paperbot · PL 论文追踪

全部论文 待读 收藏

P2 Simply typed convertibility is TOWER-complete even for safe lambda-terms

简单类型可转换性即使对安全 λ-项也是 TOWER-完全的
LMCS vol.Volume 20, Issue 32024匹配度 2/10 · 该论文研究简单类型 λ-项 β-可转换性在安全 λ-演算片段下的复杂度(TOWER-完全),属于纯类型论/复杂度与隐式自动机方向,与用户关注的 LLM/Agent 程序静态分析、浮点程序分析、贝叶斯程序分析均无直接交集,也未涉及工具实现或工业落地。引用 1 inbox
Lê Thành Dũng Nguyên
crossref: 2026-09-10

AI 速览

一句话:证明简单类型 λ-项的可转换性问题在 Blum-Ong 安全 λ-项片段上仍是 TOWER-完全。

本文研究两个简单类型 λ-项是否 β-可转换(等价于是否有相同范式)的判定问题。该问题已知是非初等的,精确复杂度为 TOWER-完全,但这一事实较少为人所知;本文也旨在普及该结果。作者的核心贡献是证明:当输入项属于 Blum 与 Ong 的安全 λ-演算——一个来自高阶递归方案研究的简单类型 λ-演算片段——时,β-可转换性仍保持 TOWER-完全。此前该安全片段问题的最好下界仅为 PSPACE-难。证明通过从 star-free 表达式等价问题进行归约,灵感来自作者与 Pradic 关于“typed λ-演算中的隐式自动机”的工作。结果同样适用于 βη-可转换性。

问题
判定两个简单类型 λ-项是否 β-可转换(即是否有相同范式),并确定该问题在一般情形及安全 λ-项片段下的精确复杂度。
方法
通过从 star-free 表达式等价问题归约,并借鉴 typed λ-演算中隐式自动机的思想,证明安全 λ-项上的 β-可转换性 TOWER-难;结合已有上界得到 TOWER-完全。
评估
理论复杂度分析:建立 TOWER-完全性,将安全 β-可转换性下界从 PSPACE-难提升至 TOWER-完全,并推广至 βη-可转换性。
原文摘要(Abstract)

We consider the following decision problem: given two simply typed $\lambda$-terms, are they $\beta$-convertible? Equivalently, do they have the same normal form? It is famously non-elementary, but the precise complexity - namely TOWER-complete - is lesser known. One goal of this short paper is to popularize this fact. Our original contribution is to show that the problem stays TOWER-complete when the two input terms belong to Blum and Ong's safe $\lambda$-calculus, a fragment of the simply typed $\lambda$-calculus arising from the study of higher-order recursion schemes. Previously, the best known lower bound for this safe $\beta$-convertibility problem was PSPACE-hardness. Our proof proceeds by reduction from the star-free expression equivalence problem, taking inspiration from the author's work with Pradic on "implicit automata in typed $\lambda$-calculi". These results also hold for $\beta\eta$-convertibility.

链接与引用

DOI 原文 · PDF(开放获取) ·

BibTeX
@article{paperbot2728,
  title = {Simply typed convertibility is TOWER-complete even for safe lambda-terms},
  author = {Lê Thành Dũng Nguyên},
  journal = {Logical Methods in Computer Science},
  volume = {Volume 20, Issue 3},
  year = {2024},
  doi = {10.46298/lmcs-20(3:21)2024}
}