OctoInformBeta

OpenAI preprint, September 2026 · Section 7. Comparison along square-step progressions

Section 7 (setting): induction parameters, fibers, and the functional: The residue lift functional and prime hypotheses used with a fixed ambient family

Parameters and fibers for scale comparison.

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

Comparison along square-step progressions uses a residue slice at the small primes and a conductor-truncated lift at the remaining prime coordinates. The ambient prime family stays fixed when the induction scale changes. For a natural length NN and a public integer set Apub⊆{1,…,N}A_{\mathrm{pub}}\subseteq\{1,\ldots,N\}, the natural-range translate is A={n∈{0,…,N−1}:n+1∈Apub}.A=\{n\in\{0,\ldots,N-1\}:n+1\in A_{\mathrm{pub}}\}.(1) The indicator 1A:ℕ→ℝ1_A:\mathbb N\to\mathbb R of the translate in (1) takes value one on AA and zero elsewhere.✦ For a natural length NN and a natural conductor bound QQ, the functional below is defined for every finite A⊆ℕA\subseteq\mathbb N, with the range restriction supplied by NN.

Let (qs)s∈S(q_s)_{s\in S} and (pj)j∈J(p_j)_{j\in J} be finite families of primes. The combined prime family p̃\widetilde p on S⊔JS\sqcup J of (qs)s∈S(q_s)_{s\in S} and (pj)j∈J(p_j)_{j\in J} is the family that agrees with qq on SS and pp on JJ.✦ The local quadratic modulus m2(q)m_2(q) is eight when q=2q=2 and is qq otherwise.✦ The fixed power exponent is β=1/100000\beta=1/100000.✦

The Residue Lift

The product residue space is Xp=∏j∈J𝔽pjX_p=\prod_{j\in J}\mathbb F_{p_j}.✦ The centered residue delta DqD_q on ℤ/qℤ\mathbb Z/q\mathbb Z and its product DUD_U on a support U⊆JU\subseteq J are Dq(x)=q1{0}(x)−1,DU(z)=∏j∈UDpj(zj).D_q(x)=q1_{\{0\}}(x)-1, \qquad D_U(z)=\prod_{j\in U}D_{p_j}(z_j).(2) Using the residue products in (2), the residue lift component and conductor-truncated lift of a real function ff on ℕ\mathbb N are LN,Upf(z)=1N∑0≤n<Nf(n)DU(z−(nmod⁡pj)j∈J),ℒN,Qpf(z)=∑U∈𝒰p(Q)LN,Upf(z),𝒰p(Q)={U⊆J:∏j∈Upj≤Q}.\begin{aligned} L_{N,U}^p f(z) &=\frac1N\sum_{0\le n<N}f(n)D_U\bigl(z-(n\bmod p_j)_{j\in J}\bigr),\\ \mathcal L_{N,Q}^p f(z) &=\sum_{U\in\mathcal U_p(Q)}L_{N,U}^p f(z), &\mathcal U_p(Q)&=\left\{U\subseteq J:\prod_{j\in U}p_j\le Q\right\}. \end{aligned}(3) Here N−1=0N^{-1}=0 at N=0N=0.

The Tuple Laws

For a positive integer hh, a permutation word on a block set is an assignment of a permutation of hh positions to each block.✦ Fix h=24h=24, T=80T=80, m=2T+1m=2T+1, r=h!/2r=h!/2, and δ=1/(10h2)\delta=1/(10h^2). The tuple vertex set is 𝒱=(Sym⁡{0,…,h−1}){0,…,m−1}\mathcal V=(\operatorname{Sym}\{0,\ldots,h-1\})^{\{0,\ldots,m-1\}}.✦ For an active block subset K⊆{0,…,m−1}K\subseteq\{0,\ldots,m-1\}, an inactive-block address a∈𝒜Ka\in\mathcal A_K is a permutation word on {0,…,m−1}\K\{0,\ldots,m-1\}\setminus K, and the addressed list indices form IK={0,…,r|K|−1}I_K=\{0,\ldots,r^{|K|}-1\}.✦ An addressed list entry is a pair in 𝒜K×IK\mathcal A_K\times I_K, where 𝒜K\mathcal A_K is the address set.✦ For a tuple vertex v∈𝒱v\in\mathcal V, the projection πK(v)∈𝒜K\pi_K(v)\in\mathcal A_K is the restriction of vv to the inactive blocks.✦ There are 161161 blocks.

For positions a<ba<b, the position transposition vk;a,bv^{k;a,b} replaces the permutation at block kk by its composition on the right with (ab)(a\ b).✦ The transposition interaction is the real matrix HK(v,w)=∑k∈K∑0≤a<b<hδb−a1{w=vk;a,b}.H_K(v,w)=\sum_{k\in K}\sum_{0\le a<b<h} \delta^{b-a}1_{\{w=v^{k;a,b}\}}.(4)✦ The cyclic shift cc of {0,…,h−1}\{0,\ldots,h-1\} is the permutation that sends each position to the next, returning the last to zero. Threshold shift adjacency of words u,wu,w means that w(k)=u(k)cw(k)=u(k)c on at least TT blocks.✦

For a real matrix HH indexed by a finite vertex set and labelings j,kj,k of that vertex set by a common label set, the label coincidence pairing is ∑v,wH(v,w)1{j(v)=k(w)}\sum_{v,w}H(v,w)1_{\{j(v)=k(w)\}}, where the sum runs over all ordered pairs of vertices.✦ The index energy of a real matrix HH and a labeling jj is half the label coincidence pairing of HH with jj in both arguments; the index Boltzmann weight of HH and jj is the exponential of that energy. The tuple Boltzmann weight for the interaction in (4) is WK(j)=exp⁡(12∑v,w∈𝒱HK(v,w)1{j(v)=j(w)}).W_K(j)=\exp\left(\frac12\sum_{v,w\in\mathcal V} H_K(v,w)1_{\{j(v)=j(w)\}}\right).(5)✦

Let EKE_K be the list constraint edge set on 𝒜K×IK\mathcal A_K\times I_K, specifying the square-difference constraints between list values. A strict square in an odd prime field is a nonzero square; write sps_p for its indicator.✦ For list values Y:𝒜K×IK→𝔽pY:\mathcal A_K\times I_K\to\mathbb F_p and a labeling j:𝒱→IKj:\mathcal V\to I_K, the square-constraint indicator and sampled tuple configuration are VK(Y)=∏(a,b)∈EKsp(Y(b)−Y(a)),zK,Y,j(v)=Y(πK(v),j(v)).V_K(Y)=\prod_{(a,b)\in E_K}s_p\bigl(Y(b)-Y(a)\bigr), \qquad z_{K,Y,j}(v)=Y\bigl(\pi_K(v),j(v)\bigr).(6) The same indicator sps_p means the nonzero square indicator when p=2p=2. Using the weight in (5) and the constraint indicator and sample in (6), the tuple component functional on real observables GG is the uniform probability average Cp,K(G)=𝔼Y𝔼jVK(Y)WK(j)G(zK,Y,j).C_{p,K}(G)=\mathbb E_Y\mathbb E_j V_K(Y)W_K(j)G(z_{K,Y,j}).(7)

The block subset mixture weight is p−|K|/16p^{-|K|/16}.✦ The restricted mixture over a family 𝒟\mathcal D of block subsets combines the components in (7); the full unnormalized mixture retains at most T+1T+1 active blocks: Bp,𝒟=∑K∈𝒟p−|K|/16Cp,K,Bp=Bp,{K:|K|≤T+1}.B_{p,\mathcal D}=\sum_{K\in\mathcal D}p^{-|K|/16}C_{p,K}, \qquad B_p=B_{p,\{K:|K|\le T+1\}}.(8) The terminal family is 𝒯={K:|K|=T+1}\mathcal T=\{K:|K|=T+1\}.✦ The full tuple law μp\mu_p, exceptional tuple sublaw εp\varepsilon_p, and its mass ηp\eta_p are obtained from the mixtures in (8) by μp=Bp(1)−1Bp,εp=Bp(1)−1Bp,𝒯,ηp=εp(1).\mu_p=B_p(1)^{-1}B_p, \qquad \varepsilon_p=B_p(1)^{-1}B_{p,\mathcal T}, \qquad \eta_p=\varepsilon_p(1).(9)

For a finite configuration space, the point mass of a functional LL at zz is L(1{z})L(1_{\{z\}}).✦ For a real function FF on Xp𝒱X_p^{\mathcal V}, the tensor product law Λp=⨂j∈Jμpj\Lambda_p=\bigotimes_{j\in J}\mu_{p_j} of the laws in (9) is the functional given by Λp(F)=∑(zj)j∈J(∏j∈Jμpj(1{zj}))F(v↦(zj(v))j∈J),zj∈𝔽pj𝒱.\Lambda_p(F)=\sum_{(z_j)_{j\in J}} \left(\prod_{j\in J}\mu_{p_j}(1_{\{z_j\}})\right) F\bigl(v\mapsto(z_j(v))_{j\in J}\bigr), \quad z_j\in\mathbb F_{p_j}^{\mathcal V}.(10) For real functions ava_v and gg on XpX_p, the multilinear integral for (10) and its diagonal specialization are IΛp((av)v)=Λp(z↦∏v∈𝒱av(z(v))),Zp(g)=IΛp((g)v∈𝒱).I_{\Lambda_p}((a_v)_v)=\Lambda_p\left(z\mapsto \prod_{v\in\mathcal V}a_v(z(v))\right), \qquad Z_p(g)=I_{\Lambda_p}((g)_{v\in\mathcal V}).(11)

The Prime Thresholds

The leading list identification sends vv to (π⌀(v),0)(\pi_{\varnothing}(v),0), where zero is the unique empty-block list index.✦ The leading constraint graph E0E_0 is the pullback of E⌀E_{\varnothing} through this bijection✦; pulling an edge set through a bijection retains exactly the pairs whose image is an edge.✦ Put e=|E⌀|e=|E_{\varnothing}|.✦ For a vertex vv of a finite graph EE, let mv(E)m_v(E) count the other vertices joined to vv in either orientation, the crossing edges for the decomposition into vv and the remaining vertices. The graph conditional energy constant is CE(v)=(2(1/2)mv(E)+4((1/2)mv(E))2)((32mv(E))1/4)2.C_E(v)=\left(\frac{2}{(1/2)^{m_v(E)}}+ \frac{4}{((1/2)^{m_v(E)})^2}\right) \left((32m_v(E))^{1/4}\right)^2.(12)✦

The leading mass lower bound ℓ\ell, Boltzmann weight upper bound WmaxW_{\max}, and conditional error numerator AcondA_{\mathrm{cond}} use the graph constants in (12); the conditional decay constant is CcondC_{\mathrm{cond}}: ℓ=2−(e+1),Wmax=exp⁡(|𝒱|(T+1)(h2)δ/2),Acond=∑v∈𝒱CE0(v)+2mWmax,Ccond=Acond/ℓ.\begin{aligned} \ell&=2^{-(e+1)},& W_{\max}&=\exp\left(|\mathcal V|(T+1)\binom h2\delta/2\right),\\ A_{\mathrm{cond}}&=\sum_{v\in\mathcal V}C_{E_0}(v)+2^mW_{\max},& C_{\mathrm{cond}}&=A_{\mathrm{cond}}/\ell. \end{aligned}(13) The interaction lower bound is κ=910δh−1\kappa=\frac9{10}\delta^{h-1}✦, and the half vertex count is M=(h!)m/2M= (h!)^m/2.✦ The marked and unmarked reflection constants are Cmarked=(exp⁡(κ)−1)M(rT+1)M,Cunmarked=(32|𝒱|2)1/4Wmax⁡2.C_{\mathrm{marked}}= \frac{(\exp(\kappa)-1)^M}{(r^{T+1})^M}, \qquad C_{\mathrm{unmarked}}=(32|\mathcal V|^2)^{1/4}W_{\max}^2.(14) The thresholds associated with (13) and (14) are the mass positivity threshold PmassP_{\mathrm{mass}}, conditional decay threshold PcondP_{\mathrm{cond}}, and reflection positivity threshold PreflP_{\mathrm{refl}}: Pmass=max⁡{3,(2e+1e)2},Pcond=max⁡{Pmass,3,Ccond32},Prefl=max⁡{3,(CunmarkedrM/Cmarked)16}.\begin{aligned} P_{\mathrm{mass}}&=\max\{3,(2^{e+1}e)^2\},\\ P_{\mathrm{cond}}&=\max\{P_{\mathrm{mass}},3,C_{\mathrm{cond}}^{32}\},\\ P_{\mathrm{refl}}&=\max\{3, (C_{\mathrm{unmarked}}r^M/C_{\mathrm{marked}})^{16}\}. \end{aligned}(15)

The Definition

Definition 1. For a nonnegative integer dd, the induction exponents are γ=164.\gamma=\frac1{64}.(16)✦ τd=βγ1000(d+1).\tau_d=\frac{\beta\gamma}{1000(d+1)}.(17)✦ σd=γτd4.\sigma_d=\frac{\gamma\tau_d}{4}.(18)✦ For a natural length NN and real exponent aa, the power cutoff is Ka(N)=⌊Na⌋ℕ,QN=Kβ(N),HN=Kβ/2(N),K_a(N)=\lfloor N^a\rfloor_{\mathbb N}, \qquad Q_N=K_\beta(N),\qquad H_N=K_{\beta/2}(N),(19)✦ where the natural floor is zero for a negative argument. For the small prime family qq, the combined small modulus is Mq=∏s∈Sm2(qs).M_q=\prod_{s\in S}m_2(q_s).(20)✦ For a natural modulus M′M', a residue s∈ℤ/M′ℤs\in\mathbb Z/M'\mathbb Z, and a finite set A⊆ℕA\subseteq\mathbb N, the normalized residue slice fM′,s,A:ℕ→ℝf_{M',s,A}:\mathbb N\to\mathbb R is the function given by fM′,s,A(n)=M′1{n≡s(mod⁡M′)}1A(n).f_{M',s,A}(n)=M'1_{\{n\equiv s\pmod{M'}\}}1_A(n).(21)✦ For natural N,QN,Q and a finite A⊆ℕA\subseteq\mathbb N, the set diagonal integral, using (11), is Yp(N,Q;A)=Zp(ℒN,Qp1A).Y_p(N,Q;A)=Z_p\bigl(\mathcal L_{N,Q}^p1_A\bigr).(22)✦ The source prime hypotheses at coverage scale NN, with thresholds from (15), are p̃:S⊔J→ℕ is injective,pj≥64,pj≥max⁡{Pmass,Pcond},pj≥Prefl,gcd⁡(Mq,pj)=1,ηpj≤pj−4(j∈J),{r:r prime,r≤N}⊆{p̃i:i∈S⊔J}.\begin{gathered} \widetilde p:S\sqcup J\to\mathbb N\text{ is injective},\\ p_j\ge64,\quad p_j\ge\max\{P_{\mathrm{mass}},P_{\mathrm{cond}}\},\quad p_j\ge P_{\mathrm{refl}},\quad \gcd(M_q,p_j)=1,\quad \eta_{p_j}\le p_j^{-4} \qquad(j\in J),\\ \{r:r\text{ prime},\ r\le N\} \subseteq\{\widetilde p_i:i\in S\sqcup J\}. \end{gathered}(23)✦ For natural bounds P,NP,N, the large prime set, its index set, and the large prime family are 𝒫(P,N)={r prime:P≤r≤N},JP,N=𝒫(P,N),pP,N:JP,N→ℕ,pP,N(r)=r.\mathcal P(P,N)=\{r\text{ prime}:P\le r\le N\}, \qquad J_{P,N}=\mathcal P(P,N),\qquad p_{P,N}:J_{P,N}\to\mathbb N,\quad p_{P,N}(r)=r.(24)✦

In the paper, the result reads: choose a fixed threshold PP such that ∏p≥P(1+p−3)−1≤1/4\prod_{p\ge P}(1+p^{-3})-1\le1/4 and |𝔤(ξ)|≤q(ξ)−1/3|\mathfrak g(\xi)|\le q(\xi)^{-1/3} for squarefree denominators q(ξ)q(\xi) supported on primes at least PP, put M0=8∏3≤p<PpM_0=8\prod_{3\le p<P}p, and, for N≥1N\ge1, set Q=NβQ=N^\beta, H=Nβ/2H=N^{\beta/2}, 𝒫N={p prime:P≤p≤Q}\mathcal P_N=\{p\text{ prime}:P\le p\le Q\}, γ=1/64\gamma=1/64, τd=βγ/(1000(d+1))\tau_d=\beta\gamma/(1000(d+1)), and σ0=γτd/4\sigma_0=\gamma\tau_d/4, define Y(N,Apub)=Z𝒫N(ℒ[N],𝒫N,Q1Apub)Y(N,A_{\mathrm{pub}})=Z_{\mathcal P_N} (\mathcal L_{[N],\mathcal P_N,Q}1_{A_{\mathrm{pub}}}) for square-difference-free Apub⊆[N]={1,…,N}A_{\mathrm{pub}}\subseteq[N]=\{1,\ldots,N\} and fs(n)=M01{n≡s(mod⁡M0)}1Apub(n)f_s(n)=M_0 1_{\{n\equiv s\pmod{M_0}\}}1_{A_{\mathrm{pub}}}(n), and let a fiber fix a set BB of prime coordinates and a small residue ss.

What Is Checked

The exponents (16), (17), and (18), cutoffs (19), modulus (20), slices (21), functional (22), hypotheses (23), and family (24) are stated over finite prime families, natural length and conductor parameters, finite subsets of ℕ\mathbb N, and real observables on finite tuple configuration spaces. The tuple component functionals, restricted and full mixture functionals, full tuple laws, exceptional tuple sublaws, and tensor product law preserve addition and multiplication by real scalars in their observables.