The cardinality of a set containing the pairwise sums of a fixed number of integers

Wouter van Doorn

For every n≥2, g₄(n)=3: any A⊆{1,…,2n} with at least n+3 elements contains all six pairwise sums of four distinct integers, and the constant cannot be lowered. The four integers need not be positive or members of A.

Sources for this record

What was already known

Choi, Erdős and Szemerédi proved only g₄(n)=O(1).

AI involvement AI-autonomous

Van Doorn explicitly says GPT-5.3 Instant autonomously came up with the proof of Theorem 3 during brainstorming. Aristotle then formalized it in Lean. Van Doorn organized the paper and supplied human exposition; Aristotle's other formalizations and improved h₄ bound are separate contributions.

The author's first-person disclosure explicitly says GPT-5.3 Instant autonomously found the assessed Theorem 3 proof.

Importance Modest

Sharpens the four-integer case to an exact constant.

Limitations

The broader problem, including exact g₅, remains open. Van Doorn’s reported Lean formalization covers the four-integer theorem, not the harder five-integer case.

Technical verification Formal (Lean 4 via Aristotle)

The author reports that each blue-checkmarked result, including Theorem 3, was formalized by Aristotle. The linked Lean file contains theorem g4 stating gFun 4 n = 3 for n≥2, with upper and lower lemmas and an axiom-print command. This review inspected the statement but did not build the file or audit all definitions.

Formal checking

Formal proof in ErdosProblem866.lean, with separate upper/lower lemmas and #print axioms g4

Checker
Lean 4 kernel through Aristotle, reported by Wouter van Doorn
Scope
ErdosProblem866.g4: gFun 4 n = 3 for all n≥2
Basis
Reported by the linked checker
Evidence source

Checking a formal statement does not establish that it matches the original problem.

Human understanding Full

Van Doorn gives a short human-readable parity lower bound and pigeonhole upper bound, including explicit four-integer formulas, explaining the proof mechanism.

Correction and publication history