OctoInformBeta

OpenAI preprint, September 2026 · Section 4. Signed multilinear forms and conductor decay

Section 4 (setting): tensor laws and the functional: Diagonal roots and exact coordinate support for real functions

Multilinear functionals for product tuple laws.

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 permutation word is a function assigning a permutation of a fixed finite set of positions to each block; a tuple configuration assigns a field value to each such word.✦ Tensor products of tuple laws give configurations valued in a product of prime fields, on which real functions are evaluated jointly across the vertices.

Permutation Configurations

Write [k]={0,…,k−1}[k]=\{0,\ldots,k-1\} for a natural number kk. The fixed height hh and block cutoff TT determine the block set BB, list base rr and interaction parameter δ\delta in (1): h=24,T=80,B=[2T+1]=[161],r=h!2,δ=110h2.h=24,\qquad T=80,\qquad B=[2T+1]=[161],\qquad r=\frac{h!}{2},\qquad \delta=\frac{1}{10h^2}.(1)

Fix a block set CC and a height kk. The word space 𝒲(C,k)\mathcal W(C,k) and the fixed tuple vertex set VV are given by (2): 𝒲(C,k)=Sym⁡([k])C,V=𝒲(B,h).\mathcal W(C,k)=\operatorname{Sym}([k])^C, \qquad V=\mathcal W(B,h).(2)

For an active block set S⊆BS\subseteq B, an inactive-block address aa is a function assigning a permutation of [h][h] to each block of B\SB\setminus S.✦ Its address space ASA_S, the addressed list index set ISI_S and the addressed list entry set ℒS\mathcal L_S are those in (3): AS=Sym⁡([h])B\S,IS=[r|S|],ℒS=AS×IS.A_S=\operatorname{Sym}([h])^{B\setminus S},\qquad I_S=[r^{|S|}],\qquad \mathcal L_S=A_S\times I_S.(3) The inactive-block projection πS:V→AS\pi_S:V\to A_S is the restriction in (4): πS(v)=v|B\S.\pi_S(v)=v|_{B\setminus S}.(4)✦

The ordered position pairs of height kk form the set PkP_k in (5): Pk={(s,t)∈[k]2:s<t}.P_k=\{(s,t)\in[k]^2:s<t\}.(5)✦ For a block set CC and height kk, let v∈𝒲(C,k)v\in\mathcal W(C,k), b∈Cb\in C and (s,t)∈Pk(s,t)\in P_k. The position transposition vb;s,t∈𝒲(C,k)v^{b;s,t}\in\mathcal W(C,k) is the word in (6), where (st)(s\ t) denotes a transposition: vb;s,t(a)={v(b)∘(st),a=b,v(a),a≠b.v^{b;s,t}(a)= \begin{cases} v(b)\circ(s\ t),&a=b,\\ v(a),&a\ne b. \end{cases}(6)✦ For CC finite, S⊆CS\subseteq C and a real parameter η\eta, the transposition interaction HS,ηH_{S,\eta} is the real matrix on 𝒲(C,k)\mathcal W(C,k) defined by the counting sum in (7): HS,η(v,w)=∑b∈S∑(s,t)∈Pkηt−s𝟏{w=vb;s,t}.H_{S,\eta}(v,w)=\sum_{b\in S}\sum_{(s,t)\in P_k} \eta^{t-s}\mathbf 1_{\{w=v^{b;s,t}\}}.(7)✦

The cyclic shift c∈Sym⁡([h])c\in\operatorname{Sym}([h]) sends each position to its successor modulo hh.✦ For a finite block set CC, a group GG, a shift a∈Ga\in G and a natural threshold qq, the threshold shift adjacency relation u∼a,qwu\sim_{a,q}w on words u,w∈GCu,w\in G^C is defined by (8): u∼a,qw⇔|{b∈C:w(b)=u(b)a}|≥q.u\sim_{a,q}w \quad\Longleftrightarrow\quad \bigl|\{b\in C:w(b)=u(b)a\}\bigr|\ge q.(8)✦

Tuple Laws

For a finite index set II, a label set KK, a real matrix HH on II and labelings j,k:I→Kj,k:I\to K, the label coincidence pairing CH(j,k)C_H(j,k) is the counting sum in (9): CH(j,k)=∑x∈I∑y∈IH(x,y)𝟏{j(x)=k(y)}.C_H(j,k)=\sum_{x\in I}\sum_{y\in I}H(x,y) \mathbf 1_{\{j(x)=k(y)\}}.(9)✦ The index energy EH(j)E_H(j) and index Boltzmann weight wH(j)w_H(j) of a labeling j:I→Kj:I\to K are the real numbers in (10): EH(j)=12CH(j,j),wH(j)=exp⁡(EH(j)).E_H(j)=\frac12 C_H(j,j),\qquad w_H(j)=\exp(E_H(j)).(10) For a labeling j:V→ISj:V\to I_S, the tuple Boltzmann weight wS(j)w_S(j) is the real number defined using the transposition interaction at the fixed parameter δ\delta in (11): wS(j)=wHS,δ(j).w_S(j)=w_{H_{S,\delta}}(j).(11)✦

For a field FF, the nonzero square indicator sF:F→ℝs_F:F\to\mathbb R is the function in (12): sF(x)={1,x≠0 and x is a square in F,0,otherwise.s_F(x)= \begin{cases} 1,&x\ne0\text{ and }x\text{ is a square in }F,\\ 0,&\text{otherwise}. \end{cases}(12)✦ Fix a prime pp. The list constraint edges form a finite set ℰS⊆ℒS×ℒS\mathcal E_S\subseteq\mathcal L_S\times\mathcal L_S of ordered pairs specifying square-difference constraints. The square-constraint indicator Qp,SQ_{p,S} of a list Y:ℒS→𝔽pY:\mathcal L_S\to\mathbb F_p is the product in (13): Qp,S(Y)=∏(e1,e2)∈ℰSs𝔽p(Y(e2)−Y(e1)).Q_{p,S}(Y)=\prod_{(e_1,e_2)\in\mathcal E_S} s_{\mathbb F_p}\bigl(Y(e_2)-Y(e_1)\bigr).(13)✦

For a natural number mm, a list Y:ℒS→ℤ/mℤY:\mathcal L_S\to\mathbb Z/m\mathbb Z and a labeling j:V→ISj:V\to I_S, the sampled tuple configuration zS,Y,j∈(ℤ/mℤ)Vz_{S,Y,j}\in(\mathbb Z/m\mathbb Z)^V assigns to each vertex vv the list value at (πS(v),j(v))(\pi_S(v),j(v)).✦ For m=pm=p prime, this is (14). zS,Y,j(v)=Y(πS(v),j(v)).z_{S,Y,j}(v)=Y\bigl(\pi_S(v),j(v)\bigr).(14)✦ For a real observable G:𝔽pV→ℝG:\mathbb F_p^V\to\mathbb R, the tuple component average 𝒞p,S(G)\mathcal C_{p,S}(G) is the uniform probability average in (15): 𝒞p,S(G)=𝔼Y∈𝔽pℒS𝔼j∈ISVQp,S(Y)wS(j)G(zS,Y,j).\mathcal C_{p,S}(G)= \mathbb E_{Y\in\mathbb F_p^{\mathcal L_S}} \mathbb E_{j\in I_S^V} Q_{p,S}(Y)w_S(j)G(z_{S,Y,j}).(15)✦

For a natural number mm and a subset S⊆BS\subseteq B, the block subset mixture weight αm(S)\alpha_m(S) is the real power of mm with exponent −|S|/16-|S|/16.✦ For a prime pp, the unnormalized tuple mixture functional ℳp\mathcal M_p on real observables G:𝔽pV→ℝG:\mathbb F_p^V\to\mathbb R is given by the counting sum in (16). αp(S)=p−|S|/16,ℳp(G)=∑S⊆B|S|≤T+1αp(S)𝒞p,S(G).\alpha_p(S)=p^{-|S|/16},\qquad \mathcal M_p(G)= \sum_{\substack{S\subseteq B\\|S|\le T+1}} \alpha_p(S)\mathcal C_{p,S}(G).(16)✦ The full tuple law μp\mu_p is the real linear functional obtained by normalizing this mixture, as in (17), where 11 is the constant observable: μp(G)=ℳp(G)ℳp(1).\mu_p(G)=\frac{\mathcal M_p(G)}{\mathcal M_p(1)}.(17)✦

Tensor Products

For a real linear functional LL on real functions on a set Ω\Omega, the point mass ρL(x)\rho_L(x) at x∈Ωx\in\Omega is the real number in (18): ρL(x)=L(𝟏{x}).\rho_L(x)=L\bigl(\mathbf 1_{\{x\}}\bigr).(18)✦ Let JJ and WW be finite sets, let XjX_j be finite for each j∈Jj\in J, and let LjL_j be a real linear functional on functions XjW→ℝX_j^W\to\mathbb R. For a real observable G:(∏j∈JXj)W→ℝG:(\prod_{j\in J}X_j)^W\to\mathbb R, the tensor product law ⨂j∈JLj\bigotimes_{j\in J}L_j of the real linear functionals LjL_j on functions XjW→ℝX_j^W\to\mathbb R is the real linear functional defined by the counting sum in (19). (⨂j∈JLj)(G)=∑z∈∏j∈JXjW(∏j∈JρLj(zj))G((zj(w))w,j).\left(\bigotimes_{j\in J}L_j\right)(G)= \sum_{z\in\prod_{j\in J}X_j^W} \left(\prod_{j\in J}\rho_{L_j}(z_j)\right) G\bigl((z_j(w))_{w,j}\bigr).(19)✦ For x∈∏j∈JXjx\in\prod_{j\in J}X_j and t∈Xjt\in X_j, write xj←tx^{j\leftarrow t} for xx with its jjth coordinate replaced by tt.

The Definition

Definition 1. Let JJ be finite and let pjp_j be prime for each j∈Jj\in J. Writing 𝒑\mathbf p for the family (pj)j∈J(p_j)_{j\in J}, the product tuple law Π𝒑\Pi_{\mathbf p} on (∏j∈J𝔽pj)V(\prod_{j\in J}\mathbb F_{p_j})^V is the tensor product in (20): Π𝒑=⨂j∈Jμpj.\Pi_{\mathbf p}=\bigotimes_{j\in J}\mu_{p_j}.(20)✦

Let WW be a finite vertex set, let XX be a set, and let LL be a real linear functional on functions XW→ℝX^W\to\mathbb R. For real functions aw:X→ℝa_w:X\to\mathbb R, the multilinear integral ML((aw)w∈W)M_L((a_w)_{w\in W}) is the real number in (21): ML((aw)w∈W)=L(z↦∏w∈Waw(z(w))).M_L((a_w)_{w\in W})= L\left(z\longmapsto\prod_{w\in W}a_w(z(w))\right).(21)✦ For a real function g:X→ℝg:X\to\mathbb R, the diagonal integral DL(g)D_L(g) is the real number obtained by specializing the multilinear integral as in (22): DL(g)=ML((g)w∈W).D_L(g)=M_L((g)_{w\in W}).(22)✦ For a real function g:X→ℝg:X\to\mathbb R and a real linear functional LL on functions XW→ℝX^W\to\mathbb R, where WW is finite and d=|W|d=|W|, the diagonal root functional RL(g)R_L(g) is the real power in (23). RL(g)=DL(g)1/d.R_L(g)=D_L(g)^{1/d}.(23)✦

Let XjX_j be finite sets indexed by a set JJ, and let U⊆JU\subseteq J be finite. The class ℋU\mathcal H_U of real functions on ∏j∈JXj\prod_{j\in J}X_j with exact coordinate support UU consists of the functions satisfying (24): f∈ℋU⇔{𝔼t∈Xjf(xj←t)=0,j∈U,x∈∏i∈JXi,f(xj←t)=f(x),j∈J\U,x∈∏i∈JXi,t∈Xj.f\in\mathcal H_U \quad\Longleftrightarrow\quad \begin{cases} \mathbb E_{t\in X_j}f(x^{j\leftarrow t})=0, &j\in U,\ x\in\prod_{i\in J}X_i,\\ f(x^{j\leftarrow t})=f(x), &j\in J\setminus U,\ x\in\prod_{i\in J}X_i,\ t\in X_j. \end{cases}(24)✦

What Is Checked

The product law (20) uses a finite family of prime fields and the fixed tuple vertices; the forms (21), (22) and (23) allow an arbitrary real linear functional on configurations over a finite vertex set; exact support (24) is stated for real functions on products of finite coordinate sets with a finite prescribed support. The unnormalized tuple mixture preserves addition of observables and multiplication by real scalars.