OctoInformBeta

OpenAI preprint, September 2026 · §4. Transverse configurations and determinant signs · §4

Lemma 4.6 [Joint determinant signs]: For even D≥12D\ge 12 and independent uniform symmetric XiX_i over 𝔽q\mathbb{F}_q, the 33 signs χ(det⁡(Xu−Xv))\chi(\det(X_u-X_v)) tend to fair independent coins as q→∞q\to\infty

Pair determinant signs are asymptotically independent and uniform.

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 graph HH of the counterexample to Sidorenko’s conjecture has 13 points and 22 faces, each face a triple of points, and 33 pairs of its points, the point pairs, are singled out. For an odd prime qq, with 𝔽q\mathbb{F}_q the field of integers modulo qq, place at each point ii a symmetric D×DD\times D matrix XiX_i over 𝔽q\mathbb{F}_q, chosen uniformly and independently, and give each point pair {u,v}\{u,v\} the sign χ(det⁡(Xu−Xv))\chi(\det(X_u-X_v)), where χ\chi is the quadratic character of 𝔽q\mathbb{F}_q. The question is how close these 33 signs come to 33 independent fair coins when the dimension DD is fixed and the prime qq grows.

Matrices, Signs and Bits

An odd prime is a natural number qq that is prime and different from 22.✦ Limits as q→∞q\to\infty are taken through the odd primes.✦

The space SDS_D of symmetric matrices of order DD over a field KK is the set of the D×DD\times D matrices MM over KK with MT=MM^{T}=M.

The uniform mean of a real function ff on a finite set AA is its average 𝔼a∈Af(a)=1|A|∑a∈Af(a).\mathbb{E}_{a\in A}\,f(a)=\frac{1}{|A|}\sum_{a\in A}f(a).(1)✦

The points are numbered 0,…,120,\dots,12. Each point pair ee has a left point ueu_e and a right point vev_e, the smaller and the larger of its two points. For a family X=(X0,…,X12)X=(X_0,\dots,X_{12}) in SD13S_D^{13} over 𝔽q\mathbb{F}_q, the pair sign of XX at ee is se(X)=χ(det⁡(Xue−Xve)),s_e(X)=\chi\bigl(\det(X_{u_e}-X_{v_e})\bigr),(2)✦ where χ\chi is the quadratic character of 𝔽q\mathbb{F}_q with χ(0)=0\chi(0)=0; thus se(X)s_e(X) is ±1\pm1 when Xue−XveX_{u_e}-X_{v_e} is invertible and 00 when it is singular.

The pair bit ηeX\eta^X_e of XX at ee is the Boolean value that is true when se(X)=1s_e(X)=1 and false when se(X)s_e(X) is −1-1 or 00.✦ The 33 pair bits form the vector ηX=(ηeX)e\eta^X=(\eta^X_e)_e in {true,false}33\{\mathrm{true},\mathrm{false}\}^{33}.

The Boolean sign σ(b)\sigma(b) of a Boolean value bb is 11 when bb is true and −1-1 when bb is false.✦

Monomials and Bordered Matrices

The Boolean monomial ηα\eta_\alpha of a vector η∈{true,false}I\eta\in\{\mathrm{true},\mathrm{false}\}^I, indexed by a finite set II, at a subset α⊆I\alpha\subseteq I is the real number ηα=∏i∈ασ(ηi).\eta_\alpha=\prod_{i\in\alpha}\sigma(\eta_i).(3)✦

For a symmetric matrix MM of order nn over a field KK, a vector z∈Knz\in K^n and a scalar a∈Ka\in K, the bordered matrix [M;z,a][M;z,a] is the matrix of order n+1n+1 [M;z,a]=(azTzM).[M;z,a]=\begin{pmatrix} a & z^{T}\\ z & M\end{pmatrix}.(4)✦

The Claim

Theorem 1. Let D≥12D\ge 12 be even, and for each odd prime qq let X=(X0,…,X12)X=(X_0,\dots,X_{12}) be uniform on SD13S_D^{13} over 𝔽q\mathbb{F}_q, the XiX_i independent and uniform on SDS_D.

  1. For every real function gg on {true,false}33\{\mathrm{true},\mathrm{false}\}^{33}, with η\eta uniform on {true,false}33\{\mathrm{true},\mathrm{false}\}^{33}, limq→∞𝔼Xg(ηX)=𝔼ηg(η).\lim_{q\to\infty}\mathbb{E}_X\,g(\eta^X)=\mathbb{E}_\eta\,g(\eta).(5)✦

  2. For every nonempty set FF of point pairs, limq→∞𝔼X∏e∈Fσ(ηeX)=0.\lim_{q\to\infty}\mathbb{E}_X\prod_{e\in F}\sigma(\eta^X_e)=0.(6)✦

  3. For every nonempty set FF of point pairs, limq→∞𝔼X∏e∈Fse(X)=0.\lim_{q\to\infty}\mathbb{E}_X\prod_{e\in F}s_e(X)=0.(7)✦

  4. Let ℒ\mathcal{L} be a filter on a set QQ, let II be a finite set, and for each q∈Qq\in Q let TqT_q be a map from a nonempty finite set AqA_q to {true,false}I\{\mathrm{true},\mathrm{false}\}^I. If 𝔼a∈AqTq(a)α\mathbb{E}_{a\in A_q}\,T_q(a)_\alpha tends to 00 along ℒ\mathcal{L} for every nonempty α⊆I\alpha\subseteq I, then for every ε∈{true,false}I\varepsilon\in\{\mathrm{true},\mathrm{false}\}^I limℒ|{a∈Aq:Tq(a)=ε}||Aq|=2−|I|.\lim_{\mathcal{L}}\frac{|\{a\in A_q: T_q(a)=\varepsilon\}|}{|A_q|}=2^{-|I|}.(8)✦

  5. Let KK be a finite field of odd characteristic with quadratic character χ\chi, let ι\iota be a finite set, and for each i∈ιi\in\iota let MiM_i be an invertible symmetric matrix of order NN over KK, wi∈KNw_i\in K^N and ai∈Ka_i\in K. If ∑i∈ιtiMi−1\sum_{i\in\iota}t_iM_i^{-1} has rank at least |ι|+1|\iota|+1 for every nonzero t∈Kιt\in K^\iota, then (𝔼z∈KN∏i∈ιχ(det⁡[Mi;z−wi,ai]))2≤1|K|.\Bigl(\mathbb{E}_{z\in K^N}\prod_{i\in\iota}\chi\bigl(\det[M_i;z-w_i,a_i]\bigr)\Bigr)^2 \le\frac{1}{|K|}.(9)✦

  6. Let KK be a finite field of odd characteristic, ψ\psi a nontrivial additive character of KK, VV a finite-dimensional vector space over KK, BB a symmetric bilinear form on VV with radical rad⁡B\operatorname{rad}B, ll a linear form on VV and c∈Kc\in K. Then |∑x∈Vψ(B(x,x)+l(x)+c)|2≤|V||rad⁡B|.\Bigl|\sum_{x\in V}\psi\bigl(B(x,x)+l(x)+c\bigr)\Bigr|^2\le|V|\,|\operatorname{rad}B|.(10)✦

In the paper, the result reads: for fixed even D≥12D\ge 12, the law of (ηeX)e∈E(\eta^X_e)_{e\in E}, where EE is the set of the 33 point pairs, converges in total variation to that of independent uniform signs on {±1}E\{\pm1\}^E.

On the finite set {true,false}33\{\mathrm{true},\mathrm{false}\}^{33}, the limit (5) for every real function gg is convergence in total variation of the law of ηX\eta^X to the uniform law, with true standing for the sign 11 and false for the signs −1-1 and 00.

What Is Checked

The claim is stated over the odd primes and limits as q→∞q\to\infty through them, the space SDS_D of symmetric matrices over 𝔽q\mathbb{F}_q with the uniform means (1) over it, the left and right points of the 33 point pairs with the pair signs (2) and pair bits they define, the Boolean signs and monomials (3), and the bordered matrices (4).

About these constructions the library proves that the symmetric matrices of order DD form a subspace of the D×DD\times D matrices✦, and that the bordered matrix of a symmetric matrix is symmetric✦.