Work

Whole-Cofactor Peeling, Extreme-Point LP Rank Rounding, and Shifted-Interval Hinge Duality for Erdős Problem 708

Year: 2026
Type: preprint
Abstract: Let 𝐴 be a set of 𝑛 integers with maximum 𝑀. Let 𝑔(𝑛) be the smallest integer such that every interval of length 𝑀 contains a subset 𝐵 of size at most 𝑔(𝑛) whose product is a multiple of the product of all elements in 𝐴. Erdős and Surányi proved that 𝑔(𝑛) is at least (2 − 𝑜(1))𝑛 in 1959. Erdős later posed the question of whether 𝑔(𝑛) ≤ (2 + 𝑜(1))𝑛. Previous methods, such as those by Chen, established bounds of 𝑔(𝑛) ≤ 11𝑛, but these methods struggle in the intermediate interval regime. We resolve this by combining extreme-point linear-programming rounding and shifted-interval hinge inequalities. We establish an 56501 unconditional bound of 𝑔(𝑛) ≤ ( + 𝑜(1))𝑛 ≈ 8.83𝑛. In other words, our method reduces the upper 6400 bound by approximately 20% compared to previous work. Finally, we verify our bounds using sliding-window evaluations up to 𝑀 = 15,120.This paper establishes the asymptotic bounds for the Erdős–Surányi function g(n). For exact small-n values and computational verification details, see the companion manuscript: 10.5281/zenodo.24366897
Language: English
Open Access status: green
OpenAlex ID: W7236090787
Cites:
Cited by: 1
Related to: 10