OctoInformBeta

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

The class š’ž(T), the probes and the rules ℬ^- and ℬ^+ (4.1), (4.2): For TāŠ†E0T \subseteq E_0, the probes fnf_n comparing a set with TT, and the rules ā„¬āˆ’(T)\mathcal{B}^-(T), ℬ+(T)\mathcal{B}^+(T) given by their limits

Probes and two rules within a density-zero class.

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

Fix a subset TT of the doubled ground set E0E_0, the union of two copies of a countable set DD, and take the sets that differ from TT by a set of density zero. Probes compare such a set with TT block by block, and two rules, ā„¬āˆ’(T)\mathcal{B}^-(T) and ℬ+(T)\mathcal{B}^+(T), select among these sets by the limits of the probes.

The doubled ground set E0=LāŠ”RE_0 = L \sqcup R is DƗ{false,true}D \times \{\mathrm{false}, \mathrm{true}\}, where DD is the countable set of pairs (m,f)(m, f) with f:{0,1}m→{0,1}f\colon \{0,1\}^m \to \{0,1\}; the pairs with second entry false form the left copy LL of DD, those with second entry true the right copy RR. For a natural number mm, the block Dm={m}Ɨ{0,1}{0,1}mD_m = \{m\} \times \{0,1\}^{\{0,1\}^m} is the finite subset of DD of the pairs with first entry mm, and it has wm=22mw_m = 2^{2^m} elements.✦ The doubled block Wm=DmƗ{L,R}W_m = D_m \times \{L, R\} is the finite subset of E0E_0 formed by the two copies of DmD_m.✦

The density of a subset XX of a set AA in a finite subset ss of AA is the real number |X∩s|/|s||X \cap s| / |s|.✦ A subset XX of E0E_0 has density zero when its density in the doubled block WmW_m tends to 00, and the subsets of density zero form the ideal š’„0\mathcal{J}_0: Xāˆˆš’„0⇔limmā†’āˆž|X∩Wm||Wm|=0.X \in \mathcal{J}_0 \iff \lim_{m \to \infty} \frac{|X \cap W_m|}{|W_m|} = 0.(1)✦

The Class and the Probes

The class š’ž(T)\mathcal{C}(T) of a subset TT of E0E_0 is the family of the subsets of E0E_0 whose symmetric difference with TT has density zero: š’ž(T)={XāŠ†E0:Xā–³Tāˆˆš’„0}.\mathcal{C}(T) = \{X \subseteq E_0 : X \mathbin{\triangle} T \in \mathcal{J}_0\}.(2)✦

The probes are relative to TT and are defined for every subset XX of E0E_0 and every n≄0n \ge 0. They are built on the prefix VnV_n, the finite subset of E0E_0 formed by the first nn doubled blocks, and on its number of elements cnc_n: Vn=W0∪W1āˆŖā€¦āˆŖWnāˆ’1,cn=|Vn|.V_n = W_0 \cup W_1 \cup \dots \cup W_{n-1}, \qquad c_n = |V_n|.(3)✦

Given a real sequence (um)(u_m) and n≄0n \ge 0, the tail supremum of (um)(u_m) at nn is the supremum of its terms from index nn on, and the tail infimum at nn is their infimum: supm≄num=sup⁡{un,un+1,un+2,…},\sup_{m \ge n} u_m = \sup\,\{u_n, u_{n+1}, u_{n+2}, \dots\},(4)✦ infm≄num=inf⁡{un,un+1,un+2,…}.\inf_{m \ge n} u_m = \inf\,\{u_n, u_{n+1}, u_{n+2}, \dots\}.(5)✦

At a doubled block WmW_m, the density difference dm(X)d_m(X) of a subset XX of E0E_0 is the density of XX in WmW_m minus the density of TT in WmW_m: dm(X)=|X∩Wm||Wm|āˆ’|T∩Wm||Wm|.d_m(X) = \frac{|X \cap W_m|}{|W_m|} - \frac{|T \cap W_m|}{|W_m|}.(6)✦ On the prefix, the count difference pn(X)p_n(X) is the number of elements of XX in VnV_n minus the number of elements of TT in VnV_n: pn(X)=|X∩Vn|āˆ’|T∩Vn|.p_n(X) = |X \cap V_n| - |T \cap V_n|.(7)✦ The probe fn(X)f_n(X) adds to pn(X)p_n(X) the size cnc_n of the prefix times the sum of the tail supremum and the tail infimum of the density differences at nn: fn(X)=pn(X)+cn(supm≄ndm(X)+infm≄ndm(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).(8)✦

For a sequence (gn)(g_n) of real functions on the subsets of a set AA and a subset XX of AA, the lower limit of gn(X)g_n(X) is the limit inferior of the real numbers gn(X)g_n(X), an element of the extended real line ā„ĀÆ=[āˆ’āˆž,+āˆž]\overline{\mathbb{R}} = [-\infty, +\infty]; the lower limit of the probes fnf_n at XX is written ā„“T(X)\ell_T(X): liminfnā†’āˆžgn(X)āˆˆā„ĀÆ,ā„“T(X)=liminfnā†’āˆžfn(X).\liminf_{n \to \infty} g_n(X) \in \overline{\mathbb{R}}, \qquad \ell_T(X) = \liminf_{n \to \infty} f_n(X).(9)✦

The Claim

Definition 1. A subset XX of a set AA has lower limit in (āˆ’1,0](-1, 0] for a sequence (gn)(g_n) of real functions on the subsets of AA when āˆ’1<liminfnā†’āˆžgn(X)≤0.-1 < \liminf_{n \to \infty} g_n(X) \le 0.(10)✦ For a subset TT of E0E_0, the lower rule ā„¬āˆ’(T)\mathcal{B}^-(T) is the family of the sets in š’ž(T)\mathcal{C}(T) that have lower limit in (āˆ’1,0](-1, 0] for the probes fnf_n relative to TT: ā„¬āˆ’(T)={Xāˆˆš’ž(T):āˆ’1<ā„“T(X)≤0}.\mathcal{B}^-(T) = \{X \in \mathcal{C}(T) : -1 < \ell_T(X) \le 0\}.(11)✦ The upper rule ℬ+(T)\mathcal{B}^+(T) is the family of the subsets of E0E_0 whose complement lies in the lower rule of the complement Tc=E0\TT^c = E_0 \setminus T: ℬ+(T)={XāŠ†E0:E0\Xāˆˆā„¬āˆ’(Tc)}.\mathcal{B}^+(T) = \{X \subseteq E_0 : E_0 \setminus X \in \mathcal{B}^-(T^c)\}.(12)✦ For a Boolean σ\sigma, the rule ā„¬Ļƒ(T)\mathcal{B}^\sigma(T) is the upper rule when σ\sigma is true and the lower rule when σ\sigma is false: ā„¬Ļƒ(T)={ℬ+(T)if σ=true,ā„¬āˆ’(T)if σ=false.\mathcal{B}^\sigma(T) = \begin{cases} \mathcal{B}^+(T) & \text{if } \sigma = \mathrm{true},\\ \mathcal{B}^-(T) & \text{if } \sigma = \mathrm{false}. \end{cases}(13)✦

In the paper, the result reads: š’ž(T)={X:Xā–³Tāˆˆš’„0}\mathcal{C}(T) = \{X : X \mathbin{\triangle} T \in \mathcal{J}_0\}; Vn=ā‹ƒm<nWmV_n = \bigcup_{m<n} W_m, cn=|Vn|c_n = |V_n|; pnp_n, dmd_m and fnf_n as in its (4.1); ā„“T=liminf⁡fn\ell_T = \liminf f_n, hT=limsup⁡fnh_T = \limsup f_n; ā„¬āˆ’(T)={Xāˆˆš’ž(T):āˆ’1<ā„“T(X)≤0}\mathcal{B}^-(T) = \{X \in \mathcal{C}(T) : -1 < \ell_T(X) \le 0\}, ℬ+(T)={Xāˆˆš’ž(T):0≤hT(X)<1}\mathcal{B}^+(T) = \{X \in \mathcal{C}(T) : 0 \le h_T(X) < 1\}.

What Is Checked

The definitions are stated over subsets TT and XX of the doubled ground set E0E_0, the doubled blocks WmW_m formed from the blocks DmD_m of DD, the density of a set in a finite set, and the ideal š’„0\mathcal{J}_0 of the subsets of E0E_0 of density zero.

The library proves that membership in the upper rule (12) is read off the upper limit hT(X)=limsup⁡nā†’āˆžfn(X)h_T(X) = \limsup_{n \to \infty} f_n(X), taken in ā„ĀÆ\overline{\mathbb{R}}, of the probes relative to TT: for every subset XX of E0E_0, Xāˆˆā„¬+(T)⇔Xāˆˆš’ž(T) and 0≤hT(X)<1.X \in \mathcal{B}^+(T) \iff X \in \mathcal{C}(T) \text{ and } 0 \le h_T(X) < 1.(14)✦