How to check an AI-generated math proof that claims to pass Lean
What happened
On November 16, Hacker News' front page discussed how to understand the claim that a theorem is correct because it passes a Lean check. The article says a Lean check gives important evidence for a theorem's correctness, but for AI-generated code you still need to verify that the code's propositions match the paper's theorems, along with the compilation result, the axioms used and the reliability of the kernel. It describes using lake comparator to check proposition consistency and axioms, and lake comparator --paranoid for multi-kernel review. Of the 290 OpenAI papers that ship Lean code, about 38 have code that does not state the paper's main theorem.
Written by AI from the coverage · updated 8 hours ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageInterpreting claims that a theorem is correct because it was checked in Lean
Lean 检查能为定理正确性提供重要依据,但 AI 生成的代码还需核验命题对应关系、编译结果、所用公理及内核可靠性。文章介绍了用 lake comparator 检查命题一致性与公理、用 lake comparator --paranoid 进行多内核复核的方法,并指出 OpenAI 附带 Lean 代码的 290 篇论文中,约 38 篇的代码未陈述论文的主要定理。
Heat over time
Heat now 6·Comparable peak 9(Nov 16 20:00)·Comparable change over 24 hours –
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.