∈Aℙp(a↔T)\mathbb P_p(o\leftrightarrow A,\ o\leftrightarrow T) \geq\mathbb P_p(o\leftrightarrow A) \min_{a\in A}\mathbb P_p(a\leftrightarrow T).

What Is Checked

The bounds (5) and (6) are stated over labelled hypergraphs with finitely many vertices under the Bernoulli product law, using connections with no forbidden vertices and real event probabilities or extended measures, respectively.