← All papers
First page of Settling the Optimal Exponent Relating Sumsets and Difference Sets

Settling the Optimal Exponent Relating Sumsets and Difference Sets

Haowei Lin, Shanda Li

math.CO Jul 29, 2026 · v1
The main theorem's proof was converted by an LLM into a Lean 4 formal proof, which is released in a public GitHub repository.
For a finite nonempty subset $A$ of an abelian group, let $σ(A)=|A+A|/|A|$ and $δ(A)=|A-A|/|A|$. The classical sum-difference inequalities state that $$σ(A)^{1/2}\leqδ(A)\leqσ(A)^2.$$ The exponent $2$ in the second inequality is known to be optimal, whereas it has remained open whether the exponent $1/2$ in the first inequality can be improved. We settle this question by constructing an explicit family of finite sets $A_K\subset\mathbb{Z}$ such that $$\frac{\logσ(A_K)}{\logδ(A_K)}\longrightarrow 2,$$ hence the exponent $1/2$ in the first inequality is also optimal. The construction and its proof were developed with the assistance of Hyra, an AI research agent based on the open-weights Hy3 model.

For finite sets A of integers, the sum-difference inequalities state σ(A)^{1/2} ≤ δ(A) ≤ σ(A)^2. The exponent 2 in the second inequality was known to be optimal. It was open whether the exponent 1/2 in the first inequality could be improved, equivalently whether sup log σ(A)/log δ(A) equals 2.

An explicit family A_K ⊂ ℤ is built from four ingredients: a base-12 digit gadget W={0,1,2,4,5,9}, a three-state carry automaton that counts modular differences, a symmetric additive basis in a cyclic group, and a Chinese remainder construction. The construction was found with the AI research agent Hyra (Hy3 model), guided by an LLM judge. The authors checked the proof by hand, and GPT-5.6 Sol converted the natural-language proof into Lean 4.

For every positive even K, C(A_K) > 2K/(K+3), so C(A_K) → 2 and the exponent 1/2 is optimal. This exceeds prior constructions and agent-search results, including Penman–Wells at 1.1259 and SimpleTES at 1.1449. A Lean 4 formal proof is released.

MethodValue
Freiman–Pigarev (1973)1.0598
Penman–Wells family1.1259
AlphaEvolve1.1219
SimpleTES with post-training1.1449
Explicit family A_K (this work)sup = 2
Selected best values of C(A)