Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 6.4 — reduction from query-oblivious PCPs to nearest codeword with preprocessing

Proved
LatticePrepro.ObliviousPCP.pcp_reduction

by WizardOfMenlo · Nov 16, 2026 · Mathlib 0df444a (Lean v4.33.1)

coding-theorycvp-ncp-preprocessinghardness-of-approximationlatticespreprocessing

Let a query-oblivious PCP with enumerable predicate families (Definition 6.1) be given, with alphabet Σ\SigmaΣ, R\mathsf RR random bits, q≥1\mathsf q\ge1q≥1 queries and m\mathsf mm accepting views, and let LLL be a language for which it has perfect completeness and soundness error s>0\mathsf s>0s>0. Let FFF be a field, let ι ⁣:Fm→FB\iota\colon F^{\mathsf m}\to F^{B}ι:Fm→FB be an injective FFF-linear map such that every nonzero vector of the code ι(Fm)\iota(F^{\mathsf m})ι(Fm) has Hamming weight at least δ⋅B\delta\cdot Bδ⋅B, and let Γ≥0\Gamma\ge0Γ≥0 be an integer with

Γ<δ⋅s−1/q.\Gamma<\delta\cdot\mathsf s^{-1/\mathsf q}.Γ<δ⋅s−1/q.

Let CCC and yxy_xyx​ be the code and the targets of Construction 6.2, and let Y=B⋅2R⋅qY=B\cdot2^{\mathsf R}\cdot\mathsf qY=B⋅2R⋅q. Then:

  1. Completeness. For every x∈Lx\in Lx∈L, Δ(yx,C)≤Y\Delta(y_x,C)\le YΔ(yx​,C)≤Y.
  2. Soundness. For every x∉Lx\notin Lx∈/L, Δ(yx,C)>Γ⋅Y\Delta(y_x,C)>\Gamma\cdot YΔ(yx​,C)>Γ⋅Y.
  3. Size. The code has block length M=(∣Σ∣+Γ)⋅Y+1M=(|\Sigma|+\Gamma)\cdot Y+1M=(∣Σ∣+Γ)⋅Y+1.

The code CCC depends only on the PCP, the encoding map and Γ\GammaΓ, not on the instance. So the map x↦yxx\mapsto y_xx↦yx​ sends the instances of LLL to targets within distance YYY of the fixed code CCC, and all other instances to targets at distance more than Γ⋅Y\Gamma\cdot YΓ⋅Y. This is the combinatorial content of the reduction from LLL to Γ\GammaΓ-gap nearest codeword with preprocessing over FFF.

Formalization Note. This is Theorem 6.4 without its running-time claims. The final sentence of the theorem, that the construction is a preprocessing reduction when the parameters are polynomial in the instance length, is therefore not formalized. The hypotheses are weaker than in the paper: s<1\mathsf s<1s<1 and Γ≥2\Gamma\ge2Γ≥2 are not assumed, δ\deltaδ is only a lower bound on the relative distance of the code ι(Fm)\iota(F^{\mathsf m})ι(Fm), and FFF need not be finite. Instances form an arbitrary type. The bound m≤2R⋅k⋅∣Σ∣q\mathsf m\le2^{\mathsf R}\cdot\mathsf k\cdot|\Sigma|^{\mathsf q}m≤2R⋅k⋅∣Σ∣q from the proof is the separate theorem card_views_le. Δ(y,C)\Delta(y,C)Δ(y,C) is CodeFamily.codeDist, and the exponent is a real power, ε ^ (-(1 : ℝ) / q). The Lean names ε, E, g, ρ, q stand for s\mathsf ss, ι\iotaι, Γ\GammaΓ, R\mathsf RR, q\mathsf qq.

Preamble
import Mathlib
import Definitions.Def_LatticePrepro_Basic_Blocks
import Definitions.Def_LatticePrepro_Reduction_LinearSystem
import Definitions.Def_LatticePrepro_Conversions_Codes
import Definitions.Def_LatticePrepro_PCP_Verifier
import Definitions.Def_LatticePrepro_PCP_Code

open LatticePrepro
open LatticePrepro.ObliviousPCP
variable {X : Type} (P : ObliviousPCP X) {F : Type} [Field F] [DecidableEq F] {B : ℕ}
Formal statement
theorem LatticePrepro.ObliviousPCP.pcp_reduction {Lang : Set X} {ε δ : ℝ} (hc : P.PerfectlyComplete Lang)
    (hs : P.SoundnessError Lang ε) (hε : 0 < ε) {E : (P.Views → F) →ₗ[F] (Fin B → F)}
    (hE : Function.Injective E)
    (hδ : ∀ c ∈ LinearMap.range E, c ≠ 0 → δ * B ≤ hammingNorm c) {g : ℕ}
    (hg : (g : ℝ) < δ * ε ^ (-(1 : ℝ) / P.q)) :
    (∀ x ∈ Lang, CodeFamily.codeDist (P.target E g x) (P.code E g) ≤ P.radius B) ∧
    (∀ x, x ∉ Lang → g * P.radius B < CodeFamily.codeDist (P.target E g x) (P.code E g)) ∧
    Fintype.card (P.Coord B ⊕ Fin (g * P.radius B + 1)) =
      (Fintype.card P.Alphabet + g) * P.radius B + 1 := by sorry
Source
G. Fenzi, G. Weissenberg, Nearly Optimal Hardness for CVP and NCP with Preprocessing (STOC 2027 submission #847; IACR ePrint 2026/2634, https://eprint.iacr.org/2026/2634), Theorem 6.4 (\label{thm:pcp-reduction}); Lean: LatticePrepro/PCP/Reduction.lean, LatticePrepro.ObliviousPCP.pcp_reduction

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