OctoInformBeta

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

Section 3 (setting): the fixed tuple and marked components: Square constraints and Boltzmann weights on addressed lists over prime fields

Permutation tuples with marked list components.

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

The finite field configurations considered here are indexed by permutation words and sampled from lists subject to nonzero square constraints. For a nonnegative integer mm, write [m]={0,…,m−1}[m]=\{0,\dots,m-1\}. A permutation word vv on a block set BB with alphabet [m][m] is a choice of a permutation vbv_b for every b∈Bb\in B, belonging to the space in (1). 𝒲(B,m)=Sym⁡([m])B.\mathcal W(B,m)=\operatorname{Sym}([m])^B.(1)✦ Permutations are composed as functions. For a,d∈[m]a,d\in[m], the permutation (ad)(a\ d) exchanges aa and dd; when a=da=d, it is the identity.

An ordered position pair (a,d)(a,d) in [m]2[m]^2 is an element of the set PmP_m in (2). Pm={(a,d)∈[m]2:a<d}.P_m=\{(a,d)\in[m]^2:a<d\}.(2)✦ For v∈𝒲(B,m)v\in\mathcal W(B,m), b∈Bb\in B and (a,d)∈Pm(a,d)\in P_m, the position transposition vb;a,dv^{b;a,d} is the word defined by (3). (vb;a,d)e={vb∘(ad),e=b,ve,e≠b.(v^{b;a,d})_e= \begin{cases} v_b\circ(a\ d),&e=b,\\ v_e,&e\ne b. \end{cases}(3)✦

For a finite block set CC, a group KK, an element c∈Kc\in K and an integer t≥0t\geq0, threshold shift adjacency →c,t\longrightarrow_{c,t} is the relation on KCK^C in (4). u→c,tv⇔|{b∈C:vb=ubc}|≥t.u\longrightarrow_{c,t}v \quad\Longleftrightarrow\quad \bigl|\{b\in C:v_b=u_bc\}\bigr|\geq t.(4)✦

Transposition Weights

For finite BB, a marked set S⊆BS\subseteq B and a real parameter η\eta, the transposition interaction HS,ηH_{S,\eta} is the real matrix on 𝒲(B,m)\mathcal W(B,m) defined by (5); its sums are counting sums and 𝟏E\mathbf 1_E denotes the indicator of a condition EE. HS,η(v,w)=∑b∈S∑(a,d)∈Pmηd−a𝟏{w=vb;a,d}.H_{S,\eta}(v,w)= \sum_{b\in S}\sum_{(a,d)\in P_m} \eta^{d-a}\,\mathbf 1_{\{w=v^{b;a,d}\}}.(5)✦

Let II be finite, let JJ be a label set, and let AA be a real matrix indexed by II. The label coincidence pairing CAC_A of label assignments j,k:I→Jj,k:I\to J is the real function defined by (6). CA(j,k)=∑x∈I∑y∈IA(x,y)𝟏{j(x)=k(y)}.C_A(j,k)=\sum_{x\in I}\sum_{y\in I}A(x,y)\, \mathbf 1_{\{j(x)=k(y)\}}.(6)✦ The index energy EAE_A and Boltzmann weight wAw_A on label assignments j:I→Jj:I\to J are the real functions defined by (7) and (8), respectively. EA(j)=12CA(j,j).E_A(j)=\frac12 C_A(j,j).(7)✦ wA(j)=exp⁡(EA(j)).w_A(j)=\exp(E_A(j)).(8)✦

For a field 𝕂\mathbb K, the nonzero square indicator σ𝕂\sigma_{\mathbb K} is the real function on 𝕂\mathbb K defined by (9). σ𝕂(x)={1,x≠0 and x is a square in 𝕂,0,otherwise.\sigma_{\mathbb K}(x)= \begin{cases} 1,&x\ne0\text{ and }x\text{ is a square in }\mathbb K,\\ 0,&\text{otherwise}. \end{cases}(9)✦

The Definition

Definition 1. The fixed alphabet size hh and threshold TT are given by (10) and (11), respectively. h=24.h=24.(10)✦ T=80.T=80.(11)✦ The number of blocks is nn in (12), and the block set is B=[n]B=[n]. n=2T+1.n=2T+1.(12)✦ The tuple vertex space VV is the fixed permutation word space in (13). V=𝒲(B,h).V=\mathcal W(B,h).(13)✦ Let c∈Sym⁡([h])c\in\operatorname{Sym}([h]) be the cyclic shift a↦a+1mod⁡ha\mapsto a+1\bmod h.✦ For each integer j≥0j\geq0, the constant cyclic vertex v(j)∈Vv^{(j)}\in V is defined by (14). vb(j)=cj(b∈B).v^{(j)}_b=c^j\qquad(b\in B).(14)✦

For a block set CC, b∈Cb\in C, α,β∈[m]\alpha,\beta\in[m] and v∈𝒲(C,m)v\in\mathcal W(C,m), the symbol reflection ρb;α,βv\rho_{b;\alpha,\beta}v is the word in (15). (ρb;α,βv)e={(αβ)∘vb,e=b,ve,e≠b.(\rho_{b;\alpha,\beta}v)_e= \begin{cases} (\alpha\ \beta)\circ v_b,&e=b,\\ v_e,&e\ne b. \end{cases}(15)✦ The positive cut condition 𝒫b;α,β(v)\mathcal P_{b;\alpha,\beta}(v) holds when α\alpha precedes β\beta in the permutation vbv_b, for v∈𝒲(C,m)v\in\mathcal W(C,m), b∈Cb\in C and α,β∈[m]\alpha,\beta\in[m], as in (16). 𝒫b;α,β(v)⇔vb−1(α)<vb−1(β).\mathcal P_{b;\alpha,\beta}(v) \quad\Longleftrightarrow\quad v_b^{-1}(\alpha)<v_b^{-1}(\beta).(16)✦ For a block set CC, a block b∈Cb\in C and symbols α,β∈[m]\alpha,\beta\in[m], the positive cut half Cb;α,β+C^+_{b;\alpha,\beta} is the subset of 𝒲(C,m)\mathcal W(C,m) defined by (17). Cb;α,β+={v∈𝒲(C,m):𝒫b;α,β(v)}.C^+_{b;\alpha,\beta}= \{v\in\mathcal W(C,m):\mathcal P_{b;\alpha,\beta}(v)\}.(17)✦

For a marked set S⊆BS\subseteq B, the fixed list parameter rr and interaction parameter δ\delta are the constants in (18). r=h!2,δ=110h2.r=\frac{h!}{2},\qquad \delta=\frac{1}{10h^2}.(18) An inactive-block address u∈𝒜Su\in\mathcal A_S is a tuple of permutations of [h][h] indexed by B\SB\setminus S✦; the address space 𝒜S\mathcal A_S and the restriction map πS:V→𝒜S\pi_S:V\to\mathcal A_S are given by (19). 𝒜S=Sym⁡([h])B\S,πS(v)=v|B\S.\mathcal A_S=\operatorname{Sym}([h])^{B\setminus S}, \qquad \pi_S(v)=v|_{B\setminus S}.(19) The list index set JSJ_S is the finite interval in (20). JS=[r|S|].J_S=[r^{|S|}].(20)✦ An addressed list entry (u,k)∈LS(u,k)\in L_S is a pair consisting of an address u∈𝒜Su\in\mathcal A_S and an index k∈JSk\in J_S, as in (21). LS=𝒜S×JS.L_S=\mathcal A_S\times J_S.(21)✦ Let ℰS⊆LS×LS\mathcal E_S\subseteq L_S\times L_S be the specified finite set of list constraint edges.

Let pp be prime. The square-constraint indicator aSa_S on list assignments Y:LS→𝔽pY:L_S\to\mathbb F_p is the real function defined by (22). aS(Y)=∏(u,v)∈ℰSσ𝔽p(Y(v)−Y(u)).a_S(Y)=\prod_{(u,v)\in\mathcal E_S} \sigma_{\mathbb F_p}\bigl(Y(v)-Y(u)\bigr).(22)✦ For a label assignment j:V→JSj:V\to J_S, the sampled configuration zY,j∈𝔽pVz_{Y,j}\in\mathbb F_p^V is given by (23). zY,j(v)=Y(πS(v),j(v)).z_{Y,j}(v)=Y\bigl(\pi_S(v),j(v)\bigr).(23)✦ The interaction HS,δH_{S,\delta} gives label assignments j:V→JSj:V\to J_S the tuple Boltzmann weight wSw_S in (24). wS(j)=wHS,δ(j).w_S(j)=w_{H_{S,\delta}}(j).(24)✦

The marked-list component ΛS\Lambda_S is the functional on real functions G:𝔽pV→ℝG:\mathbb F_p^V\to\mathbb R in (25), with both averages taken with respect to uniform probability. ΛS(G)=𝔼Y∈𝔽pLS𝔼j∈JSVaS(Y)wS(j)G(zY,j).\Lambda_S(G)= \mathbb E_{Y\in\mathbb F_p^{L_S}}\, \mathbb E_{j\in J_S^V} a_S(Y)w_S(j)G(z_{Y,j}).(25)✦ For b∈Bb\in B and α,β∈[h]\alpha,\beta\in[h], take the positive cut half C+=Cb;α,β+⊆VC^+=C^+_{b;\alpha,\beta}\subseteq V and write ρ=ρb;α,β\rho=\rho_{b;\alpha,\beta}. The cut quadratic form on real functions F:𝔽pC+→ℝF:\mathbb F_p^{C^+}\to\mathbb R is QS;b;α,βQ_{S;b;\alpha,\beta} in (26). QS;b;α,β(F)=ΛS(z↦F(z|C+)F((z∘ρ)|C+)).Q_{S;b;\alpha,\beta}(F)= \Lambda_S\!\left(z\mapsto F(z|_{C^+})\,F\bigl((z\circ\rho)|_{C^+}\bigr)\right).(26)✦

The fixed weight bound constant MM is defined by (27). M=exp⁡(|V|(T+1)(h2)δ2).M=\exp\!\left(\frac{|V|(T+1)\binom h2\,\delta}{2}\right).(27)✦

What Is Checked

The component functionals and cut forms are defined over every prime field 𝔽p\mathbb F_p, for arbitrary marked sets S⊆BS\subseteq B and real test functions on the corresponding configuration spaces.