aiqbench analysis
This paper establishes a rigorous lower bound on the number of ways specific trees can be grown on a square lattice, extending Bouch's construction to all sizes via an AI-assisted proof. The research highlights implications for operator growth and imaginary-time evolution in quantum spin systems, directly linking combinatorial tree structures to fundamental dynamics in quantum physics.
Trees that can be grown in "too many" ways: Bouch's construction and a bound for every size
arXiv:2412.16912v4 Announce Type: replace-cross Abstract: We show that, for every positive integer $L$, there is a rooted tree on the square lattice that can be grown in at least $L!/C^L$ distinct ways, where $L$ denotes the number of bonds and $C>1$ is a constant independent of $L$. Bouch's hierarchical construction [Bouch2015] gives such trees along an unbounded sequence of sizes. We carefully review his construction and combine it with an elementary extraction lemma to obtain the bound for every size. (As discussed in Section~IV.A of [ParkerCaoAvdoshkinScaffidiAltman2019], Appendix~A.3 of [ShiraishiTasaki2024], and the video [TasakiVideo2025], the existence of such trees has an implication for operator growth and imaginary-time evolution in quantum spin systems in two or higher dimensions.) I wrote the first version of this note as a self-contained exposition of Bouch's construction. The stronger lower bound for every tree size presented here was proved by ChatGPT on the basis of that version (see the end of Section~1). I wish to make clear that I made no creative contribution to this beautiful and, I hope, useful theorem, whose proof is due to Bouch and ChatGPT.
Abstract:We show that, for every positive integer $L$, there is a rooted tree on the square lattice that can be grown in at least $L!/C^L$ distinct ways, where $L$ denotes the number of bonds and $C>1$ is a constant independent of $L$. Bouch's hierarchical construction [Bouch2015] gives such trees along an unbounded sequence of sizes. We carefully review his construction and combine it with an elementary extraction lemma to obtain the bound for every size. (As discussed in Section~IV.A of [ParkerCaoAvdoshkinScaffidiAltman2019], Appendix~A.3 of [ShiraishiTasaki2024], and the video [TasakiVideo2025], the existence of such trees has an implication for operator growth and imaginary-time evolution in quantum spin systems in two or higher dimensions.) I wrote the first version of this note as a self-contained exposition of Bouch's construction. The stronger lower bound for every tree size presented here was proved by ChatGPT on the basis of that version (see the end of Section~1). I wish to make clear that I made no creative contribution to this beautiful and, I hope, useful theorem, whose proof is due to Bouch and ChatGPT.
12 pages, 6 figures. Expository note not to be submitted to any journals. Versions 1-3: A self-contained exposition of Bouch's construction. Version 4: Title changed. A new proof obtained by ChatGPT extends Bouch's bound to every tree size. The author claims no creative contribution to the proof
Mathematical Physics (math-ph); Statistical Mechanics (cond-mat.stat-mech); Pattern Formation and Solitons (nlin.PS); Quantum Physics (quant-ph)
arXiv:2412.16912 [math-ph]
(or arXiv:2412.16912v4 [math-ph] for this version)
Full text under the publisher's open licence · 247 words
Related coverage

Filling in the convex hull
LessWrong · Oct 9, 2026

Social Theory and Agent Foundations
LessWrong · Oct 8, 2026
Untrusted advice for AI control: Short, strong advice significantly uplifts weak LLMs
LessWrong · Jul 27, 2026

Research directions in condensation: varieties of objectivity
AI Alignment Forum · Jul 28, 2026