λ(d,1) asymptotic limit 1/e hypercube

Conjecture on $\lambda(d,1)$ · arXiv:2410.20498

arXiv Informal medium confidence— first stated 2024-10-27

Authors: Noga Alon, Maria Axenovich, John Goldwasser

Status open high confidence

No follow-up paper has proven or disproven the conjecture that $\lambda(d,1)=(1+o(1))/e$. A March 2025 paper (arXiv:2503.03408) studies the same $\lambda(d,s)$ statistics framework using flag algebras, proving exact values for $(d,s)\in\{(3,2),(4,2),(4,4)\}$, but explicitly treats $\lambda(2,1)$ as an intriguing open problem without addressing the $d\to\infty$ asymptotic. An April 2026 preprint (arXiv:2604.13402) extends the framework to affine (non-axis-aligned) subspaces of $\mathbb{F}_2^n$, motivated by the source paper, but its results concern a different statistic $\lambda^*(d,s)$ and do not bear on the axis-aligned conjecture.

Cited literature (1)

partial Some exact values of the inducibility and statistics constants for hypercubes (2025)

unknown (authors not returned by abstract page fetch) · arXiv preprint · arXiv:2503.03408

Proves exact values $\lambda(3,2)=8/9$, $\lambda(4,2)=264/343$, $\lambda(4,4)=26/27$ via flag algebras, and explicitly identifies $\lambda(2,1)$ as an open problem, but does not address the $d\to\infty$ asymptotic of $\lambda(d,1)$.

Reviewer notes. The conjecture is recent (October 2024) and no resolution was found. arXiv:2503.03408 (March 2025) is the closest follow-up in the same research line: it proves exact values for other (d,s) pairs using flag algebras, cites Alon-Axenovich-Goldwasser as the initiating work, and mentions $\lambda(d,1)$ cases as open problems, confirming the conjecture remains unresolved as of early 2025. arXiv:2604.13402 (April 2026) is motivated by the source paper but works with affine subspaces and a different statistic $\lambda^*(d,s)$; for odd $s$ it proves $\lambda^*(d,s)\le 1/2$, which is structurally different from the axis-aligned setting.

Auto-reviewed 2026-05-14 with claude-sonnet-4-6 (web search enabled).

Informal. $\lambda(d,1)=(1+o(1))/e$ where the $o(1)$-term tends to $0$ as $d$ tends to infinity.

Context

The lower bound in Theorem 2, $(1-2^{-d})^{2^{d}-1}$, approaches $e^{-1}\approx 0.37$ as $d\to\infty$. The authors suspect this lower bound is asymptotically sharp, making $1/e$ the true limit of $\lambda(d,1)$.

Notes. Stated in the abstract with the words 'We suspect'; no labelled conjecture environment.

Source paper

On hypercube statistics
Noga Alon, Maria Axenovich, John Goldwasser · 2024-10-27
https://arxiv.org/abs/2410.20498