OctoInformBeta

OpenAI preprint, September 2026 · §3. Ladder cuts and a coefficient lower bound · §3

Lemma 3.4 [Positive probability of cuts]: For p∈(1/2,1)p \in (1/2, 1), shortened left pieces have integrable length and the right remainder has positive no-survivor probability

Positive probability and density of inventory piece cuts.

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

In the burger-order inventory representation of Fortuin-Kasteleyn planar maps with Potts parameter q>4q > 4, infinite streams of burger and order letters are factored into independent word blocks called left pieces and right pieces. A boundary between consecutive pieces is a left cut when forward prefixes leave no unmatched orders, and a right cut when backward prefixes leave no unmatched burgers; the questions are whether shortened left pieces have finite mean length and whether left and right cuts occur with positive probability and positive density.

Alphabet, Reduction, and Stream Laws

An inventory order o∈𝒪o \in \mathcal{O} is either a fixed order of boolean type b∈{false,true}b \in \{\mathrm{false}, \mathrm{true}\} or the flexible order *\ast.✦ The burger-order alphabet 𝒜\mathcal{A} is the finite set consisting of the burger letters BbB_b for b∈{false,true}b \in \{\mathrm{false}, \mathrm{true}\} and the order letters OoO_o for o∈𝒪o \in \mathcal{O}.✦ A word w∈𝒜*w \in \mathcal{A}^* is a finite sequence of letters from 𝒜\mathcal{A}, with length |w|∈ℕ={0,1,2,…}|w| \in \mathbb{N} = \{0, 1, 2, \dots\} and reversal wrevw^{\mathrm{rev}}.

An inventory stack evaluation state e=(eord,eburg)e = (e_{\mathrm{ord}}, e_{\mathrm{burg}}) consists of a list eorde_{\mathrm{ord}} of pending orders in 𝒪\mathcal{O} and a freshest-first list eburge_{\mathrm{burg}} of boolean burger types in {false,true}\{\mathrm{false}, \mathrm{true}\}.✦ The word evaluation run(w,e)\mathrm{run}(w, e) of a word w∈𝒜*w \in \mathcal{A}^* at an evaluation state ee is the evaluation state that equals ee on the empty word [][], and on a letter aa followed by a suffix uu recursively evaluates uu after updating ee by the single-step transition on aa, in which each order consumes the freshest eligible burger.✦ The word reduction red(w)=(redord(w),redburg(w))\mathrm{red}(w) = (\mathrm{red}_{\mathrm{ord}}(w), \mathrm{red}_{\mathrm{burg}}(w)) of w∈𝒜*w \in \mathcal{A}^* is the evaluation state run(w,([],[]))\mathrm{run}(w, ([], [])) starting from empty pending order and burger stacks.✦

A fixed word w∈𝒜*w \in \mathcal{A}^* is a word in 𝒜*\mathcal{A}^* in which the flexible order O*O_\ast does not appear.✦ For b∈{false,true}b \in \{\mathrm{false}, \mathrm{true}\}, the letter net type count δb:𝒜→ℤ\delta_b : \mathcal{A} \to \mathbb{Z} is the function given by (1), δb(a)={1,a=Bb,−1,a=Ob,0,otherwise,\delta_b(a) = \begin{cases} 1, & a = B_b, \\ -1, & a = O_b, \\ 0, & \text{otherwise}, \end{cases}(1)✦ and the word net type count of w∈𝒜*w \in \mathcal{A}^* is the integer Δb(w)=∑i=0|w|−1δb(wi)\Delta_b(w) = \sum_{i=0}^{|w|-1} \delta_b(w_i) summed over the letters w0,…,w|w|−1w_0, \dots, w_{|w|-1} of ww.✦

The letter increment ξ:𝒜→ℤ\xi : \mathcal{A} \to \mathbb{Z} is the function assigning +1+1 to every burger letter BbB_b and −1-1 to every order letter OoO_o.✦ For a sequence x=(xi)i≥0∈𝒜ℕx = (x_i)_{i \ge 0} \in \mathcal{A}^{\mathbb{N}} and n∈ℕn \in \mathbb{N}, the letter increment sum Sn(x)∈ℤS_n(x) \in \mathbb{Z} of xx up to nn is the sum ∑i=0n−1ξ(xi)\sum_{i=0}^{n-1} \xi(x_i).✦

The model parameter is a real number p∈(1/2,1)p \in (1/2, 1).✦ For p∈(1/2,1)p \in (1/2, 1), the letter mass function mp:𝒜→ℝm_p : \mathcal{A} \to \mathbb{R} is the function given by (2), mp(a)={1/4,a=Bb for b∈{false,true},p/2,a=O*,(1−p)/4,a=Ob for b∈{false,true},m_p(a) = \begin{cases} 1/4, & a = B_b \text{ for } b \in \{\mathrm{false}, \mathrm{true}\}, \\ p/2, & a = O_\ast, \\ (1 - p)/4, & a = O_b \text{ for } b \in \{\mathrm{false}, \mathrm{true}\}, \end{cases}(2)✦ and the letter probability distribution πp\pi_p on 𝒜\mathcal{A} is the probability distribution obtained by normalizing mpm_p.✦

For any index set ι\iota and p∈(1/2,1)p \in (1/2, 1), the independent and identically distributed (i.i.d.) inventory stream measure ℙpι\mathbb{P}_p^{\iota} on 𝒜ι\mathcal{A}^{\iota} is the infinite product measure with marginal πp\pi_p at every coordinate.✦ Given a probability measure ν\nu on a measurable space α\alpha, the two-sided i.i.d. sequence law ν⊗ℤ\nu^{\otimes \mathbb{Z}} on αℤ\alpha^{\mathbb{Z}} is the infinite product measure with marginal ν\nu at every integer coordinate.✦ The natural pieces law ν⊗ℕ\nu^{\otimes \mathbb{N}} of a probability measure ν\nu on 𝒜*\mathcal{A}^* is the infinite product measure on (𝒜*)ℕ(\mathcal{A}^*)^{\mathbb{N}} with marginal ν\nu at every index n∈ℕn \in \mathbb{N}.✦

Stopping Times and Piece Constructions

Given a space Ω\Omega and an integer-valued process F:ℕ×Ω→ℤF : \mathbb{N} \times \Omega \to \mathbb{Z}, the first hitting time of zero τ0(F,ω)∈ℕ∪{∞}\tau_0(F, \omega) \in \mathbb{N} \cup \{\infty\} at ω∈Ω\omega \in \Omega is the least n∈ℕn \in \mathbb{N} such that F(n,ω)=0F(n, \omega) = 0, or ∞\infty if no such nn exists.✦

For a sequence of nonnegative integer defect counts X=(Xj)j≥0X = (X_j)_{j \ge 0} on Ω\Omega, the defect walk Wn(X,x)∈ℤW_n(X, x) \in \mathbb{Z} at step n∈ℕn \in \mathbb{N} and x∈Ωx \in \Omega is the sum ∑j=0n−1(1−Xj(x))\sum_{j=0}^{n-1} (1 - X_j(x))✦, and the defect walk survival event An(X)⊆ΩA_n(X) \subseteq \Omega at step nn is the subset {x∈Ω:Wk(X,x)≤0 for all k≤n}\{x \in \Omega : W_k(X, x) \le 0 \text{ for all } k \le n\}.✦ The defect stopping index Nstop(X,x)∈ℕN_{\mathrm{stop}}(X, x) \in \mathbb{N} is the first step at which Wk(X,x)>0W_k(X, x) > 0, or 00 if Wk(X,x)≤0W_k(X, x) \le 0 for all k∈ℕk \in \mathbb{N}.✦

For a one-sided sequence x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} and n∈ℕn \in \mathbb{N}, the revealed word Rn(x)∈𝒜*R_n(x) \in \mathcal{A}^* of length nn is the reversed prefix (x0,…,xn−1)rev(x_0, \dots, x_{n-1})^{\mathrm{rev}} (with R0(x)=[]R_0(x) = [])✦, and the shifted letter stream θnx∈𝒜ℕ\theta^n x \in \mathcal{A}^{\mathbb{N}} at shift nn is the sequence defined by (θnx)k=xn+k(\theta^n x)_k = x_{n+k} for k∈ℕk \in \mathbb{N}.✦ Given a map T:𝒜ℕ→ℕ∪{∞}T : \mathcal{A}^{\mathbb{N}} \to \mathbb{N} \cup \{\infty\} and x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}}, the stopped word Wstop(T,x)∈𝒜*W_{\mathrm{stop}}(T, x) \in \mathcal{A}^* at TT is the word Rm+1(x)R_{m+1}(x), where m=T(x)m = T(x) if T(x)<∞T(x) < \infty and m=0m = 0 if T(x)=∞T(x) = \infty.✦

For x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} and step n∈ℕn \in \mathbb{N}, the survivor indicator ϕ(n,x)∈ℤ\phi(n, x) \in \mathbb{Z} is 11 if redburg(Rn+1(x))=[]\mathrm{red}_{\mathrm{burg}}(R_{n+1}(x)) = [] and 00 otherwise✦, and the first backward survivor time σ(x)∈ℕ∪{∞}\sigma(x) \in \mathbb{N} \cup \{\infty\} of xx is τ0(ϕ,x)\tau_0(\phi, x).✦ The survivor order count K(x)∈ℕK(x) \in \mathbb{N} of x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} is the integer equal to 00 if σ(x)=∞\sigma(x) = \infty, and to |redord(Rn+1(x))||\mathrm{red}_{\mathrm{ord}}(R_{n+1}(x))| if σ(x)=n<∞\sigma(x) = n < \infty.✦ Writing m=σ(x)m = \sigma(x) when σ(x)<∞\sigma(x) < \infty and m=0m = 0 when σ(x)=∞\sigma(x) = \infty, the first survivor type β(x)∈{false,true}\beta(x) \in \{\mathrm{false}, \mathrm{true}\} of x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} is the freshest element of redburg(Rm+1(x))\mathrm{red}_{\mathrm{burg}}(R_{m+1}(x)) or false\mathrm{false} if that burger stack is empty✦, and the first survivor data (K(x),β(x))∈ℕ×{false,true}(K(x), \beta(x)) \in \mathbb{N} \times \{\mathrm{false}, \mathrm{true}\} of xx is the pair formed by the survivor order count and first survivor type.✦

The first survivor tail Θ(x)∈𝒜ℕ\Theta(x) \in \mathcal{A}^{\mathbb{N}} of x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} is the shifted stream θm+1x\theta^{m+1} x, where m=σ(x)m = \sigma(x) if σ(x)<∞\sigma(x) < \infty and m=0m = 0 if σ(x)=∞\sigma(x) = \infty.✦ For j∈ℕj \in \mathbb{N} and x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}}, the parsing step survivor data at step jj is the pair (K(Θj(x)),β(Θj(x)))∈ℕ×{false,true}(K(\Theta^j(x)), \beta(\Theta^j(x))) \in \mathbb{N} \times \{\mathrm{false}, \mathrm{true}\}✦, the parsing step defect count Kj(x)∈ℕK_j(x) \in \mathbb{N} at step jj is K(Θj(x))K(\Theta^j(x))✦, and with the defect sequence K•=(Kj)j≥0K_\bullet = (K_j)_{j \ge 0}, the left parsing step count L(x)∈ℕL(x) \in \mathbb{N} of xx is the defect stopping index Nstop(K•,x)N_{\mathrm{stop}}(K_\bullet, x).✦

The short piece prefix Pshort(n,x)∈𝒜*P_{\mathrm{short}}(n, x) \in \mathcal{A}^* of x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} at step n∈ℕn \in \mathbb{N} is defined recursively by Pshort(0,x)=[]P_{\mathrm{short}}(0, x) = [], while Pshort(n+1,x)P_{\mathrm{short}}(n + 1, x) concatenates the reduced word of the stopped word Wstop(σ,Θn(x))W_{\mathrm{stop}}(\sigma, \Theta^n(x)) of xx with Pshort(n,x)P_{\mathrm{short}}(n, x).✦ The short left piece word WLshort(x)∈𝒜*W_L^{\mathrm{short}}(x) \in \mathcal{A}^* of x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} is the word Pshort(L(x),x)P_{\mathrm{short}}(L(x), x)✦, and for p∈(1/2,1)p \in (1/2, 1) the shortened left piece law μL,p\mu_{L,p} on 𝒜*\mathcal{A}^* is the pushforward of ℙpℕ\mathbb{P}_p^{\mathbb{N}} under WLshortW_L^{\mathrm{short}}.✦

For an initial level h∈ℕh \in \mathbb{N}, a stream x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}}, and step n∈ℕn \in \mathbb{N}, the downward inventory path Dh(n,x)∈ℤD_h(n, x) \in \mathbb{Z} is h+Sn+1(x)h + S_{n+1}(x)✦, and the first downward hitting time τh↓(x)∈ℕ∪{∞}\tau^{\downarrow}_h(x) \in \mathbb{N} \cup \{\infty\} from hh is τ0(Dh,x)\tau_0(D_h, x).✦ The reversed right piece word WRrev(x)∈𝒜*W_R^{\mathrm{rev}}(x) \in \mathcal{A}^* of x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} is the stopped word Wstop(τ1↓,x)W_{\mathrm{stop}}(\tau^{\downarrow}_1, x).✦ For p∈(1/2,1)p \in (1/2, 1), the reversed right piece law μR,prev\mu_{R,p}^{\mathrm{rev}} on 𝒜*\mathcal{A}^* is the pushforward of ℙpℕ\mathbb{P}_p^{\mathbb{N}} under WRrevW_R^{\mathrm{rev}}✦, and the right piece law μR,p\mu_{R,p} on 𝒜*\mathcal{A}^* is the pushforward of μR,prev\mu_{R,p}^{\mathrm{rev}} under word reversal w↦wrevw \mapsto w^{\mathrm{rev}}.✦

Piece Prefixes, Cuts, and Remainder Scans

For a sequence of words y=(yn)n≥0∈(𝒜*)ℕy = (y_n)_{n \ge 0} \in (\mathcal{A}^*)^{\mathbb{N}}, the concatenated piece prefix prefn(y)∈𝒜*\mathrm{pref}_n(y) \in \mathcal{A}^* at index n∈ℕn \in \mathbb{N} is defined recursively by pref0(y)=[]\mathrm{pref}_0(y) = [] and the concatenation prefn+1(y)=prefn(y)yn\mathrm{pref}_{n+1}(y) = \mathrm{pref}_n(y) \, y_n.✦ The left-cut event 𝒞L⊆(𝒜*)ℕ\mathcal{C}_L \subseteq (\mathcal{A}^*)^{\mathbb{N}} is the set of sequences y∈(𝒜*)ℕy \in (\mathcal{A}^*)^{\mathbb{N}} such that redord(prefn(y))=[]\mathrm{red}_{\mathrm{ord}}(\mathrm{pref}_n(y)) = [] for every n∈ℕn \in \mathbb{N}✦, and the right-cut event 𝒞R⊆(𝒜*)ℕ\mathcal{C}_R \subseteq (\mathcal{A}^*)^{\mathbb{N}} is the set of sequences y∈(𝒜*)ℕy \in (\mathcal{A}^*)^{\mathbb{N}} such that redburg(prefn(y)rev)=[]\mathrm{red}_{\mathrm{burg}}(\mathrm{pref}_n(y)^{\mathrm{rev}}) = [] for every n∈ℕn \in \mathbb{N}.✦

The letter scan scan(y)∈𝒜ℕ\mathrm{scan}(y) \in \mathcal{A}^{\mathbb{N}} of a sequence of words y∈(𝒜*)ℕy \in (\mathcal{A}^*)^{\mathbb{N}} is the letter sequence whose jj-th entry (j∈ℕj \in \mathbb{N}) is the jj-th letter of prefj+1(y)\mathrm{pref}_{j+1}(y) if |prefj+1(y)|>j|\mathrm{pref}_{j+1}(y)| > j, and the flexible order O*O_\ast otherwise.✦ For p∈(1/2,1)p \in (1/2, 1), the right scan law ηR,p\eta_{R,p} on 𝒜ℕ\mathcal{A}^{\mathbb{N}} is the pushforward of (μR,prev)⊗ℕ(\mu_{R,p}^{\mathrm{rev}})^{\otimes \mathbb{N}} under scan\mathrm{scan}✦, and the right remainder law ρR,p\rho_{R,p} on 𝒜ℕ\mathcal{A}^{\mathbb{N}} is the pushforward of ηR,p\eta_{R,p} under the one-step shift θ1\theta^1.✦

On a two-sided word sequence y=(yk)k∈ℤ∈(𝒜*)ℤy = (y_k)_{k \in \mathbb{Z}} \in (\mathcal{A}^*)^{\mathbb{Z}}, the concatenated interval word Ia,n(y)∈𝒜*I_{a,n}(y) \in \mathcal{A}^* starting at a∈ℤa \in \mathbb{Z} and spanning n∈ℕn \in \mathbb{N} pieces is prefn((ya+j)j≥0)\mathrm{pref}_n((y_{a+j})_{j \ge 0}).✦ A sequence y∈(𝒜*)ℤy \in (\mathcal{A}^*)^{\mathbb{Z}} has a left cut at a∈ℤa \in \mathbb{Z} if redord(Ia,n(y))=[]\mathrm{red}_{\mathrm{ord}}(I_{a,n}(y)) = [] for every n∈ℕn \in \mathbb{N}✦, and has a right cut at b∈ℤb \in \mathbb{Z} if redburg(Ib−n,n(y))=[]\mathrm{red}_{\mathrm{burg}}(I_{b-n,n}(y)) = [] for every n∈ℕn \in \mathbb{N}.✦ For p∈(1/2,1)p \in (1/2, 1), the left-cut density dL,p∈ℝd_{L,p} \in \mathbb{R} is the probability under μL,p⊗ℤ\mu_{L,p}^{\otimes \mathbb{Z}} that a two-sided sequence y∈(𝒜*)ℤy \in (\mathcal{A}^*)^{\mathbb{Z}} has a left cut at 00✦, and the right-cut density dR,p∈ℝd_{R,p} \in \mathbb{R} is the probability under μR,p⊗ℤ\mu_{R,p}^{\otimes \mathbb{Z}} that yy has a right cut at 00.✦

The Claim

Theorem 1. Let p∈(1/2,1)p \in (1/2, 1). For the shortened left piece law μL,p\mu_{L,p}, word length satisfies (3) and the left-cut event satisfies (4): ∫𝒜*|w|dμL,p(w)<∞,\int_{\mathcal{A}^*} |w| \, d\mu_{L,p}(w) < \infty,(3)✦ and μL,p⊗ℕ(𝒞L)>0.\mu_{L,p}^{\otimes \mathbb{N}}(\mathcal{C}_L) > 0.(4)✦ For the right piece constructions, the right remainder law satisfies (5) and the reversed right piece law satisfies (6): ρR,p({x∈𝒜ℕ:σ(x)=∞})>0,\rho_{R,p}(\{x \in \mathcal{A}^{\mathbb{N}} : \sigma(x) = \infty\}) > 0,(5)✦ and (μR,prev)⊗ℕ(𝒞R)>0.(\mu_{R,p}^{\mathrm{rev}})^{\otimes \mathbb{N}}(\mathcal{C}_R) > 0.(6)✦ Moreover, the left-cut and right-cut densities satisfy (7) and (8): dL,p>0,d_{L,p} > 0,(7)✦ and dR,p>0.d_{R,p} > 0.(8)✦ More generally, let ν\nu be a probability measure on 𝒜*\mathcal{A}^* such that ν\nu-almost every word is a fixed word and w↦|w|w \mapsto |w| is ν\nu-integrable. Suppose further that for every b∈{false,true}b \in \{\mathrm{false}, \mathrm{true}\}, the net type count w↦Δb(w)w \mapsto \Delta_b(w) is ν\nu-integrable with ∫𝒜*Δb(w)dν(w)=1/2\int_{\mathcal{A}^*} \Delta_b(w) \, d\nu(w) = 1/2, while ν({Bb})>0\nu(\{B_b\}) > 0. Then the natural pieces law of ν\nu satisfies (9): ν⊗ℕ(𝒞L)>0.\nu^{\otimes \mathbb{N}}(\mathcal{C}_L) > 0.(9)✦

In the paper, the result reads: each fixed boundary has positive probability of being a left cut (𝒞L\mathcal{C}_L) and positive probability of being a right cut (𝒞R\mathcal{C}_R) in the respective piece laws, and consequently both cut sets have positive limiting densities in both index directions. Here μL,p\mu_{L,p} and μR,prev\mu_{R,p}^{\mathrm{rev}} are the piece laws, with ρR,p\rho_{R,p} governing the right remainder and ν\nu allowing general fixed-order laws, while the coordinate-zero probabilities dL,pd_{L,p} and dR,pd_{R,p} under μL,p⊗ℤ\mu_{L,p}^{\otimes \mathbb{Z}} and μR,p⊗ℤ\mu_{R,p}^{\otimes \mathbb{Z}} represent the limiting cut densities.

What Is Checked

For p∈(1/2,1)p \in (1/2, 1), the claims are stated over the piece and remainder laws μL,p\mu_{L,p}, μR,prev\mu_{R,p}^{\mathrm{rev}}, and ρR,p\rho_{R,p} together with the cut events 𝒞L\mathcal{C}_L and 𝒞R\mathcal{C}_R, and over the cut densities dL,pd_{L,p} and dR,pd_{R,p} under μL,p⊗ℤ\mu_{L,p}^{\otimes \mathbb{Z}} and μR,p⊗ℤ\mu_{R,p}^{\otimes \mathbb{Z}}. The general left-cut statement is stated over any probability measure ν\nu on 𝒜*\mathcal{A}^* concentrated almost surely on fixed words, for which word length is integrable and each b∈{false,true}b \in \{\mathrm{false}, \mathrm{true}\} has integrable net type count Δb\Delta_b with mean 1/21/2 and positive single-burger mass ν({Bb})>0\nu(\{B_b\}) > 0.