小型 Transformer 学会 Coxeter 群语言并催生数学猜想证明
Can small transformers learn the language of Coxeter groups?
数学家和 AI 研究者合训了个小 Transformer，让它学三角形铺砌对称群的语言，然后靠看模型注意力模式想出并证明了一个数学猜想。想做可解释性或数学交叉研究的人可以看看这个范式。
一篇 arXiv 论文（编号 2611.10936）展示了数学家与机制可解释性研究者的合作成果。研究团队训练了一个小型 Transformer 学习 Coxeter 群 affine A2 的语言——该群最广为人知的刻画是等边三角形对欧氏平面的铺砌对称性。
训练完成后，研究者观察模型的注意力模式，据此 formulate 并证明了一个关于 affine A2 的词理论（word-theoretic）猜想。也就是说，模型内部学到的表征直接启发了数学命题的提出与严格证明。
这项工作说明机制可解释性可以成为数学研究的工具：注意力权重不只是解释对象，还能反哺猜想生成。对研究可解释性与数学交叉方向的学者，这是少见的闭环案例。
Can small transformers learn the language of Coxeter groups?
The field of mechanistic interpretability is well-suited to collaborative projects between mathematicians and AI researchers. This work describes one such project. We trained a small transformer to learn the language of the Coxeter group affine A2, best known as the symmetries of the tiling of the Euclidean plane by equilateral triangles. We then formulated and proved a word-theoretic conjecture about affine A2 after observing the attention patterns of the trained model.
This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Transformer, 机制可解释性, Coxeter群, arXiv.
Source attribution: curated from arXiv cs.LG (@J. Elisenda Grigsby, Yiyao Huang, Arushi Katyal, Ryan Lin, Pranav Ramakrishnan, Weiya Wang).
