1
姚班校友主导，Claude攻克费马大定理首个完整形式化证明
最后靠Harness救回来
2
面向数学形式化证明：Mistral AI 发布 Leanstral 1.5 低使用成本开源模型
数学证明成本骤降：Mistral AI开源Leanstral 1.5，解决同类问题仅需4美元，远超竞品。
IT之家 7 月 6 日消息，欧洲人工智能企业 Mistral AI 当地时间本月 2 日宣布推出 面向数学形式化证明程序语言 Lean 4 的 Leanstral 1.5 模型 。该模型总共拥有 119B 参数，激活 6B 参数，以 Apache-2.0 许可开源。 Mistral AI 表示，L…
3
PRA-RAG: Provably Robust Aggregation in Retrieval-Augmented Generation against Retrieval Corruption
面对检索投毒，PRA-RAG以可证明鲁棒的聚合机制守住生成安全，是RAG防御的新解法。
arXiv:2607.00012v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external…
4
A Machine-Verified Proof of a Quantum-Optimization Conjecture
机器验证攻克量子优化十年悬案：QAOA在环上近似比精确等于(2p+1)/(2p+2)，形式化证明无懈可击。
arXiv:2606.29687v1 Announce Type: cross Abstract: We report a machine-verified resolution of a problem open for over a decade in quantum optimization:…
5
Evaluating the Robustness of Proof Autoformalization in Lean 4
arXiv新作评测LLM将自然语言证明翻译成Lean 4形式代码的稳健性，揭示数学AI的前沿挑战。
arXiv:2606.14867v1 Announce Type: cross Abstract: Proof autoformalization aims to translate a mathematical informal proof written in natural language …
6
Formally Solving Answer-Construction Problems in Lean
聚焦Lean形式化证明新挑战：提出解决数学竞赛“答案构造”问题的形式化方法，结合LLM进展，拓展自动推理边界。
arXiv:2505.18492v5 Announce Type: replace Abstract: Mathematical competition problems fall into two broad types: theorem proving, which asks for a pro…
7
Advancing Mathematics Research with AI-Driven Formal Proof Search
大型语言模型在数学推理中进步巨大，但可靠性问题如何解决？这篇论文用130道IMO竞赛题验证了AI驱动形式化证明搜索的潜力。
arXiv:2605.22763v1 Announce Type: new Abstract: Large language models (LLMs) increasingly excel at mathematical reasoning, but their unreliability lim…
8
From LLM-Generated Conjectures to Lean Formalizations: Automated Polynomial Inequality Proving via Sum-of-Squares Certificates
提出用LLM生成猜想并转化为Lean形式化验证，基于平方和证书实现自动多项式不等式证明，开辟数学推理新路径。
arXiv:2605.15445v1 Announce Type: new Abstract: Automated proving of polynomial inequalities is a fundamental challenge in automated mathematical reas…