Beyond the AI Proof Opaque Crisis: Re-solving via Modularization, N-Line Refactoring Rules, and Decompilation Governance for AI-Generated Mathematical Proofs
Recent AI-generated mathematical manuscripts have highlighted critical challenges regarding unverified theorems, bloated machine-level code, and human comprehension gaps. To maintain mathematicians' reasonable and target achievable working environment, to preserve humans' mathematical knowledge and research, and to maintain humans' civilization, this paper introduces a governance taxonomy distinguishing conjectures from claimed theorems, establishes refactoring and anti-bloat controls, and outlines verification strategies to prevent cognitive degradation in mathematical research.
Authors
- Weixing Shao (ORCID: https://orcid.org/0009-0008-9505-4840)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24248051
- Primary Topic
- Mathematics, Computing, and Information Processing
- Type
- preprint