Behrend-style sets avoiding mixed-sign affine patterns

Informal question on Behrend-style sets avoiding affine patterns with mixed-sign coefficients · arXiv:1903.04863

arXiv Informal medium confidence— first stated 2019-07-02

Authors: Jacob Fox, Ashwin Sah, Mehtaab Sawhney, David Stoner, Yufei Zhao

Status open low confidence

The question asks whether there exist subsets of $[N]$ with $N^{1-o(1)}$ elements avoiding linear patterns with two positive and two negative coefficients (e.g., $2x+2y=3w+z$), in the spirit of Behrend's construction. No follow-up paper resolving this question was found in a targeted web search covering the period 2019–2026. The question is related to known limitations of the Ruzsa–Bergelson–Host–Kra popular-difference approach, and Asaf Shapira's earlier work on Behrend-type constructions for linear equations treats adjacent problems, but the specific mixed-sign regime identified in the Fox–Sah–Sawhney–Stoner–Zhao paper appears to remain open. Given that the question is now roughly 7 years old, the absence of indexed results warrants low confidence.

Reviewer notes. No follow-up resolving this question was found after 5 web calls. The closest related work is Shapira's 'Behrend-Type Constructions for Sets of Linear Equations', which predates the Fox et al. paper and treats the problem for different coefficient regimes. Recent improvements to Behrend's construction (arXiv:2401.16106, arXiv:2406.12290) concern 3-AP-free sets and do not address the mixed-sign setting. The question remains open with low confidence due to the 7-year age of the conjecture.

Auto-reviewed 2026-05-15 with claude-sonnet-4-6 (web search enabled).

Informal. We do not know if there are subsets of $[N]$, in the style of Behrend's construction, with $N^{1-o(1)}$ elements avoiding such patterns (e.g., $2x + 2y = 3w + z$).

Context

In Section 3.1, the authors discuss why the Ruzsa–Bergelson–Host–Kra approach for constructing popular-difference counterexamples fails for certain linear identities $a_1P(b_1)+\cdots+a_4P(b_4)=0$ in which there are two positive and two negative $a_i$'s (rather than the symmetric case). They note that it is unknown whether large Behrend-style sets avoiding such patterns exist.

Notes. Stated in running prose in Section 3.1 without a labelled environment. PDF source.

Source paper

Triforce and Corners
Jacob Fox, Ashwin Sah, Mehtaab Sawhney, David Stoner, Yufei Zhao · 2019-07-02
https://arxiv.org/abs/1903.04863 PDF source