MATH · 数学阅读
数学前沿
80 条内容数学进展、研究者文章与数学 × AI。
The myths of scientific edition
从方程到海岸灾害评估:海啸的数学建模
文章以作者团队开发的 Tsunami-HySEA 模型为主线,说明如何从球坐标下的浅层水方程出发,用有限体积格式、TVD-WAF 限制器和干湿处理来模拟海啸传播与漫滩,再借助多分辨率嵌套网格和 GPU 高性能计算,把理论模型转化为可在 Copernicus EMS 框架下运行的海岸危险性评估工具。文中特别强调,数据一致性与海岸线定义会直接影响漫滩模拟:把海底地形与陆地高程拼接成无缝的 topobathymetric 模型时,岸线附近的微小不一致就可能引起虚假的干湿变化。应用部分给出西班牙海岸的算例:大西洋最坏情景(由 Marquês de Pombal 与 Horseshoe 断层组成的复合震源)约 50–60 分钟到达、局部波高接近 8 米;地中海多个候选震源则可能在约 15 分钟内到达,但波幅较小。作者还介绍了四层嵌套网格(分辨率从 320 米细化到 5 米),以及基于最大淹没水深和淹没范围生成的疏散路线与垂直疏散产品(PEMAREA 平台)。读这篇长文时,可重点看数学表述如何与运行流程的各个环节对应起来。
Book Review: “Mathematica: A Secret World of Intuition and Curiosity” by David Bessis (translated by Kevin Frey)
MATRIX – Residential research institute for the mathematical sciences in Australia
A message from the president // Brief words from the editor-in-chief
第 1 页 A message from the president Photo by Jim Høyer, University of Copenhagen. It has been a very intense summer pe- riod for the EMS and the worldwide mathematical community in general. At the end of June, we had the EMS council meeting. The council is the gov- erning body of the EMS. The EMS’s activities are being reviewed, and the budget suggested by the Executive Committee for the next two years is approved. The most exciting item on the council this time was the election of Professor Helge Holden as the new president of the EMS for the 2027–2030 period. Congratulations to Helge and to
On Stefan Cohn-Vossen and his family
第 1 页 On Stefan Cohn-Vossen and his family Bernd Kawohl and Michael Weigel Many mathematicians know and appreciate the book “Geometry and the imagination” by Hilbert and Cohn-Vossen [15] . It was first published in German in 1932 and has been translated into many languages. In contrast to David Hilbert, the second author, Stefan Cohn-Vossen, is less well known. In an effort to find out more about him, we realized that his life and that of his family had been affected by many dramatic and historic events. 1 Stefan Cohn-Vossen Bernhard Arnold Stefan Cohn-Vossen was born on May 28, 1902,
New editor appointed
有限群群代数中稀疏厄米平方和分解的理论与计算(杨剑霆、叶科、支丽红)
近日,中国科学院数学与系统科学研究院数学机械化实验室博士后杨剑霆与叶科研究员、支丽红研究员合作完成的论文“ Sparse Sum of Hermitian Squares in Group Algebras of Finite Groups ”被国际数学期刊 Foundations of Computational Mathematics 接收。 厄米平方和( Sum of Hermitian Squares ,简称 SOHS )是经典平方和分解在非交换代数中的推广,为非负性验证提供了重要的代数工具,在量子信息、组合优化与多项式优化等领域具有广泛应用。如何高效构造非负元素的稀疏 SOHS 分解,是该领域的重要研究问题。现有相关研究主要集中于有限交换群及具有次数限制的多项式,对一般非交换情形仍缺乏系统的稀疏 SOHS 松弛分层与误差分析。本工作将有限交换群上的稀疏傅里叶平方和研究拓展至任意有限群的群代数,并借助逼近论,建立了稀疏 SOHS 分解的松弛分层方法及相应的误差分析理论。 在松弛分层方面,研究团队证明,利用给定非负元素的平方根,可以显式构造最稀疏 SOHS 分解问题的一个凸松弛问题的最优解。在此基础上,提出基于多项式逼近的松弛分层方法,避免直接计算平方根,并降低计算规模,为稀疏 SOHS 分解提供了可计算的途径。 在误差分析方面,研究团队分别对系数的二范数残差和常数项扰动残差
On the Navier–Stokes Millennium Prize Problem
We’re sharing an AI-generated solution to the Navier–Stokes Millennium Prize Problem, including a writeup and a formal proof in Lean.
中国科学院数学与系统科学研究院携手 Project Numina:三维挂谷猜想证明完成形式化验证
近日,中国科学院数学与系统科学研究院何伟鲲研究员带领数学机械化实验室 MMAT 团队,与开源 AI4Math 组织 Project Numina 及相关高校合作,完成了从三维粘性挂谷猜想到一般三维挂谷猜想的形式化推导,并结合南开大学与 Seed 团队的相关成果,共同完成了三维挂谷猜想的形式化证明。至此,三维挂谷猜想的完整证明已在 Lean 4 中通过机器核验。 本项目以 Larry Guth 、王虹和 Joshua Zahl 于 2026 年给出的简化证明为主要参考,在将三维粘性挂谷情形中的关键估计作为外部输入的前提下,形式化了从该估计出发,经过指数迭代、管族体积估计,最终得到三维挂谷集合具有完整 Hausdorff 维数的整条推导链。 本形式化项目共包含约 41.7 万行 Lean 代码、 907 个源文件和 7575 条定理及引理声明,主要依托 Project Numina 内部开发的人机交互平台 Fuse 完成。项目由数学所何伟鲲研究员统筹指导,数学机械化重点实验室王家祺、黄逸鹤、李云斐负责数学框架梳理,路线规划与定义校验,曹一川,邱瑞晨,刘俊杞组织和调度智能体开展大规模形式化工作。 完整代码已通过构建与最终公理依赖检查。下面这行简洁的 Lean 代码,是整个项目最终抵达的形式化结论。 > `theorem KakeyaDimensionThree : ∀ (S : Set
Announcing FrontierMath Erdős
A benchmark of 68 significant Erdős problems, open as of August 2026, curated by Thomas Bloom and formalized in Lean. AI systems must prove or disprove them within a fixed budget.
MechGeo:首次在 Lean 4 中大规模完成IMO 几何题自动形式化与机器证明
MechGeo :首次在 Lean 4 中大规模完成 IMO 几何题自动形式化与机器证明 从自然语言题目到可由 Lean 内核独立核验的完整证明 核心成果: MechGeo 对 2000—2026 年 44 道 IMO 几何题构造了忠实形式化与完整 Lean 证明。据作者所知,这是目前公开报道中规模最大的自动化、经通用证明助手内核核验的 IMO 几何证明集合。 近日,中国科学院数学与系统科学研究院支丽红研究员、沈皓(博士生)和肖羽轩(硕士生),与中山大学郭俊余(硕士生)、山东大学崔添(本科生)共同提出了可信几何推理框架 MechGeo。该系统基于 Lean 4 及其大型数学库 Mathlib,将自然语言几何问题的自动形式化、命题诊断、证明构造与反例验证整合到统一的验证闭环中。 20 世纪 70 年代末,吴文俊先生提出以特征列为核心的“吴方法”,把初等几何命题转化为多项式问题并实现机械化证明,开创了数学机械化的重要方向。此后,高小山、周咸青、张景中等进一步发展几何不变量、可读证明生成和演绎数据库方法,持续推动机器证明从代数判定走向更短、更接近传统几何语言的证明。 MechGeo 延续了这一学术脉络。其名称取自“mechanized geometry”,也寄托着对吴文俊先生在几何定理机器证明领域奠基性贡献的纪念。系统进一步将大语言模型、计算机代数与通用证明助手相结合,使自动生成的证明能
平稳与非平稳混合变量的主成分分析(奚晋与合作者)
主成分分析是从大量经济金融指标中提取共同信息、监测经济周期的重要工具,但传统方法需逐一判断变量是否平稳并选择转换方式,主观性强,且易受极端冲击干扰。本文提出一种自动化方法,无需预先识别变量的平稳性或非平稳类型,即可从混合数据中提取共同周期因子。基于美国 FRED-MD 数据库的研究表明,该方法能更清晰地识别经济周期转折,有效处理金融危机、新冠疫情等极端冲击,为构建宏观经济景气指数以及预测通胀和工业产出提供了更稳健的计量工具。成果于8月被计量经济学国际顶级期刊《 Journal of Econometrics 》接收。 Publication: Journal of Econometrics Author: James D. Hamilton (corresponding author) University of California, San Diego Email:jhamilton@ucsd.edu Xinwei Ma University of California, San Diego Email: x1ma@ ucsd.edu Jin Xi Academy of Mathematics and Systems Science Email: xijin@amss.ac.cn
Ten advances in mathematics and theoretical computer science
OpenAI shares new results on long-standing open problems in mathematics and theoretical computer science, including advances in geometry, cryptography, and complexity.
Accelerating Mathematical and Scientific Discovery with Gemini Deep Think
Research papers point to the growing impact of Deep Think across fields
GPT-5 and the future of mathematical discovery
UCLA Professor Ernest Ryu and GPT-5 solved a key question in optimization theory, showcasing AI’s role in accelerating mathematical discovery.
Improving mathematical reasoning with process supervision
We’ve trained a model to achieve a new state-of-the-art in mathematical problem solving by rewarding each correct step of reasoning (“process supervision”) instead of simply rewarding the correct final answer (“outcome supervision”). In addition to boosting performance relative to outcome supervision, process supervision also has an important alignment benefit: it directly trains the model to produce a chain-of-thought that is endorsed by humans.
Solving (some) formal math olympiad problems
We built a neural theorem prover for Lean that learned to solve a variety of challenging high-school olympiad problems, including problems from the AMC12 and AIME competitions, as well as two problems adapted from the IMO.
Generative language modeling for automated theorem proving
GamePad: A learning environment for theorem proving