Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Top-layer-free resilience (1): 64 ∣{u∉L(∅):256k σu(V)>N}∣≥n64\,|\{u\notin L(\varnothing):256k\,\sigma_u(V)>N\}|\ge n64∣{u∈/L(∅):256kσu​(V)>N}∣≥n for covering pairs (Erdős 624)

Proved
Erdos624.top_layer_free_resilience

by Ethan Cheung · Oct 24, 2026 · Mathlib 0df444a (Lean v4.33.1)

combinatoricserdos-problems

Throughout, UUU is a finite set of inputs, CCC a finite set of colours with ∣C∣=n=2k|C|=n=2^k∣C∣=n=2k, and g0,g1:P(U)→Cg_0,g_1:\mathcal P(U)\to Cg0​,g1​:P(U)→C are two arbitrary maps (a pair); write L(A)={g0(A),g1(A)}L(A)=\{g_0(A),g_1(A)\}L(A)={g0​(A),g1​(A)} for the list of colours at the input AAA. Here also ∣U∣=n|U|=n∣U∣=n. For a pool V⊆UV\subseteq UV⊆U with N=∣V∣N=|V|N=∣V∣ and a colour u∉L(∅)u\notin L(\varnothing)u∈/L(∅), the proper minimal witnesses of uuu in W⊆VW\subseteq VW⊆V are the sets A⊆WA\subseteq WA⊆W with 0<∣A∣<k0<|A|<k0<∣A∣<k, u∈L(A)u\in L(A)u∈L(A) and u∉L(B)u\notin L(B)u∈/L(B) for every B⊊AB\subsetneq AB⊊A; the family is affine-coherent if some w:U→F2w:U\to\mathbf F_2w:U→F2​ has ∑x∈Aw(x)=1\sum_{x\in A}w(x)=1∑x∈A​w(x)=1 on every member. Let σu(V)\sigma_u(V)σu​(V) be the least ∣R∣|R|∣R∣, R⊆VR\subseteq VR⊆V, such that the proper minimal witnesses of uuu in V∖RV\setminus RV∖R are affine-coherent.

Theorem. Let k≥3k\ge3k≥3, N≥228k4N\ge2^{28}k^4N≥228k4, and suppose the pair covers every kkk-subset of VVV: for each Y∈(Vk)Y\in\binom VkY∈(kV​) and each colour uuu there is A⊆YA\subseteq YA⊆Y with u∈L(A)u\in L(A)u∈L(A). Then there is a set HHH of colours with 64∣H∣≥n64|H|\ge n64∣H∣≥n such that every u∈Hu\in Hu∈H satisfies u∉L(∅)u\notin L(\varnothing)u∈/L(∅) and 256k σu(V)>N256k\,\sigma_u(V)>N256kσu​(V)>N, i.e. for every R⊆VR\subseteq VR⊆V with 256k∣R∣≤N256k|R|\le N256k∣R∣≤N the proper minimal witnesses of uuu in V∖RV\setminus RV∖R admit no F2\mathbf F_2F2​ certificate. Equivalently

64 ∣{u∉L(∅): 256k σu(V)>N}∣  ≥  n.(1)64\,\bigl|\{u\notin L(\varnothing):\ 256k\,\sigma_u(V)>N\}\bigr|\;\ge\; n.\qquad(1)64​{u∈/L(∅): 256kσu​(V)>N}​≥n.(1)

(For V=UV=UV=U the size hypothesis holds for all k≥64k\ge64k≥64.)

Context: Erdős Problem 624 asks whether H(n)−log⁡2n→∞H(n)-\log_2 n\to\inftyH(n)−log2​n→∞, where H(n)H(n)H(n) is the least mmm such that some f:P(X)→Xf:\mathcal P(X)\to Xf:P(X)→X with ∣X∣=n|X|=n∣X∣=n satisfies f[P(Y)]=Xf[\mathcal P(Y)]=Xf[P(Y)]=X for all ∣Y∣≥m|Y|\ge m∣Y∣≥m. By an earlier track-amplification result it suffices to find common holes for fixed numbers of tracks; this is an auxiliary finite inequality from Sol's Round 10 analysis of the two-track case. It does not settle Erdős Problem 624, PAIR or Step 17.

Preamble
import Mathlib
open Finset
Formal statement
namespace Erdos624

theorem top_layer_free_resilience {U : Type*} [Fintype U] [DecidableEq U]
    {C : Type*} [Fintype C] [DecidableEq C] (k : ℕ) (hk : 3 ≤ k)
    (hU : Fintype.card U = 2 ^ k) (hC : Fintype.card C = 2 ^ k) (g0 g1 : Finset U → C)
    (V : Finset U) (hV : 2 ^ 28 * k ^ 4 ≤ #V)
    (hcov : ∀ Y ∈ powersetCard k V, ∀ u : C, ∃ A ⊆ Y, u ∈ ({g0 A, g1 A} : Finset C)) :
    ∃ H : Finset C, 2 ^ k ≤ 64 * #H ∧ ∀ u ∈ H, u ∉ ({g0 ∅, g1 ∅} : Finset C) ∧
      ∀ R ⊆ V, 256 * k * #R ≤ #V → ¬ ∃ w : U → ZMod 2, ∀ A ⊆ V \ R, 0 < #A → #A < k →
        u ∈ ({g0 A, g1 A} : Finset C) → (∀ B ⊂ A, u ∉ ({g0 B, g1 B} : Finset C)) →
          ∑ x ∈ A, w x = 1 := by sorry

end Erdos624
Source
Erdős Problem 624 (https://www.erdosproblems.com/624); Round 10 reports (consolidated report §'parity transport', §'many holes one level below'; Step 17 report 'resilient relations below the top layer', eqs. (9)-(26)); derived from GPT 6.1 Sol's analysis, relayed by Ethan Cheung; proof reconstructed, hand-checked and Lean-verified by Grok Bot for Ethan Cheung (Oct 24 2026). Auxiliary result only; does NOT settle PAIR / Step 17 / #624.

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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me