OctoInformBeta

OpenAI preprint, September 2026 · Section 3. Finite reflection-positive probability laws

Lemma 3.2: Reflection invariance at every block, with arbitrary real functions on the positive cut half

Marked components have symmetric cut forms.

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

Configurations indexed by tuples of permutations combine nonzero square differences with an interaction between label assignments. Fix a prime pp and use the parameters in (1), where BB is the block set: h=24,T=80,B={0,…,2T},r=h!2,δ=110h2.h=24,\qquad T=80,\qquad B=\{0,\ldots,2T\},\qquad r=\frac{h!}{2},\qquad \delta=\frac{1}{10h^2}.(1) A permutation word vv on a block set CC with word length ℓ\ell is a map assigning to each block a permutation of {0,…,ℓ−1}\{0,\ldots,\ell-1\}. Write [h]={0,…,h−1}[h]=\{0,\ldots,h-1\} and 𝔖h\mathfrak S_h for its permutation group. The tuple vertex set VV is the set in (2): V=𝔖hB.V=\mathfrak S_h^B.(2)✦

For a marked block set S⊆BS\subseteq B, an inactive-block address a∈ASa\in A_S is a map assigning a permutation in 𝔖h\mathfrak S_h to each block outside SS.✦ A list index n∈JSn\in J_S is an integer between zero and r|S|−1r^{|S|}-1, and an addressed list entry e∈LSe\in L_S is a pair consisting of an inactive-block address in ASA_S and a list index in JSJ_S. Their sets, together with the inactive-block projection πS\pi_S, are given in (3): AS=𝔖hB\S,JS={0,…,r|S|−1},LS=AS×JS,πS(v)=v|B\S.A_S=\mathfrak S_h^{B\setminus S},\qquad J_S=\{0,\ldots,r^{|S|}-1\},\qquad L_S=A_S\times J_S,\qquad \pi_S(v)=v|_{B\setminus S}.(3)

For maps u,v:C→Gu,v:C\to G from a finite block set to a group, threshold shift adjacency ∼a,t\sim_{a,t} with shift a∈Ga\in G and nonnegative integer threshold tt is the relation in (4). The fixed cyclic shift c∈𝔖hc\in\mathfrak S_h rotates the positions by one.✦ u∼a,tv⇔|{d∈C:v(d)=u(d)a}|≥t.u\sim_{a,t}v \quad\Longleftrightarrow\quad \bigl|\{d\in C:v(d)=u(d)a\}\bigr|\geq t.(4)✦

The list constraint edges ℰS⊆LS×LS\mathcal E_S\subseteq L_S\times L_S are the ordered pairs of addressed list entries on which a nonzero square difference is imposed.✦ The nonzero square indicator qp:𝔽p→ℝq_p:\mathbb F_p\to\mathbb R and the square-constraint indicator QSQ_S for a residue label field Y:LS→𝔽pY:L_S\to\mathbb F_p are defined by (5): qp(x)=𝟏{x≠0,x is a square},QS(Y)=∏(e,f)∈ℰSqp(Y(f)−Y(e)).q_p(x)=\mathbf1_{\{x\ne0,\ x\text{ is a square}\}},\qquad Q_S(Y)=\prod_{(e,f)\in\mathcal E_S}q_p\bigl(Y(f)-Y(e)\bigr).(5)

The Interaction and the Cut

An ordered position pair (i,k)(i,k) is an element of Ph={(i,k)∈[h]2:i<k}P_h=\{(i,k)\in[h]^2:i<k\}.✦ For v∈Vv\in V, b∈Bb\in B, and (i,k)∈Ph(i,k)\in P_h, the position transposition σb,i,kv\sigma_{b,i,k}v changes the permutation at bb by right composition with (ik)(i\ k), as in (6): (σb,i,kv)(d)={v(b)∘(ik),d=b,v(d),d≠b.(\sigma_{b,i,k}v)(d)= \begin{cases} v(b)\circ(i\ k),&d=b,\\ v(d),&d\ne b. \end{cases}(6)✦ The transposition interaction HS:V×V→ℝH_S:V\times V\to\mathbb R is the counting sum in (7): HS(v,w)=∑b∈S∑(i,k)∈Phδk−i𝟏{w=σb,i,kv}.H_S(v,w)=\sum_{b\in S}\sum_{(i,k)\in P_h} \delta^{k-i}\mathbf1_{\{w=\sigma_{b,i,k}v\}}.(7)✦

For a finite index set II, a label set JJ, a real matrix AA on II, and label assignments j,k:I→Jj,k:I\to J, the label coincidence pairing XA(j,k)X_A(j,k) is the real number given in (8). For a finite index set II, a label set JJ, a real matrix HH on II, and a label assignment j:I→Jj:I\to J, the index energy EH(j)E_H(j) is the real number equal to half the label coincidence pairing of jj with itself, and the index Boltzmann weight WH(j)W_H(j) is the real number equal to the exponential of EH(j)E_H(j), as in (8): XA(j,k)=∑x∈I∑y∈IA(x,y)𝟏{j(x)=k(y)},EH(j)=12XH(j,j),WH(j)=exp⁡(EH(j)).X_A(j,k)=\sum_{x\in I}\sum_{y\in I} A(x,y)\mathbf1_{\{j(x)=k(y)\}},\qquad E_H(j)=\frac12X_H(j,j),\qquad W_H(j)=\exp(E_H(j)).(8) For a label assignment j:V→JSj:V\to J_S, the tuple Boltzmann weight WS(j)=WHS(j)W_S(j)=W_{H_S}(j) is the real number given by (7) and (8).

The sampled tuple configuration zY,j:V→𝔽pz_{Y,j}:V\to\mathbb F_p reads the residue label at the address and index selected by each vertex.✦ The component functional νp,S\nu_{p,S} on real functions Φ:𝔽pV→ℝ\Phi:\mathbb F_p^V\to\mathbb R is the weighted average in (9); both averages are uniform probability averages, and YY and jj are independent: zY,j(v)=Y(πS(v),j(v)),νp,S(Φ)=𝔼Y∈𝔽pLS𝔼j∈JSVQS(Y)WS(j)Φ(zY,j).z_{Y,j}(v)=Y\bigl(\pi_S(v),j(v)\bigr),\qquad \nu_{p,S}(\Phi)= \mathbb E_{Y\in\mathbb F_p^{L_S}}\mathbb E_{j\in J_S^V} Q_S(Y)W_S(j)\Phi(z_{Y,j}).(9)

For b∈Bb\in B and α,β∈[h]\alpha,\beta\in[h], the symbol reflection θ=θb,α,β:V→V\theta=\theta_{b,\alpha,\beta}:V\to V changes block bb by left composition with (αβ)(\alpha\ \beta).✦ The positive cut half Cθ⊆VC_\theta\subseteq V of the symbol reflection θ=θb,α,β\theta=\theta_{b,\alpha,\beta} is the set of words in which α\alpha precedes β\beta in block bb, as specified in (10): (θv)(d)={(αβ)∘v(b),d=b,v(d),d≠b,Cθ={v∈V:v(b)−1(α)<v(b)−1(β)}.(\theta v)(d)= \begin{cases} (\alpha\ \beta)\circ v(b),&d=b,\\ v(d),&d\ne b, \end{cases} \qquad C_\theta=\{v\in V:v(b)^{-1}(\alpha)<v(b)^{-1}(\beta)\}.(10) For a configuration z∈𝔽pVz\in\mathbb F_p^V, write Rθz=z∘θR_\theta z=z\circ\theta. The cut bilinear form ℬp,S,θ\mathcal B_{p,S,\theta} on real functions F,G:𝔽pCθ→ℝF,G:\mathbb F_p^{C_\theta}\to\mathbb R is defined in (11): ℬp,S,θ(F,G)=νp,S(z↦F(z|Cθ)G((Rθz)|Cθ)).\mathcal B_{p,S,\theta}(F,G)= \nu_{p,S}\bigl(z\mapsto F(z|_{C_\theta})G((R_\theta z)|_{C_\theta})\bigr).(11)

The Claim

Theorem 1. Let pp be prime, let S⊆BS\subseteq B, and choose b∈Bb\in B and α,β∈[h]\alpha,\beta\in[h]. Set θ=θb,α,β\theta=\theta_{b,\alpha,\beta}. For every real function Φ:𝔽pV→ℝ\Phi:\mathbb F_p^V\to\mathbb R, the component satisfies (12): νp,S(Φ∘Rθ)=νp,S(Φ).\nu_{p,S}(\Phi\circ R_\theta)=\nu_{p,S}(\Phi).(12)✦ For every pair of real functions F,G:𝔽pCθ→ℝF,G:\mathbb F_p^{C_\theta}\to\mathbb R, the cut form satisfies (13): ℬp,S,θ(F,G)=ℬp,S,θ(G,F).\mathcal B_{p,S,\theta}(F,G)=\mathcal B_{p,S,\theta}(G,F).(13)✦

In the paper, the result reads: for every allowed mark set SS and every reflection θ\theta, νp,S\nu_{p,S} is invariant under z↦(zθv)v∈Vz\mapsto(z_{\theta v})_{v\in V}, and its matrix at that cut is symmetric.

What Is Checked

The statements are over every prime pp, every marked block set S⊆BS\subseteq B, and every block and pair of symbols, with arbitrary real functions on the full configuration space for (12) and on the positive cut half’s configuration space for (13).