§2: at most profiles per list; each original triple lies in exactly one subinstance
ProvedBddExp3SUM.profile_count3sumabelian-groupsp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1randomized-algorithms
Let have size at most . Then each list has at most
distinct profile vectors , and every triple belongs to exactly one subinstance, that is, to exactly one triple of profile classes.
The count bounds the number of subinstances the algorithm has to process, and the partition is what makes it enough to find a witness in one subinstance.
Differences from the source:
- generalization: the standing assumptions of p.2 (n ≥ 16, R ≥ M, |S_i| > √n) → only |S_i| ≤ n; reason: the two claims do not use the others; the source’s setting is a special case
Preamble
import Mathlib import Definitions.Def_BddExp3SUM_Profiles
Formal statement
namespace BddExp3SUM
/-- §2, p.2 (Zeng, Wang, Cao, Dai, Wang, STOC 2027 submission, manuscript sha256 19546c9a…): "There
are at most `(1 + ⌊log₂ n⌋)^L` profiles per list." and "Every original triple belongs to exactly one
such subinstance."
**Formalization Note.** `S i` are finite subsets of `G = Z_M^r` of size at most `n`. Neither clause
needs the other standing assumptions of p.2, which are therefore not hypotheses here. -/
theorem profile_count {M r n : ℕ} (S : Fin 3 → Finset (Grp M r)) (hS : ∀ i, (S i).card ≤ n) :
(∀ i, profileCount (S i) ≤ (1 + Nat.log 2 n) ^ Lcard M) ∧
∀ x : Fin 3 → Grp M r, (∀ i, x i ∈ S i) →
∃! U : Fin 3 → Finset (Grp M r), IsProfileTriple S U ∧ ∀ i, x i ∈ U i := by sorry
end BddExp3SUM
Source
Zeng, Wang, Cao, Dai, Wang, Truly Subquadratic 3SUM over Every Bounded-Exponent Abelian Group, STOC 2027 submission (manuscript sha256 19546c9a…), §2, p.2 (PDF p.3)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.