---
id: 20261030-T0-03
title: "程序重构后LLM生成的证明还成立吗？新基准RefactorProof给出答案"
title_en: "RefactorProof Tests Whether LLM-Generated Proofs Survive Code Refactoring"
url: https://ai.daily.yangsir.net/daily/20261030-T0-03
issue_date: 2026-10-30
publish_date: 2026-10-29T04:00:00.000Z
category: research
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2610.36328
---

# 程序重构后LLM生成的证明还成立吗？新基准RefactorProof给出答案

可验证代码生成的基准测试通常只检验证明对一个固定实现是否有效。新论文提出RefactorProof，追问：当程序经过保持行为不变的重构后，同一份证明是否仍然成立？该基准衍生出458个经Lean认证的样本，用于压力测试LLM生成证明的鲁棒性。结论指向一个实际问题：如果模型生成的证明无法在重构后复用，形式化验证的工程价值会大打折扣。

## English Version

**RefactorProof Tests Whether LLM-Generated Proofs Survive Code Refactoring**

Benchmarks for verifiable code generation typically check a proof against one fixed implementation. A new paper introduces RefactorProof, asking whether the same proof remains valid after behavior-preserving program refactoring. The benchmark derives 458 Lean-certified samples to stress-test the robustness of LLM-generated proofs. The findings point to a practical problem: if generated proofs can't be reused after refactoring, the engineering value of formal verification drops sharply.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2610.36328)

**详情页**：https://ai.daily.yangsir.net/daily/20261030-T0-03

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*