Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

The tuned normalized physical probe with Gram scales ZℓxZ^{\ell_x}Zℓx​, ZℓyZ^{\ell_y}Zℓy​

Definition
Transport_PhysicalLow_NormalizedProbe

by Himan · Nov 15, 2026 · Mathlib 0df444a (Lean v4.33.1)

analytic-number-theorynumber-theoryqrh-lw56e2-b83dbbdtowards-grh

Throughout, O=Z[ζ3]\mathcal O=\mathbb Z[\zeta_3]O=Z[ζ3​], a Hecke character is a finite-order Hecke character of Q(ζ3)\mathbb Q(\zeta_3)Q(ζ3​) in OpenAI's family-003 development, and β\betaβ is the supremum of the real parts of the zeros of the associated Hecke LLL-functions (pole exception excluded, sentinel at 1/21/21/2). The tuned constants of LW56e2's refinement are p=214651/1250000000p=214651/1250000000p=214651/1250000000, the Gram exponent b=192821/1562500b=192821/1562500b=192821/1562500, the total slot length ℓ=16+p\ell=\tfrac16+pℓ=61​+p, the row and column exponents ℓx=(1−ℓ−b)/2\ell_x=(1-\ell-b)/2ℓx​=(1−ℓ−b)/2 and ℓy=(1−ℓ+b)/2\ell_y=(1-\ell+b)/2ℓy​=(1−ℓ+b)/2, the conductor exponent h=(1+3ℓ+b)/2h=(1+3\ell+b)/2h=(1+3ℓ+b)/2, the cushion ζ=1/(8⋅1012)\zeta=1/(8\cdot 10^{12})ζ=1/(8⋅1012), the boundary B=1112−ℓ4=4374785349/5000000000B=\tfrac{11}{12}-\tfrac{\ell}{4}=4374785349/5000000000B=1211​−4ℓ​=4374785349/5000000000 and the signal exponent sig⁡(s)=s+ℓx/2−1+h/6=s−2147607/3125000\operatorname{sig}(s)=s+\ell_x/2-1+h/6=s-2147607/3125000sig(s)=s+ℓx​/2−1+h/6=s−2147607/3125000.

This file defines the probe function of the tuned argument. It is OpenAI's normalized physical probe with the three numerical exponents of the original construction replaced by the tuned ones.

Let M⊂OM\subset\mathcal OM⊂O be a nonzero ideal, HHH a subgroup of (O/M)×(\mathcal O/M)^\times(O/M)×, SSS a finite set of maximal ideals with nonzero product, ℓ1,…,ℓK\ell_1,\dots,\ell_Kℓ1​,…,ℓK​ real slot lengths, 0<a≤b0<a\le b0<a≤b a support window, W1,…,WK:R→RW_1,\dots,W_K:\mathbb R\to\mathbb RW1​,…,WK​:R→R slot profiles, W0,W1W^0,W^1W0,W1 Schwartz functions and η\etaη a Hecke character. For Z>0Z>0Z>0 put Yj=ZℓjY_j=Z^{\ell_j}Yj​=Zℓj​, and let TjT_jTj​ be the pool of prime ideals outside SSS in the identity ray class modulo (M,H)(M,H)(M,H) with norm in the window [aYj,bYj][aY_j,bY_j][aYj​,bYj​]. The tuned normalized probe is

J(Z)=Pη(Zℓx, Zℓy, Z)c(W0,W1,S)⋅Π(Z,ℓ,(mj)j),J(Z)=\frac{\mathcal P_\eta\big(Z^{\ell_x},\,Z^{\ell_y},\,Z\big)}{c(W^0,W^1,S)\cdot \Pi\big(Z,\ell,(m_j)_j\big)},J(Z)=c(W0,W1,S)⋅Π(Z,ℓ,(mj​)j​)Pη​(Zℓx​,Zℓy​,Z)​,

where Pη(X,Y,Z)\mathcal P_\eta(X,Y,Z)Pη​(X,Y,Z) is the source's compensated physical probe of η\etaη built from the calibration data of SSS, the canonical slot supports of the pools TjT_jTj​, the slot profiles WjW_jWj​ at scales YjY_jYj​, and the Gram scales X,YX,YX,Y; c(W0,W1,S)c(W^0,W^1,S)c(W0,W1,S) is the source residue constant for the modulus ∏P∈SP\prod_{P\in S}P∏P∈S​P; and Π(Z,ℓ,m)\Pi(Z,\ell,m)Π(Z,ℓ,m) is the principal scalar over all KKK slots with total length ℓ\ellℓ and slot masses mjm_jmj​ computed from the pools and the residue weights of the WjW_jWj​.

OpenAI's original normalizedProbe is the same expression with (ℓx,ℓy,ℓ)=(17/48, 23/48, 1/6)(\ell_x,\ell_y,\ell)=(17/48,\,23/48,\,1/6)(ℓx​,ℓy​,ℓ)=(17/48,23/48,1/6). The low-side and the high-side estimates of the refinement are both statements about this one function, which is why it is isolated as a definition.

Formalization Note Transcribed verbatim from the source. compensatedPhysicalProbe, calibrationForSet, canonicalSlotSupport, pool, sourceResidueConstant, Probe.principalScalar, slotMass and residueWeights are the platform's port of OpenAI's objects (bundles up to OAIHecke78B022). The value at ZZZ is a quotient in C\mathbb CC, with Lean's convention x/0=0x/0=0x/0=0.

Definition code
import Mathlib
import Definitions.Def_HeckeSevenEighths
import Definitions.Def_OAIHecke78B022
import Definitions.Def_Transport_TunedCandidate

set_option autoImplicit false

/-! The actual tuned normalized probe.
Adapted from OpenAI Detector/FinalAssemblySource.lean, pinned source commit
fd4aeeb2ee4fc729c18d98444fed42fd0529eeeb.
This definition is shared verbatim by the low, high and principal proofs. -/

namespace RH.Transport.PhysicalLow

noncomputable section
open scoped Classical ContDiff
open Filter Asymptotics
open OAI OAI.SevenEighths
open HeckeFamily ProbePhysical ProbeRaySlots PrincipalSignalComparison PrincipalMellinResidues
open ProbePrincipalResidueActual
local notation "O" => HeckeFamily.O

def tunedNormalizedProbe {K : ℕ} (M : Ideal O) [NeZero M] [Finite (O ⧸ M)]
    (H : Subgroup (O ⧸ M)ˣ) (S : Finset (Ideal O)) (hmax : ∀P∈S,P.IsMaximal)
    [NeZero (∏P∈S,P)] (ell : Fin K→ℝ) (a b : ℝ) (W : Fin K→ℝ→ℝ)
    (W0 W1 : SchwartzMap ℝ ℂ) (η : Character) (Z : ℝ) : ℂ :=
  let Yp := fun j=>Z^(ell j)
  let T := fun j=>pool (RayQuotient.identityClass M H) S a b (Yp j)
  compensatedPhysicalProbe η (calibrationForSet S hmax) W0 W1
    (fun j=>canonicalSlotSupport (T j)) (fun j y=>(W j y:ℂ)) Yp
    (Z^TunedCandidate.lx) (Z^TunedCandidate.ly) Z /
    (sourceResidueConstant W0 W1 (∏P∈S,P)*
      (Probe.principalScalar Finset.univ Z TunedCandidate.ell (slotMass T (residueWeights W Yp)) : ℂ))

end
end RH.Transport.PhysicalLow
Source
LW56e2, quasi-riemann-improved, commit b83dbbd914c5751acf9bc3389cde58970437278f, lean/Transport/PhysicalLow/NormalizedProbe.lean, def tunedNormalizedProbe (L18-L28), adapted from OpenAI math, Detector/FinalAssemblySource.lean (normalizedProbe): https://github.com/LW56e2/quasi-riemann-improved/blob/b83dbbd914c5751acf9bc3389cde58970437278f/lean/Transport/PhysicalLow/NormalizedProbe.lean#L18-L28

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me