OctoInformBeta

OpenAI preprint, September 2026 · Section 5. A weighted cancellation argument

§5 (setting): the opposite graph and the weighted sum: Cyclic phases multiply polynomial state weights for paired triangle transversals

Opposite edges and weighted pair terms.

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 choice of one corner in each triangle determines a collection of opposite sides. Paired choices also carry a cyclic sign, which enters the weight attached to the pair.

A triangle system TT on a vertex set VV, indexed by a finite set AA, is a family of injective corner maps vt:ℤ/3ℤ→Vv_t:\mathbb Z/3\mathbb Z\to V such that two triangles sharing two distinct corners have the same index.✦ Fix a subset X⊆VX\subseteq V. A triangle transversal, or state, ss onto XX is a choice of corner index s(t)s(t) for each t∈At\in A such that the selected vertex map in (1) is a bijection from AA onto XX: sˆ(t)=vt(s(t)).\widehat s(t)=v_t(s(t)).(1)✦ Two states rr and ss are partners when their selected vertices differ in every triangle, as in (2): rˆ(t)≠sˆ(t)(t∈A).\widehat r(t)\ne\widehat s(t)\qquad(t\in A).(2)✦ A paired triangle transversal (r,s)(r,s) is an ordered pair of partners, with first state rr and second state ss.✦

Cyclic Signs and Polynomial Factors

The cyclic signed edge weight δ\delta on the corners of a triangle is defined by (3), with corner indices read modulo three: δ(vt(i),vt(j))={0,i=j,1,j=i+1,−1,otherwise.\delta(v_t(i),v_t(j))= \begin{cases} 0,&i=j,\\ 1,&j=i+1,\\ -1,&\text{otherwise}. \end{cases}(3)✦ The total cyclic increment R(r,s)R(r,s) of two states is the integer in (4): R(r,s)=∑t∈Aδ(rˆ(t),sˆ(t)).R(r,s)=\sum_{t\in A}\delta(\widehat r(t),\widehat s(t)).(4)✦

For a finite index set BB and coefficients btb_t in a commutative ring, the linear factor product LB(b;x)L_B(b;x) is the polynomial in (5): LB(b;x)=∏t∈B(1+btx).L_B(b;x)=\prod_{t\in B}(1+b_t x).(5)✦ Let a(t,v)a(t,v) be a real incidence weight for every corner vv of every triangle tt, and let i\mathrm{i} denote the imaginary unit.

The Definition

Definition 1. Let TT be a triangle system with finite triangle index set AA, let ss be a state onto X⊆VX\subseteq V, and let (r,s)(r,s) be a paired triangle transversal. The opposite graph QsQ_s is the simple graph on VV with adjacency given by (6): v∼Qsw⇔v≠w and there exist t∈A,i,j∈ℤ/3ℤ such thatv=vt(i),w=vt(j),i≠s(t),j≠s(t).\begin{split} v\sim_{Q_s}w\quad\Longleftrightarrow\quad &v\ne w\ \text{ and there exist }t\in A, \ i,j\in\mathbb Z/3\mathbb Z\text{ such that}\\ &v=v_t(i),\quad w=v_t(j),\quad i\ne s(t),\quad j\ne s(t). \end{split}(6)✦ For an integer nn, the phase factor ϕ(n)∈ℂ\phi(n)\in\mathbb C is the complex number given by (7): ϕ(n)=i⌊n/3⌋.\phi(n)=\mathrm{i}^{\lfloor n/3\rfloor}.(7)✦ For real incidence weights a(t,v)a(t,v), the state weight polynomial Ws(x)∈ℂ[x]W_s(x)\in\mathbb C[x] is the product in (8): Ws(x)=∏t∈A(1+a(t,sˆ(t))x).W_s(x)=\prod_{t\in A}\bigl(1+a(t,\widehat s(t))x\bigr).(8)✦ The weighted pair term τr,s(x)∈ℂ[x]\tau_{r,s}(x)\in\mathbb C[x] of a paired triangle transversal (r,s)(r,s) is the polynomial given by (9): τr,s(x)=ϕ(R(r,s))Ws(x).\tau_{r,s}(x)=\phi\bigl(R(r,s)\bigr)W_s(x).(9)✦

In the paper, the result reads: in a plane triangulation with triangles colored black and white, let AA consist of the black triangles other than the outer triangle; for a state ss, QsQ_s contains the edge opposite sˆ(t)\widehat s(t) in every nonouter black triangle t∈At\in A; for real incidence weights a(t,v)a(t,v), put ω(s)=∑t∈Aa(t,sˆ(t))\omega(s)=\sum_{t\in A}a(t,\widehat s(t)) and J(r,s)=13∑t∈Aδ(rˆ(t),sˆ(t))J(r,s)=\frac13\sum_{t\in A}\delta(\widehat r(t),\widehat s(t)); then Z(x)=∑(r,s)iJ(r,s)exp⁡(xω(s))Z(x)=\sum_{(r,s)}\mathrm{i}^{J(r,s)}\exp(x\omega(s)), where the sum runs over paired states.

What Is Checked

The definitions are stated over triangle systems, states onto a fixed vertex subset, and ordered pairs of pointwise distinct states, with a finite triangle index set for the cyclic sum and polynomial weights. The adjacency relation in (6) is symmetric and irreflexive.