OctoInformBeta

OpenAI preprint, September 2026 · §4. Bases within a density-zero class

Proposition 4.6: For I⊆YI \subseteq Y and TT, E0\TE_0 \setminus T positive, ℬσ(T)\mathcal{B}^\sigma(T) meets [I,Y][I, Y] when T\IT \setminus I or Y\TY \setminus T has density zero

When a local rule meets an interval of sets.

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.

Proposition 4.6 concerns two families of subsets of E0E_0 attached to a subset TT of E0E_0, the local lower rule and the local upper rule, whose members lie in the class of TT, and asks where their members fall relative to an interval [I,Y][I, Y] of subsets of E0E_0, according to how II and YY compare with TT modulo 𝒥0\mathcal{J}_0.

The Setting

The ground set is E0=L⊔R=D×{false,true}E_0 = L \sqcup R = D \times \{\mathrm{false}, \mathrm{true}\}.✦ Here DD is the countable set of pairs (m,f)(m, f) with m≥0m \ge 0 and f:{0,1}m→{0,1}f : \{0,1\}^m \to \{0,1\}, and the left copy LL and the right copy RR of DD in E0E_0 are the pairs with second entry false and true respectively. The block Dm={m}×{0,1}{0,1}mD_m = \{m\} \times \{0,1\}^{\{0,1\}^m} of DD has wm=22mw_m = 2^{2^m} elements✦, and the doubled block Wm=Dm×{L,R}W_m = D_m \times \{L, R\} is the finite subset of E0E_0 made of both copies of DmD_m✦.

For a subset XX of E0E_0 and m≥0m \ge 0, the density of XX in WmW_m is |X∩Wm|/|Wm||X \cap W_m| / |W_m|.✦ The ideal 𝒥0\mathcal{J}_0 is the family of subsets of E0E_0 whose density in WmW_m tends to 00: 𝒥0={X⊆E0:limm→∞|X∩Wm||Wm|=0}.\mathcal{J}_0 = \Bigl\{ X \subseteq E_0 : \lim_{m \to \infty} \frac{|X \cap W_m|}{|W_m|} = 0 \Bigr\}.(1)✦ A subset XX of E0E_0 is positive when XX is not in 𝒥0\mathcal{J}_0. A subset XX of E0E_0 lies in the class of a subset TT of E0E_0 when the symmetric difference X△TX \mathbin{\triangle} T is in 𝒥0\mathcal{J}_0.✦

For subsets I⊆YI \subseteq Y of E0E_0, the interval [I,Y][I, Y] is the family of subsets BB of E0E_0 with I⊆B⊆YI \subseteq B \subseteq Y. A subset BB of E0E_0 is an outcome for [I,Y][I, Y] when B⊆IorI⊆B⊆YorY⊆B.B \subseteq I \quad\text{or}\quad I \subseteq B \subseteq Y \quad\text{or}\quad Y \subseteq B.(2)✦ A family ℬ\mathcal{B} of subsets of E0E_0 meets the interval [I,Y][I, Y] when some member of ℬ\mathcal{B} is an outcome for [I,Y][I, Y].

The Probes and the Local Rules

The probes compare a subset XX of E0E_0 with a fixed subset TT of E0E_0. For n≥0n \ge 0 the prefix Vn=W0∪…∪Wn−1V_n = W_0 \cup \dots \cup W_{n-1} has cn=|Vn|c_n = |V_n| elements.✦ The density difference dm(X)d_m(X) is the density of XX in WmW_m minus that of TT✦, and the prefix count difference is pn(X)=|X∩Vn|−|T∩Vn|p_n(X) = |X \cap V_n| - |T \cap V_n|✦. The probe fn(X)f_n(X) and its lower limit ℓT(X)\ell_T(X), taken in the extended reals, are fn(X)=pn(X)+cn(supm≥ndm(X)+infm≥ndm(X)),ℓT(X)=liminfn→∞fn(X).f_n(X) = p_n(X) + c_n \Bigl( \sup_{m \ge n} d_m(X) + \inf_{m \ge n} d_m(X) \Bigr), \qquad \ell_T(X) = \liminf_{n \to \infty} f_n(X).(3)

The local lower rule ℬ−(T)\mathcal{B}^-(T) at TT is the family of subsets of E0E_0 in the class of TT whose lower limit lies in (−1,0](-1, 0]: ℬ−(T)={X⊆E0:X△T∈𝒥0 and −1<ℓT(X)≤0}.\mathcal{B}^-(T) = \{ X \subseteq E_0 : X \mathbin{\triangle} T \in \mathcal{J}_0 \text{ and } -1 < \ell_T(X) \le 0 \}.(4)✦ The local upper rule ℬ+(T)\mathcal{B}^+(T) at TT is the family of subsets of E0E_0 whose complement lies in the local lower rule at E0\TE_0 \setminus T: ℬ+(T)={X⊆E0:E0\X∈ℬ−(E0\T)}.\mathcal{B}^+(T) = \{ X \subseteq E_0 : E_0 \setminus X \in \mathcal{B}^-(E_0 \setminus T) \}.(5)✦ For a Boolean σ\sigma, the local rule ℬσ(T)\mathcal{B}^\sigma(T) is ℬ+(T)\mathcal{B}^+(T) when σ\sigma is true and ℬ−(T)\mathcal{B}^-(T) when σ\sigma is false.✦

The Claim

Theorem 1. Let σ\sigma be a Boolean, let TT be a subset of E0E_0 such that TT and E0\TE_0 \setminus T are positive, and let I⊆YI \subseteq Y be subsets of E0E_0. If one of the conditions T\I∈𝒥0,T \setminus I \in \mathcal{J}_0,(6)✦ Y\T∈𝒥0,Y \setminus T \in \mathcal{J}_0,(7)✦ I△T∈𝒥0andY△T∈𝒥0I \mathbin{\triangle} T \in \mathcal{J}_0 \quad\text{and}\quad Y \mathbin{\triangle} T \in \mathcal{J}_0(8)✦ holds, then ℬσ(T)\mathcal{B}^\sigma(T) meets [I,Y][I, Y]. If Y\IY \setminus I is positive and II lies in the class of TT, then some B∈ℬσ(T)B \in \mathcal{B}^\sigma(T) satisfies B⊆IorI⊆B⊆Y,B \subseteq I \quad\text{or}\quad I \subseteq B \subseteq Y,(9)✦ and if Y\IY \setminus I is positive and YY lies in the class of TT, then some B∈ℬσ(T)B \in \mathcal{B}^\sigma(T) satisfies I⊆B⊆YorY⊆B.I \subseteq B \subseteq Y \quad\text{or}\quad Y \subseteq B.(10)✦

In the paper, the result reads: for σ∈{−,+}\sigma \in \{-, +\} and I⊆Y⊆E0I \subseteq Y \subseteq E_0, if T\I∈𝒥0T \setminus I \in \mathcal{J}_0 or Y\T∈𝒥0Y \setminus T \in \mathcal{J}_0, then some B∈ℬσ(T)B \in \mathcal{B}^\sigma(T) satisfies B⊆IB \subseteq I, or I⊆B⊆YI \subseteq B \subseteq Y, or Y⊆BY \subseteq B, with −- and ++ in place of false and true.

What Is Checked

The claim is stated over the ground set E0E_0 with its doubled blocks WmW_m, the ideal 𝒥0\mathcal{J}_0 of (1) and the classes modulo it, the probes fnf_n relative to TT with their lower limit ℓT\ell_T, the local rules of (4) and (5) chosen by σ\sigma, and the outcomes of (2).

The local lower rule has two further properties. Let TT, XX and GG be subsets of E0E_0 with XX in the class of TT and GG positive. If GG is disjoint from XX and ℓT(X)≤−1\ell_T(X) \le -1, then some B∈ℬ−(T)B \in \mathcal{B}^-(T) satisfies X⊆B⊆X∪G,X \subseteq B \subseteq X \cup G,(11)✦ and if G⊆XG \subseteq X and ℓT(X)>0\ell_T(X) > 0, then some B∈ℬ−(T)B \in \mathcal{B}^-(T) satisfies X\G⊆B⊆X.X \setminus G \subseteq B \subseteq X.(12)✦