K₄ graph-code density vanishing

Informal question: is $d_{K_4}(n) = o(1)$? · arXiv:2301.13305

arXiv Informal medium confidence— first stated 2023-02-06

Status partial medium confidence

The question whether $d_{K_4}(n) = o(1)$ for general graph codes remains open as of May 2026. Versteegen (arXiv:2310.19891, 2023) made partial progress by proving that any linear graph code on $n$ vertices containing no copy of a graph $H$ with an even number of edges has size $O(2^{\binom{n}{2}}/\log n)$, which settles the analogue for linear codes; however, the general (non-linear) case of the question from Conjecture 1.1 / Question 1.1 of arXiv:2301.13305 is not resolved by this work. Several further papers cite arXiv:2301.13305 (including Gishboliner–Jin–Sudakov 2312.06610, Conlon–Lee–Versteegen 2404.17467, and Rancourt–Dodos–Tyros 2602.03298) but none were verified to resolve the full question.

Cited literature (1)

  • Leo Versteegen · arXiv preprint · arXiv:2310.19891

    Proves that a linear graph code on $n$ vertices with no copy of $H$ (even number of edges) has size $O(2^{\binom{n}{2}}/\log n)$, settling the linear-code variant of the $d_{K_4}(n)=o(1)$ question but leaving the general non-linear case open.

Reviewer notes. The Semantic Scholar citation list for arXiv:2301.13305 returned 20 citing papers (as of May 2026). The most directly relevant verified follow-up is arXiv:2310.19891 (Versteegen), which addresses only linear graph codes. The papers arXiv:2312.06610, arXiv:2404.17467, and arXiv:2602.03298 also cite the source paper but their specific results regarding $d_{K_4}(n)=o(1)$ were not verified within the 5-call budget. The lower bound $d_{K_4}(n) \geq 1/n^{o(1)}$ is established within the source paper itself via Hunter–Mubayi.

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

Informal. Is it true that $d_{K_4}(n) = o(1)$?

Context

In the concluding remarks the authors single this out as an interesting special case of Question 1.1. The companion question of whether $d_{K_4}(n) \geq 1/n^{o(1)}$ is answered affirmatively within the paper by appealing to an edge-coloring construction of Hunter and Mubayi [5] (modifying [11] and [4]); the upper-bound direction $d_{K_4}(n) = o(1)$ remains open.

Notes. The lower bound $d_{K_4}(n) \geq 1/n^{o(1)}$ is established in the paper; only the question of whether $d_{K_4}(n) = o(1)$ remains open.

Source paper

Graph-codes
Noga Alon · 2023-02-06
https://arxiv.org/abs/2301.13305 PDF source