OctoInformBeta

OpenAI preprint, September 2026 · §6. Infinite paths and weak-Cauchy block sequences

Proposition 6.3: Simultaneous evaluation by the infinite paths of a fixed pure or mixed family

Bounded block sequences admit pathwise Cauchy subsequences.

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 block sequence in the real norming completion EE over ω1\omega_1 can be tested by the bounded linear functionals associated with infinite paths. The selection problem asks for one subsequence that works for all paths in a fixed operation family.

The coordinate set is Γ=ω1\Gamma=\omega_1, the ordinals below the initial ordinal of cardinality ℵ1\aleph_1. Fix the countable unbounded partition given by a bijection π:Γ→ℕ×ω1\pi:\Gamma\longrightarrow\mathbb N\times\omega_1. The color c(a)c(a) is the first component of π(a)\pi(a), and Ck={a:c(a)=k}C_k=\{a:c(a)=k\} is its kkth color class.✦ A rational array is a rational-valued function on Γ\Gamma with finite support; write 𝒜=c00(Γ,ℚ)\mathcal A=c_{00}(\Gamma,\mathbb Q).✦

An operation family ff is either the pure family of a color kk, with parameter sf=k+2s_f=k+2, or the mixed family, with parameter sf=2s_f=2. Two arrays u,v∈𝒜u,v\in\mathcal A have successive supports for ff when every a∈supp⁡ua\in\operatorname{supp}u and b∈supp⁡vb\in\operatorname{supp}v satisfy a<ba<b, and also c(a)<c(b)c(a)<c(b) when ff is mixed.✦

The Norming Operations

For a natural parameter ss, the depth dnd_n, recursive state (ts,n,bs,n)(t_{s,n},b_{s,n}), block length Ls,nL_{s,n} and normalization factor ms,nm_{s,n} are the sequences in (1); subtraction in their exponents is natural-number subtraction. dn=n+11,(ts,0,bs,0)=(d02+2d0+2,0),bs,n+1=bs,n+2⋅2s(s+1)ts,n,ts,n+1=ts,n+(bs,n+1+1)(dn+1+1)+1+dn+1+dn+12,Ls,n=2s(s+1)ts,n,ms,n=2(s−1)(s+1)ts,n+dn.\begin{aligned} d_n&=n+11,& (t_{s,0},b_{s,0})&=(d_0^2+2d_0+2,0),\\ b_{s,n+1}&=b_{s,n}+2\cdot2^{s(s+1)t_{s,n}},\\ t_{s,n+1}&=t_{s,n}+(b_{s,n+1}+1)(d_{n+1}+1) +1+d_{n+1}+d_{n+1}^2,\\ L_{s,n}&=2^{s(s+1)t_{s,n}},& m_{s,n}&=2^{(s-1)(s+1)t_{s,n}+d_n}. \end{aligned}(1) Sequence indices begin at zero. For a positive weight jj, the family parameters are Ljf=Lsf,j−1L^f_j=L_{s_f,j-1} and mjf=msf,j−1m^f_j=m_{s_f,j-1}. The exponent scale ts,nt_{s,n} is the first component of the recursive parameter state for parameter ss at index nn.

A Type I expression ee in ff consists of a weight j≥1j\geq1 and nonzero children u0,…,uℓ−1∈𝒜u_0,\ldots,u_{\ell-1}\in\mathcal A, where 1≤ℓ≤Ljf1\leq\ell\leq L^f_j and the children have successive supports for ff.✦ Its normalized value is given by (2). val⁡(e)=1mjf∑h=0ℓ−1uh.\operatorname{val}(e)=\frac1{m^f_j}\sum_{h=0}^{\ell-1}u_h.(2)✦

Raw Paths and Finite Paths

The coherent finite closure system assigns increasing finite sets Hn(β)H_n(\beta) to β<ω1\beta<\omega_1, each containing β\beta and contained in [0,β][0,\beta], with union [0,β][0,\beta]. Its coherence condition is Hn(α)=Hn(β)∩[0,α]H_n(\alpha)=H_n(\beta)\cap[0,\alpha] whenever α∈Hn(β)\alpha\in H_n(\beta). A predecessor closure family at β\beta is a family of finite subsets of Γ\Gamma indexed by endpoints α<β\alpha<\beta and zero-based indices nn.✦ It is coherent below β\beta when its sets contain their endpoints, are contained in the corresponding initial segments, increase with nn, exhaust those initial segments, and satisfy the coherence condition for finite closures. A coherent endpoint extension of a predecessor closure family at β\beta is an increasing, exhausting finite-set sequence at β\beta, containing β\beta and lying below it, whose intersection with [0,α][0,\alpha] equals that family’s set at endpoint α\alpha and the same index whenever it contains α<β\alpha<\beta.✦ The endpoint closure construction assigns a finite-set sequence to a predecessor family, and the system HH is defined recursively in the endpoint using this construction.✦

The positive-indexed finite closures and their entry ranks are defined in (3). Fp(β)=Hmax⁡{p−1,0}(β),ρ(α,β)={min⁡{p≥1:α∈Fp(β)},α≤β,0,α>β.F_p(\beta)=H_{\max\{p-1,0\}}(\beta),\qquad \rho(\alpha,\beta)= \begin{cases} \min\{p\geq1:\alpha\in F_p(\beta)\},&\alpha\leq\beta,\\ 0,&\alpha>\beta. \end{cases}(3)

A raw infinite path rr consists of nonzero original rational pieces un∈𝒜u_n\in\mathcal A, declared weights jn∈ℕj_n\in\mathbb N and metadata pn∈ℕp_n\in\mathbb N, all indexed from zero.✦ Its prefix support, endpoint and closure are the finite sets and ordinal in (4). Sn(r)=⋃i=0nsupp⁡ui,βn(r)=max⁡Sn(r),Dn(r)=Fpn(βn(r)).S_n(r)=\bigcup_{i=0}^n\operatorname{supp}u_i,\qquad \beta_n(r)=\max S_n(r),\qquad D_n(r)=F_{p_n}(\beta_n(r)).(4) The prefix support Sn(r)S_n(r) is nonempty.✦

For a finite coordinate set DD and a∈Γa\in\Gamma, the order position of aa in DD is pos⁡D(a)=|{b∈D:b<a}|\operatorname{pos}_D(a)=|\{b\in D:b<a\}|.✦ The positional rational array code of u∈𝒜u\in\mathcal A on DD is AD(u)={(pos⁡D(a),u(a)):a∈D}A_D(u)=\{(\operatorname{pos}_D(a),u(a)):a\in D\}; it retains zero coefficients.✦ A finite path code is a finite list of triples consisting of a natural weight, natural metadata and a finite set of position-coefficient pairs.✦ The prefix code of rr is the list in (5), with each piece coded on the common closure Dn(r)D_n(r). cn(r)=[(ji,pi,ADn(r)(ui))]i=0n.c_n(r)=\bigl[(j_i,p_i,A_{D_n(r)}(u_i))\bigr]_{i=0}^n.(5)✦ The maximum metadata of a code is the largest metadata entry, with value zero for the empty list.✦ The injective coding map σ\sigma pairs a code’s canonical natural-number encoding with one plus its maximum metadata; its value exceeds every metadata entry in the code.✦

A raw path rr is admissible when its pieces have successive ordinal supports, its metadata are strictly increasing, and the conditions in (6) hold. 1≤jn≤pn,1≤pn,jn+1=σ(cn(r)),ρ(a,b)≤pnfor a,b∈Sn(r),a≤b.\begin{gathered} 1\leq j_n\leq p_n,\qquad 1\leq p_n,\qquad j_{n+1}=\sigma(c_n(r)),\\ \rho(a,b)\leq p_n \quad\text{for }a,b\in S_n(r),\ a\leq b. \end{gathered}(6)✦ A valid prefix of length ℓ\ell of raw data rr is its initial segment at indices below ℓ\ell, whose pieces have successive ordinal supports, whose metadata are strictly increasing, and which satisfies (6) at those indices, with the next-weight rule required only when n+1<ℓn+1<\ell.

A finite path PP in ff consists of raw infinite data rr and a positive length ℓ\ell with valid prefix, together with Type I expressions eie_i for i<ℓi<\ell whose values and weights are uiu_i and jij_i. The original pieces have successive supports for ff. The finite path sum of a finite path PP of length ℓ\ell is val⁡(P)=∑i<ℓui\operatorname{val}(P)=\sum_{i<\ell}u_i, where the uiu_i are its original uncropped pieces.✦

The Completion and Its Blocks

For a base B⊆𝒜B\subseteq\mathcal A and family ff, the norming stages Sn(B,f)S_n(B,f) start with S0(B,f)=BS_0(B,f)=B. The next stage retains the preceding stage and adjoins Type I values whose children lie there, and norming values obtained from original paths, called Type II values, for which every child of every original uncropped piece on every original path lies there. Write δa\delta_a for the rational array equal to one at aa and zero elsewhere. The signed monochromatic base, pure norming sets and full norming set are those in (7). Bk={0}∪{δa,−δa:a∈Ck},Kk=⋃n≥0Sn(Bk,purek),K=⋃n≥0Sn(⋃k≥0Kk,mixed).\begin{aligned} B_k&=\{0\}\cup\{\delta_a,-\delta_a:a\in C_k\},\\ K_k&=\bigcup_{n\geq0}S_n(B_k,\mathrm{pure}\ k),\\ K&=\bigcup_{n\geq0}S_n\left(\bigcup_{k\geq0}K_k,\mathrm{mixed}\right). \end{aligned}(7) The base BfB_f of a pure family of color kk is BkB_k, and the mixed base is ⋃kKk\bigcup_kK_k.✦ The norming family of ff is Kf=⋃n≥0Sn(Bf,f)K_f=\bigcup_{n\geq0}S_n(B_f,f).✦

A norming set JJ on Γ\Gamma is a subset of 𝒜\mathcal A containing every δa\delta_a and satisfying |u(a)|≤1|u(a)|\leq1 for u∈Ju\in J and a∈Γa\in\Gamma.✦ The bounded function space ℓ∞(J)\ell_\infty(J) of a norming set JJ on Γ\Gamma is the real space of bounded functions on JJ with the supremum norm.✦ For a finite vector x∈c00(Γ,ℝ)x\in c_{00}(\Gamma,\mathbb R), the coordinate evaluation vectors vav_a, finite support inclusion TJT_J and closed evaluation span EJE_J are defined in (8). va(u)=u(a),TJx=∑a∈supp⁡xx(a)va,EJ=TJ(c00(Γ,ℝ))¯∥⋅∥∞⊆ℓ∞(J).\begin{aligned} v_a(u)&=u(a),& T_Jx&=\sum_{a\in\operatorname{supp}x}x(a)v_a,\\ E_J&=\overline{T_J(c_{00}(\Gamma,\mathbb R))}^{\,\|\cdot\|_\infty} \subseteq\ell_\infty(J). \end{aligned}(8) The norming completion EJE_J associated with a norming set JJ on Γ\Gamma is the closed evaluation span in (8), with its inherited norm; the inclusion of finite vectors into EJE_J is TJT_J with this restricted target.✦ The norming evaluation functional uˆ\widehat u associated with u∈Ju\in J is the real linear functional on EJE_J defined by uˆ(z)=z(u)\widehat u(z)=z(u) for z∈EJz\in E_J.✦ The full norming structure uses J=KJ=K✦, and E=EKE=E_K is the constructed real Banach space.✦

An admissible block sequence (ξn)(\xi_n) in ff is a sequence of nonzero vectors in c00(Γ,ℝ)c_{00}(\Gamma,\mathbb R) with supports successive for ff. For a pure family of color kk, all their support coordinates must lie in CkC_k. The embedded block vectors of an admissible block sequence (ξn)(\xi_n) in ff are xn=TKξn∈Ex_n=T_K\xi_n\in E.✦

Infinite Paths and Their Functionals

An infinite path PP in ff consists of admissible raw data rr and Type I expressions ene_n whose values and weights agree with unu_n and jnj_n. The original pieces have successive supports for ff, and every child of every ene_n belongs to KfK_f. The finite prefix P|nP|_n of an infinite path PP in ff is the finite path that retains its raw data and the expressions at indices zero through nn; its length is n+1n+1.✦

The path partial sums are the bounded norming evaluation functionals FP,n=val⁡(P|n)ˆF_{P,n}=\widehat{\operatorname{val}(P|_n)} on EE.✦ The path functional FPF_P is a chosen continuous real linear functional with the properties in (9)✦; such a functional exists for every infinite path. ∥FP∥≤1,limn→∞FP,n(z)=FP(z)(z∈E).\|F_P\|\leq1,\qquad \lim_{n\to\infty}F_{P,n}(z)=F_P(z)\quad(z\in E).(9)✦

The Claim

Theorem 1. Let ff be a pure or mixed operation family, and let (xn)(x_n) be the embedded vectors of an admissible block sequence in ff. Suppose that a real number MM satisfies ∥xn∥≤M\|x_n\|\leq M for every nn. There is a strictly increasing map φ:ℕ→ℕ\varphi:\mathbb N\longrightarrow\mathbb N such that (10) holds for every infinite path PP in ff: (FP(xφ(n)))n≥0is a Cauchy sequence in ℝ.\bigl(F_P(x_{\varphi(n)})\bigr)_{n\geq0} \quad\text{is a Cauchy sequence in }\mathbb R.(10)✦

In the paper, the result reads: every bounded block sequence in a fixed family has a weak-Cauchy subsequence, meaning convergence under every member of E*E^*.

What Is Checked

The claim is stated over the full real norming completion, its embedded admissible block sequences, and the infinite paths whose Type I children belong to the fixed family’s norming set. The full set KK contains every positive coordinate array δa\delta_a✦, and every coefficient of every member of KK has absolute value at most one.✦ For each infinite path PP, every finite prefix sum val⁡(P|n)\operatorname{val}(P|_n) belongs to KK.✦