一篇给出 Transformer 代数系统的理论研究论文
An Algebra of Transformers
AI 导读
arXiv 论文《An Algebra of Transformers》为 Transformer 构建了一套用于推理的代数系统，基于 Transformer 的函数分解结果，以函数分解为原子动作，包含顺序组合、并行组合、递归反馈与抽象隐藏层等结构。作者称这是首次尝试对 Transformer 进行逻辑建模，可用于验证基于 Transformer 的应用的正确性。
正文 · 原文
Abstract:On the theoretic analyses on transforms, they mainly focus on the computability analysis, such the Turing-completeness of unbounded precision transformers, the advantage of parallelizability of unbounded precision transformers and the limitations of bounded precision transformers. This paper give transformers an algebra for reasoning, which relies heavily on the functional decomposition results of transformers. This algebra uses functional decompositions as atomic actions, includes sequential composition and parallel composition, recursion for feedback and abstraction for hidden layers. As we know, it is the first attempt to model transformers logically and can be used to verify the correctness of transformer based applications.
| Subjects: | Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.36795 [cs.LO] |
| (or arXiv:2610.36795v1 [cs.LO] for this version) | |
| https://doi.org/10.48550/arXiv.2610.36795 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yong Wang [view email]
[v1]
Tue, 27 Oct 2026 07:59:05 UTC (35 KB)
来源：arXiv cs.AI · arxiv.org