Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 4.5, p. 159 — for p ≥ 2, J′(F) has optima, all binary; x(F) = 1 iff (3.3) holds; value 2n_p + 1 − x(F)

Proved
JeroslowMLP.Value.theorem_4_5

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

multilevel-programmingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1polynomial-hierarchyquantified-boolean-formulareduction

Let p≥2p\ge2p≥2, let FFF be a propositional formula over blocks X1,…,XpX_1,\dots,X_pX1​,…,Xp​ of sizes n1,…,npn_1,\dots,n_pn1​,…,np​, and let S=Sp+1S=S_{p+1}S=Sp+1​ be the set of optimal solutions of the (p+1)(p+1)(p+1)-level game J′(F)J'(F)J′(F). Then:

  1. S≠∅S\neq\emptysetS=∅;
  2. at every x∈Sx\in Sx∈S all atom variables and all x(G)x(G)x(G) are binary;
  3. at every x∈Sx\in Sx∈S, x(F)=1x(F)=1x(F)=1 if and only if the Σp\Sigma_pΣp​ sentence (3.3) (∃Xp)(∀Xp−1)⋯(Q1X1)[F=1](\exists X_p)(\forall X_{p-1})\cdots(Q_1X_1)[F=1](∃Xp​)(∀Xp−1​)⋯(Q1​X1​)[F=1] is true;
  4. the value of J′(F)J'(F)J′(F) (player ppp's criterion on SSS) is
Zp=2np+1−x(F)={2np,(3.3) true,2np+1,(3.3) false;Z_p=2n_p+1-x(F)=\begin{cases}2n_p,&\text{(3.3) true},\\2n_p+1,&\text{(3.3) false};\end{cases}Zp​=2np​+1−x(F)={2np​,2np​+1,​(3.3) true,(3.3) false;​
  1. if (3.3) is true, a vector v∈Rnpv\in\mathbb R^{n_p}v∈Rnp​ is player ppp's block in some optimal solution if and only if vvv is binary and
(Qp−1Xp−1)⋯(Q1X1) [F(X1,…,Xp−1,α)=1](4.17)(Q_{p-1}X_{p-1})\cdots(Q_1X_1)\,[F(X_1,\dots,X_{p-1},\alpha)=1] \qquad (4.17)(Qp−1​Xp−1​)⋯(Q1​X1​)[F(X1​,…,Xp−1​,α)=1](4.17)

holds, where α\alphaα is the truth valuation of vvv.

The paper states this as: "the (p+1)(p+1)(p+1) level game J′(F)J'(F)J′(F) has a binary value, and this value is zero iff F∈BpF\in B_pF∈Bp​. Furthermore, all optimal solutions are binary vectors." The theorem reduces truth of Σp\Sigma_pΣp​ sentences to the value of a linear multi-level program with fixed criteria, so computing such values is Σp\Sigma_pΣp​-hard.

Formalization Note The paper's "binary value, zero iff F∈BpF\in B_pF∈Bp​" describes the explicit term 1−x(F)1-x(F)1−x(F) of Zp=1−x(F)+2∑jP(xpj)Z_p=1-x(F)+2\sum_jP(x_{pj})Zp​=1−x(F)+2∑j​P(xpj​); the term 2∑jP(xpj)=2np2\sum_j P(x_{pj})=2n_p2∑j​P(xpj​)=2np​ is the same constant in every optimum (the proof of Lemma 4.4 computes Zk=2nkZ_k=2n_kZk​=2nk​ or 1+2nk1+2n_k1+2nk​). Items 3 and 4 state the value exactly. "All optimal solutions are binary vectors" is read as the atom variables and the x(G)x(G)x(G): the gadget variable zzz of (4.1) equals 222 at y=1,ξ=0y=1,\xi=0y=1,ξ=0. Item 5 is the theorem's last sentence ("the additional fact emerged in the proof of Lemma 4.4"). Player ppp is index p of the game and its block is ⟨p - 1, _⟩.

Preamble
import Mathlib
import Definitions.Def_JeroslowMLP_Value_Game
Formal statement
namespace JeroslowMLP.Value

open MultilevelProgram GVar

theorem theorem_4_5 (p : ℕ) (hp : 2 ≤ p) (n : Fin p → ℕ) (F : Formula (Atom p n)) :
    let S := (jGame p n F).solSet (p + 1)
    let top : Fin p := ⟨p - 1, by omega⟩
    S.Nonempty ∧
    (∀ x ∈ S, (∀ a, IsBinary (x (atom a))) ∧ (∀ g, IsBinary (x (node g)))) ∧
    (∀ x ∈ S, (xF x = 1 ↔ Sentence33 p n F)) ∧
    (Sentence33 p n F → (jGame p n F).HasValue (p + 1) (2 * (n top : ℝ))) ∧
    (¬ Sentence33 p n F → (jGame p n F).HasValue (p + 1) (2 * (n top : ℝ) + 1)) ∧
    (Sentence33 p n F → ∀ w : Fin (n top) → ℝ,
      ((∃ x ∈ S, ∀ j, x (atom ⟨top, j⟩) = w j) ↔
        ((∀ j, IsBinary (w j)) ∧
          QHolds p n F (p - 1) (fun a => if h : a.1 = top then
            decide (w (Fin.cast (by rw [h]) a.2) = 1) else false)))) := by sorry

end JeroslowMLP.Value
Source
Jeroslow, The polynomial hierarchy and a simple model for competitive analysis, Math. Programming 32 (1985), p. 159, Theorem 4.5, (4.17)
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

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