跳至正文

发表了就算证明吗

目录

机制裁决红队风第九十三篇 · 对称双向红队第八十八篇 · section F 数学 / 逻辑 / 认识论被神化 · 全库第 151 篇

核验图例:✓ 主笔亲核(本人取回官方原文/官方 PDF/官方元数据并逐字比对)· ◐ 一手可及但仅摘要级或仅经引述(未读全文,主张不越出可见部分)· ⚠ 二手转述或单侧来源(已标明立场)· ○ 仅题名可核(不作承重引用)。

去重与分界声明:本库同行评议与引用指标篇已裁「过评议≠正确」这条一般命题;本篇是它的最纯活体案例,并反向补一件那篇没有的东西——当期刊的接受与领域专家的判断长期相反时,还剩什么裁判。上游不重做:可证伪性归波普尔篇,复现失败的统计学归复制危机篇,数学与世界的关系归数学「不合理有效性」篇,升格病理的判据本身归方法论封顶篇一条主动的防联想:本篇争议与哥德尔不完备性无关——IUT 的争点不是「某命题不可判定」,而是「某一步推导有没有被写出来」;把二者混为一谈正是哥德尔滥用篇点名的射程越界。全库查重:abc 猜想 / IUT / 望月 / Mochizuki / Scholze 此前零覆盖。

结构胎记=发表跳 × 理解跳 × 共识跳(三跳升格链):发表跳——把「通过了某本期刊的审稿」冒充「定理成立」;理解跳——双向的,批评侧把「我读不懂」冒充「这里有洞」,支持侧把「你读不懂」冒充「你没资格反对」,两边都在拿认知状态冒充数学状态;共识跳——把「社区不接受」冒充「已被证伪」,或者把「一百万美元悬赏三年没人领」冒充「没有缺陷」。


〇 母裁决 · 七层硬度光谱

命题 硬度 判据
abc 猜想是一个真问题 ✓ 硬 Masser–Oesterlé 1985;Elkies 1991 由 abc 推出 Mordell 猜想;直接蕴含费马大定理至多有限多反例
IUT 是一次真发表 ✓ 硬 PRIMS 57 卷 1/2 期特刊,四篇分列 pp. 3–207 / 209–401 / 403–626 / 627–723,2021-03-05 出版,四个 DOI 实查
论文作者同时是该刊主编 ✓ 硬(今日实查) EMS Press「PRIMS 编委会」页 2026-07-26 仍列 Editor-in-Chief: Shinichi Mochizuki
争点可定位到一步 ✓ 硬 四方(望月/Scholze–Stix/Joshi/LANA)一致指向 IUT III「定理 3.11 ⟹ 推论 3.12」
Scholze–Stix 的那张六边形不交换 ✓ 硬(且三方同意) SS 2018 原文;LANA 2026「we concur that the diagram in Figure 7 does not commute」;望月本人亦如是说
但六边形不交换是否构成缺陷 ⚠ 未结案(独占核心) LANA:「does not, by itself, establish a defect」;真正的关口被改写成 η^q = η^anab_S
「abc 已证」与「IUT 是骗局」双向皆软 × 软 LANA 2026-07 同一段话同时拒绝两侧;Scholze 署名评论与望月早年硬定理并存于同一份文本

母裁决:abc 猜想是硬的,IUT 的发表是真的,争点是可定位的——但「abc 已被证明」未立,「IUT 是骗局」同样未立。 三件事必须分开数:第一,猜想本身立得住(1985 年起源,Elkies 1991 证明它蕴含 Mordell 猜想,Scholze 在自己署名的评论里写它「arguably the most central open Diophantine problem」)。第二,望月新一 2012 年释出的四篇论文确实在 2021 年由 PRIMS 作为特刊正式发表,页码、DOI、出版日期全部可查,另有一篇五作者的显式估计论文发表于 Kodai Math. J.——这不是「预印本没人认」,这是「真期刊真发表」。第三,从 2012 年到 2026 年,全世界没有出现过一次公开的、双方都承认的验证收敛:Scholze 与 Stix 2018 年逐字写下「there is no proof」,望月逐字回以「fundamental misunderstandings」且「does not imply the existence of any flaws whatsoever」,Joshi 2025 年逐字写下「every assertion of [Scholze–Stix] … is mathematically false」同时又写「Mochizuki’s proof is also incomplete」——三份文件互相否证,谁也没能让对方或让第三方收敛。十四年里唯一真正往前挪了一格的事,发生在 2026 年:一个把 IUT 内部专家请进来的中立团队,花了一年半,终于把那场关于「谁读懂了」的争吵,压成了一条能写进 Lean 的等式 η^q = η^anab_S;然后当众说,我们没有它的证明,也没有它的反证。 越界在「发表=证明」和「不被接受=已被驳倒」这两跳,不在这些论文本身。

灵魂句:数学本来是人类唯一能把对错做成客观事实的地方——可 abc 这十四年证明的是,客观性不会自动送达:它不是期刊盖的章,不是专家点的头,也不是一百万美元悬赏三年没人领;它是一条别人能独立走完的路,而在那一步被写成机器能读的一行字之前,所有人手里握着的都只是关于彼此的判断。


一 真锚一:abc 猜想是一个真问题

1.1 陈述

abc 猜想的最简形式,用 Scholze 本人在 zbMATH 上署名评论里写的版本 [一手逐字]:

“Recall that in its simplest form, the ABC conjecture states that for all ε > 0 there is some constant C_ε such that for all coprime positive integers a, b, c satisfying a + b = c, one has c ≤ C_ε (∏_{p|abc} p)^{1+ε}. Here the product runs over all primes p dividing one of a, b and c, but crucially not counting multiplicity.”

来源:Peter Scholze, 关于 Mochizuki《Inter-universal Teichmüller theory I》的 zbMATH 评论,Zbl 1465.14002(✓ 主笔亲核;zbMATH 官网当前对本环境返回 Cloudflare 拦截,全文经 Wayback 2025-03-22 存档 PDF 逐字取回)

那个「crucially not counting multiplicity」是全部要害:radical(根基)只数不同素因子各一次。所以猜想说的是——a + b = c 这种加法关系,会限制三个数的乘法结构有多「浓缩」。加法与乘法的相互约束,正是数论最核心的那根神经。

Scholze–Stix 2018 年那份报告在正文第一页给的是等价的对数形式,并顺带把归属钉死 [一手逐字]:

“The abc-conjecture goes back to Masser and Oesterlé in 1985: Conjecture 1. Let ε > 0. Then for all coprime integers a, b, c with a + b + c = 0 we have log(max{|a|, |b|, |c|}) < (1 + ε) · Σ_{p|abc} log(p) + O(1) where the constant O(1) depends only on ε.”

来源:Peter Scholze, Jakob Stix,《Why abc is still a conjecture》(2018-08-23 版本),math.uni-bonn.de/people/scholze/WhyABCisStillaConjecture.pdf(✓ 主笔亲核全文 PDF;望月官网镜像同一文件)

Masser 1985 与 Oesterlé 1988 的原始文献本人未取回,标 ○,不作承重——归属这件事由上面两份亲核文件承担,两份文件分属对立双方,交叉一致。

1.2 为什么它值一场十四年的争论

abc 是那种「一旦成立就顺手推倒一排」的猜想。Scholze 同一条评论里点了两件事 [一手逐字]:

“For example, applied to a = x^n, b = y^n and c = z^n it formally implies that there are at most finitely many counterexamples to Fermat’s Last Theorem; if C_ε can be made explicit, it thus in principle reduces it to a finite computation. A proof of the ABC conjecture would also lead to a new proof of Mordell’s conjecture [N. D. Elkies, Int. Math. Res. Not. 1991, No. 7, 99–109].”

Elkies 那篇的元数据实查为:International Mathematics Research Notices 1991 年第 7 期,起始页 99,DOI 10.1155/s1073792891000144(✓ Crossref 实查)。也就是说,abc 蕴含 Faltings 定理(Mordell 猜想)——一个已经拿过菲尔兹奖的定理,会变成 abc 的推论。

同一条评论还写清了它与 Szpiro 猜想的关系 [一手逐字]:

“The ABC conjecture is known to imply the Szpiro conjecture, but the converse fails, essentially because the discriminant of elliptic curves does not contain an ‘Archimedean factor’. The author observed in [Math. J. Okayama Univ. 52, 1–28 (2010)] that if one formulates suitably uniform versions of both conjectures for number fields, they become equivalent.”

这段值得注意的地方是:它出自那位说「这不是证明」的人之手,而且它在承认望月早期工作的贡献。 这一条会在第八节的红线里再被用到。

1.3 数值证据:它「几乎总是」成立,而且余量很薄

abc 猜想有一个方便的量化指标:对互素正整数 a + b = c,令 r = rad(abc),质量 q = log c / log r。猜想说 q > 1 + ε 的三元组只有有限多个。

已知质量最高的三元组是 2 + 3¹⁰·109 = 23⁵,q = 1.62991,由 Eric Reyssat 发现。来源:Abderrahmane Nitaj(诺曼底大学 LMNO)维护的 abc conjecture home page 的「Table I. The top ten good abc-examples」(◐ 一手页面亲核,但该表为社区长期维护的记录汇编,非同行评议文献)。

这个数字很说明问题:七十年的搜索、包括分布式计算项目大规模枚举,把质量的经验上界推到 1.63 就再也推不动了。这不是证明,但它是那种让人相信猜想为真的证据类型——本库模型成功与科学实在论篇的分级里,这属于「现象层复现许可」,不许可任何本体或证明层结论。

这一节的作用是划出真锚:后面所有的争论都不是「abc 是不是个伪问题」。abc 是真的,重要,而且至今是猜想。


二 真锚二:IUT 是一次真发表,不是一次「网上贴了篇文章」

2.1 时间线

  • 2012 年 8 月:望月新一(京都大学数理解析研究所 RIMS)在个人主页释出《Inter-universal Teichmüller Theory I–IV》四篇,并投稿。
  • 2015 年 12 月 / 2016 年 7 月:牛津、RIMS 两场大型国际研讨会。
  • 2018 年 3 月 15–20 日:Scholze 与 Stix 访京都,与望月、星裕一郎(Yuichiro Hoshi)闭门讨论五天。
  • 2018 年 5 月 / 8 月:Scholze–Stix 两版报告;望月两份逐条回应;同年 9 月《量子杂志》报道。
  • 2020 年 4 月 3 日:京都大学召开记者会宣布接受。
  • 2021 年 3 月 5 日:PRIMS 第 57 卷第 1/2 期特刊正式出版。
  • 2022 年 6 月:五作者《Explicit estimates in IUT》发表于 Kodai Math. J.。
  • 2023 年 7 月 7 日:IUT 挑战者奖(100 万美元)与 IUT 创新者奖(2–10 万美元/年)设立。
  • 2024–2025 年:Joshi 系列预印本与望月的逐条反驳;2025 年 10 月望月发布现状报告并首次大篇幅讨论 Lean。
  • 2026 年 3 月 31 日 / 7 月 17 日:LANA 项目公开存在、发布中间报告。
  • 2026 年 4 月:望月本人在「AI 与定理证明器」研讨会上作 IUT 形式化进展报告。

2.2 发表事实(Crossref / EMS Press 实查)

论文 卷/期/页 DOI 出版日
IUT I: Construction of Hodge Theaters 57 (1/2), 3–207 10.4171/prims/57-1-1 2021-03-04
IUT II: Hodge–Arakelov-Theoretic Evaluation 57 (1/2), 209–401 10.4171/prims/57-1-2 2021
IUT III: Canonical Splittings of the Log-Theta-Lattice 57 (1/2), 403–626 10.4171/prims/57-1-3 2021
IUT IV: Log-Volume Computations and Set-Theoretic Foundations 57 (1/2), 627–723 10.4171/prims/57-1-4 2021
Preface to the Special Issue 57 (1/2), p. 1 10.4171/prims/57-1-0 2021-03-05

(✓ 主笔亲核:五条元数据经 Crossref APIEMS Press 该刊页面 双路实查。特刊序言正文在付费墙后,未读,不作承重。)

第五篇是:Shinichi Mochizuki, Ivan Fesenko, Yuichiro Hoshi, Arata Minamide, Wojciech Porowski,《Explicit estimates in inter-universal Teichmüller theory》,Kodai Math. J. 45 (2022), 175–236,DOI 10.2996/kmj45201(✓ Crossref 实查元数据;页码取自 IUGC 官方新闻稿)。

望月本人在 2025 年 10 月的现状报告里对这一层的表述是 [一手逐字]:

“IUT is a well established mathematical theory that currently consists of five mathematical papers published in two internationally recognized mathematical journals and, in particular, has passed peer reviews for these journals.”

来源:Shinichi Mochizuki,《Report on the Current Situation Surrounding Inter-universal Teichmüller Theory (IUT)》(October 2025),kurims.kyoto-u.ac.jp/~motizuki/IUT-report-2025-10.pdf(✓ 主笔亲核全文 PDF)

这句话每一个字都是事实。它也正是本篇要审的那一跳的完美样本:从「passed peer reviews」到「well established mathematical theory」之间,隔着一整篇文章。

2.3 作者同时是主编——这条今天仍然为真

EMS Press 的 PRIMS 编委会页面,本人于 2026-07-26 实查,页面首行为 [一手逐字]:

“Editor-in-Chief — Shinichi Mochizuki — RIMS, Kyoto University”

来源:ems.press/journals/prims/editorial-board(✓ 主笔亲核,2026-07-26 取回)

关于回避:多处二手报道称望月在本刊审稿中回避了。本人未能取回任何 RIMS 或 PRIMS 的官方声明原文——Nature 2020 年的报道在付费墙后,本人只读到标题与首段。因此这一条标 ⚠,且不作任何方向的承重:既不据此认定程序不当,也不据此认定程序无瑕。可核的只有两件事:作者是该刊主编(✓),以及当时公开的质疑存在(下引)。

Peter Woit 在发表消息见报当天的博文里写 [有争议]:

“It’s completely unheard of for a major journal to publish a proof of an important result when experts have publicly stated that the proof is flawed and are standing behind that statement. That Mochizuki is the chief editor of the journal and that the announcement was made by two of his RIMS colleagues doesn’t help at all with the situation.”

来源:Peter Woit,「Latest on abc」,Not Even Wrong,2020-04-03,math.columbia.edu/~woit/wordpress/?p=11709(✓ 原帖亲核;作者为公开的批评侧,立场已标注)

同一篇博文转引了 Nature 报道中的两句话(◐ 经引述,Nature 原文付费墙未读):Scholze——「My judgment has not changed in any way since I wrote that manuscript with Jakob Stix.」;Kedlaya——「I think it is safe to say that there has not been much change in the community opinion since 2018.」记住 Kedlaya 这句 2020 年的话,第七节它会以同一个人的名义被本人修正。


三 争点是可定位的:一切都压在推论 3.12 上

这是本案最反直觉、也最重要的一件事:争议不是「六百页没人看得懂」,而是「所有人都同意问题出在哪一步」。

3.1 那一步

IUT IV 的定理 1.10 由 IUT III 的推论 3.12 推出;推论 3.12 又该由 IUT III 的定理 3.11(多重径向算法,multiradial algorithm)推出。Scholze–Stix、望月、Joshi、LANA 四方所有文本,指的都是同一步:3.11 ⟹ 3.12

Adam Topaz(阿尔伯塔大学,LANA 核心成员)2026 年 7 月的演讲稿把它说得最白 [一手逐字]:

“The statement at the center of the controversy is Corollary 3.12 in the third IUT paper. This is a statement asserting an inequality between two real numbers. This corollary is meant to follow from Theorem 3.11 of IUT3, which describes the so-called ‘Multiradial Algorithm’ of IUT. In fact, one may consider Theorem 3.11 as a summary of the whole of IUT itself.”

来源:《Speech drafts in LANA interim report July 2026》(2026-07-18 版),zen-univ.jp/en/zmc/topics/arn1ligpl(✓ 主笔亲核 PDF 全文)

《量子杂志》2018 年的报道里也记下了同一件事的社会学侧面(◐ 媒体报道,含直接引语):Scholze 完成四篇的「粗读」后卡在 3.12;Brian Conrad 2015 年牛津会议后发帖,几天内收到三位数学家的来信,「For each of these people, the proof that had stumped them was for 3.12」。来源:Erica Klarreich,「Titans of Mathematics Clash Over Epic Proof of ABC Conjecture」,Quanta Magazine,2018-09-20,quantamagazine.org

还有一条硬指标,出自 Scholze 的 zbMATH 评论 [一手逐字]:

“In parts II and III, with the exception of the critical Corollary 3.12, the reader will not find any proof that is longer than a few lines; the typical proof reads ‘The various assertions of Corollary 2.3 follow immediately from the definitions and the references quoted in the statements of these assertions.'”

推论 3.12 是中间两篇里唯一一个证明超过几行的结论,占九页。 这既是批评侧的抱怨,也是一个客观的定位信号:靶子小到可以被围住。

3.2 Scholze–Stix 说了什么(逐字)

他们的结论句在报告开头第二段 [一手逐字]:

“We, the authors of this note, came to the conclusion that there is no proof. We are going to explain where, in our opinion, the suggested proof has a problem, a problem so severe that in our opinion small modifications will not rescue the proof strategy.”

技术主张则极其具体——不是「看不懂」,是「那个 j² 因子不该在那里」。他们论证在一致地识别各份实数拷贝之后,推导实际给出的是缺了 j² 的版本,于是 [一手逐字]:

“Starting from the corrected inequality we obtain 0 ⪅ d(P) which is essentially free of content.”

以及最后一节的结论 [一手逐字]:

“The conclusion of this discussion is that with consistent identifications of copies of real numbers, one must in (1.5) omit the scalars j² that appear, which leads to an empty inequality.”

还有对望月现场答复的评估 [一手逐字]:

“On the fifth and final day, Mochizuki tried to explain to us why this is not a problem after all. In particular, he claimed that up to the ‘blurring’ given by certain indeterminacies the diagram does commute; it seems to us that this statement means that the blurring must be by a factor of at least O(ℓ²) rendering the inequality thus obtained useless.”

一个常被忽略的脚注,值得单独摘出来,因为它划了批评的边界 [一手逐字]:

“We pause to observe that with the simplifications outlined above, such as identifying identical copies of objects along the identity, the critical [IUTT-3, Theorem 3.11] does not become false, but trivial.”

Scholze–Stix 从来没有主张 IUT 自相矛盾或定理 3.11 是假的。他们主张的是:在他们能一致解读的框架里,那一步给出的不等式是空的。这一区分在 2025–2026 年被反复误传,望月的现状报告有整整一节在纠正这件事(见 3.3)。

3.3 望月说了什么(逐字)

望月 2019 年 2 月定稿的讨论报告,核心判断在第二节 [一手逐字]:

“The negative position of SS is a consequence of certain fundamental misunderstandings … on the part of SS concerning IUTch, and, in particular, does not imply the existence of any flaws whatsoever in IUTch.”

来源:Shinichi Mochizuki,《Report on Discussions, Held during the Period March 15 – 20, 2018, Concerning Inter-universal Teichmüller Theory (IUTch)》(February 2019),kurims.kyoto-u.ac.jp/~motizuki/Rpt2018.pdf(✓ 主笔亲核全文 PDF)

他给的「非常粗糙的」类比模型 (Smm) 是理解他立场最快的入口 [一手逐字]:设 A、B 为正实数满足 −2B = −A(对应 Θ-link),定理给出 −2B ≤ −2A + 1(对应定理 3.11),两者合起来得 A ≤ 1(对应推论 3.12);而在他看来 Scholze–Stix 的做法相当于额外假设 A = B,那当然立刻推出 A = B = 0 的「矛盾」——

“the ‘contradiction’ A = B = 0 is nothing more than a superficial consequence of the extraneous assumption ‘A = B’ and, in particular, does not imply the existence of any flaws whatsoever in IUTch.”

他随即把这条上升为一般逻辑原则 (GLR1)/(GLR2) [一手逐字]:

“Given any mathematical argument, it is always easy to derive a contradiction by arbitrarily identifying mathematical objects that must be regarded as distinct in the situation discussed in the argument. On the other hand, this does not, by any means, imply the existence of any logical flaws in the original mathematical argument!”

以及关于时间的那句,后来被反复引用 [一手逐字]:

“substantial progress in understanding IUTch always requires discussions over an extended period of time, typically on the order of months.”

2025 年 10 月的现状报告里,他把双方立场的差别做了一次自己的归纳 [一手逐字]:

“(SSA1) they do not see how to derive any nontrivial diophantine consequences, i.e., such as height inequalities, from the theory of [IUTchI, II, III, IV].”

“(ExPr) the existence of an algorithm … that is unsuccessful in yielding any nontrivial consequences … does not, in any way, imply the non-existence/falsehood of an algorithm … that is successful in yielding nontrivial consequences.”

同一份报告里他还明确否认了流传的「IUT 的设定本身含矛盾」版本,并指出 Scholze–Stix 从未这么主张 [一手逐字]:

“On the contrary, Scholze-Stix repeatedly emphasized — i.e., during the March 2018 discussions, as well as in their manuscript — that they do not contest the mathematical validity of [IUTchIII], Theorem 3.11.”

这段值得停一下:望月本人在 2025 年出面澄清「批评者并没有说我的主定理是假的」——这既是他的辩护,也是本篇能确认的事实,而且它与 Scholze–Stix 那条脚注互相印证。双方对「争的是什么」的描述,比大多数报道以为的要一致得多。

3.4 发表跳的最纯活体:同一篇论文,两份永久记录

这是全案最锋利的一件事,而且它不是修辞:

  • 记录一:PRIMS 第 57 卷 1/2 期特刊,四篇论文,2021 年 3 月 5 日出版,通过了该刊的审稿。
  • 记录二:zbMATH Open(数学界两大标准评论数据库之一)中 Zbl 1465.14002 条目下,署名 Peter Scholze (Bonn) 的评论,针对的正是已发表版本,逐字写着 [一手逐字]:

“Unfortunately, the argument given for Corollary 3.12 is not a proof, and the theory built in these papers is clearly insufficient to prove the ABC conjecture.”

“In any case, at some point in the proof of Corollary 3.12, things are so obfuscated that it is completely unclear whether some object refers to the q-values or the Θ-values, as it is somehow claimed to be definitionally equal to both of them, up to some blurring of course, and hence you get the desired result.”

“Together with J. Stix, the reviewer has spent a week in Kyoto to discuss these issues with the author … The concerns expressed in this manuscript have not been addressed in the published version.”

这两条记录都在,都永久,都可引用,而且它们互相矛盾。 没有任何制度机制去消解它。一篇论文可以同时是「PRIMS 发表的定理」和「标准评论数据库里被点名说不是证明的东西」——如果「过评议=正确」这条推论成立,那这个世界就不该长成这样。

同行评议篇给出的是这条推论的统计学与制度史反驳;本案给出的是它的存在性反例,而且是最强形式:不是劣质期刊、不是掠夺性出版、不是审稿疏忽,是一本真刊、多年审稿、被公开质疑之后照样出版,并且质疑至今在库。

3.5 语气档案:归措辞的归措辞

两侧的公开文本都出现过远超学术惯例的措辞。这一节只记录,不引申——因为措辞不是数学证据,而本篇的红线之一是不污名泛化。

望月对 Scholze–Stix 2018 年 8 月版第一条评论 [一手逐字]:

“I can only say that it is a very challenging task to document the depth of my astonishment when I first read this Remark! This Remark may be described as a breath-takingly (melo?)dramatic self-declaration, on the part of SS, of their profound ignorance of the elementary theory of heights, at the advanced undergraduate/beginning graduate level.”

来源:《Comments on the Manuscript (2018-08 version) by Scholze-Stix》(September 2018),kurims.kyoto-u.ac.jp/~motizuki/Cmt2018-08.pdf(✓ 主笔亲核)

他 2024 年对 Joshi 系列预印本的评价 [一手逐字]:

“(ShtAns) … it is conspicuously obvious to any reader of these preprints who is equipped with a solid, rigorous understanding of the actual mathematical content of inter-universal Teichmüller theory that the author of this series of preprints is profoundly ignorant of the actual mathematical content of inter-universal Teichmüller theory, and, in particular, that this series of preprints does not contain … any meaningful mathematical content whatsoever.”

来源:《Report on the Recent Series of Preprints by K. Joshi》(March 2024),kurims.kyoto-u.ac.jp/~motizuki/Report on a certain series of preprints (2024-03).pdf(✓ 主笔亲核)

批评侧同样有过尖锐表达:望月在《量子杂志》上被引述说,他向同事描述 Scholze–Stix 的反对意见时,「were met with a remarkably unanimous response of utter astonishment and even disbelief (at times accompanied by bouts of laughter!)」(◐ 经媒体引述其书面回应);而 Fesenko 在 2023 年的奖项新闻稿中把公开质疑者的行为直接指为违反欧洲数学会行为准则(见 5.3)。

裁决:这些文本是这场争议社会学的一手材料,不是任何一方数学正确性的证据,也不构成对任何个人的品格判断。本篇不做人身评价。


四 不是二元对立:五种立场的分账表

把这件事讲成「望月 vs 全世界」或者「天才 vs 傲慢的西方主流」,两种叙事都不成立。截至 2026 年 7 月,公开可查的立场至少有五种,而且互不相容的方式各不相同

立场 核心主张 对 Scholze–Stix 的态度 对「abc 已证」的态度 硬度
望月及 RIMS 团队 IUT 正确,批评源于基本误解 是误解,不构成缺陷 已证,且有显式常数版本 五篇两刊已发表;批评者未撤回
Scholze–Stix 那一步不是证明,且小修补救不了 —— 未证 报告未在期刊发表;zbMATH 署名评论在库
Kirti Joshi SS 的每一条断言都是错的,同时望月的证明也不完整;补全需要他自己的理论 全部驳倒 「由 Joshi 系列+望月系列合起来成立」 四篇关键预印本至今无期刊发表记录
Kedlaya(2020→2026) 2020:社区意见没变;2026:那个「不一致」的外观源于沟通失误,状态仍未决 其结论造成了「已定案」的错误共识 未证,但可能补得上 立场公开转变,两端均可查
LANA 项目(2026) 争点=η^q = η^anab_S;我们没有它的证明,也没有它的反证 六边形确实不交换,但不构成缺陷 不能说已证,也不能说已驳倒 一年半集体工作+中间报告公开

(表中每一条的原始出处见第三、五、七节及末尾来源清单。)

Joshi 那一行值得特别注意,因为它是唯一的「双否」立场,而且它自己写得极清楚 [一手逐字]:

“As Table §1.3 shows, every assertion of [Scholze and Stix, 2018] and [Scholze, 2021] is mathematically false. On the other hand, Mochizuki’s proof is also incomplete (see §1.2). A robust version of the theory claimed by Mochizuki is provided by my work.”

“At the very center of the issue is that Mochizuki’s quantification of what it means to be an Arithmetic Holomorphic Structure is mathematically inadequate to quantitatively assert that one has two or more such structures.”

“Peter Scholze and Jakob Stix recognized this problem (2018) – but they extrapolated and argued (incorrectly) that many such structures cannot exist.”

来源:Kirti Joshi,《Final Report on the Mochizuki-Scholze-Stix Controversy》,arXiv:2505.10568(2025-04-29 提交;✓ 主笔亲核 arXiv HTML 全文)

而望月对这一整条线的回应是全盘否定(3.5 已引),并附了一条纯事实陈述 [一手逐字]:「(NotPb) none of these preprints has been published in an internationally recognized mathematical journal.」——本人于 2026-07-26 逐一实查 Joshi 四篇关键预印本的 arXiv 页面(2106.114522210.116352401.135082403.10430),均无期刊发表记录(✓)。

这张表的意义是杀死一个流行叙事。 「数学界一致否定 IUT」不成立(Joshi 在为它的核心辩护、Kedlaya 在为它翻案、LANA 明确不下否定判断);「只有望月圈内人支持」也已经过期(LANA 五名核心成员里有三位与望月无历史关联,且项目起点是刻意保持独立)。同时,「IUT 已被接受」更不成立。


五 制度层:期刊、奖金、行为准则,以及一个被设错的判定门槛

5.1 一百万美元的挑战者奖

2023 年 7 月 7 日,日本财团与 DWANGO 教育机构筹备会在 ZEN 大学(当时筹建中)宇宙际几何中心(IUGC,主任加藤文元)发布新闻稿,设立两个奖 [一手逐字]:

“The IUT Innovator Prize will be awarded annually within a scope from $20,000 to $100,000 to the best paper containing new and important developments in IUT theory and related fields.”

“The IUT Challenger Prize of $1,000,000 will be awarded to the first mathematician to write a paper on the IUT theory that shows an inherent flaw in the theory.”

评审规则也写明了 [一手逐字]:

“The judging will be conducted by Nobuo Kawakami on his own initiative. The method of judging will not be made public, but the papers to be judged must be peer-reviewed and published in a mathematical journal that is covered by MathSciNet and has published at least 10 papers on arithmetic geometry in the past 10 years.”

创新者奖一侧则是 [一手逐字]:「The papers will be evaluated by a panel of experts in the IUT theory appointed by IUGC in a closed review process independent of peer review in the journal to which the paper is submitted.」

来源:ZEN 大学新闻稿,「Establishment of International Award for IUT Theory and First IUGC Conference on IUT Theory」,2023-07-07,zen.ac.jp/news/0ul6zqed9-0(✓ 主笔亲核全文)

出资人川上量生本人的说明相当坦率,也值得逐字保留 [一手逐字]:

“It is a common phenomenon in politics and society that things that are supposed to be true or false continue to be debated without ever being settled. However, I am honored to have witnessed, as a witness of history, a rare case in which the same thing is happening in mathematics, where such a thing should not happen. In my opinion, the root of the problem is that the cost of learning to argue the correctness of the IUT theory has become too high, even for superlative mathematicians. … Of course, I am not a mathematician, so I am not competent to judge the correctness or incorrectness of the arguments. My hope is that the IUT theory, if it is wrong, will be settled in the mathematical community under a sound mathematical debate.”

2024 年 4 月 2 日,第一届创新者奖 10 万美元授予前述 Kodai 五作者论文;四位 RIMS 作者受奖(Fesenko 谢绝奖金),并表示希望把奖金捐给 RIMS。来源:zen.ac.jp/news/d-5ye560_l(✓ 亲核)与望月本人的受奖声明(✓ 亲核)。

5.2 共识跳的解剖:「三年没人领」证明不了任何事

一个流行推论是:既然一百万美元挂了三年(2023-07 至今)没人拿走,说明没人能指出缺陷。这条推论不成立,理由是判定门槛本身,而不是任何一方的诚信问题

  1. 门槛要求「期刊发表的反驳论文」。而 Scholze–Stix 的报告从未投期刊——它是作者主页上的手稿,加上一条 zbMATH 署名评论。按规则,这份数学界最著名的反对意见先天不具备参评资格
  2. 反驳性论文本来就极难发表。数学期刊很少发「某文第 X 页有误」类稿件,尤其当被反驳方是该领域少数专家之一时。
  3. 评审规则不公开、由非数学家个人裁定——这是出资人自己写明的,并非隐瞒;但它意味着奖项无法充当共同体裁判。
  4. 对称地看,它也不是坏事:这笔钱是明码标价押注自己一方可能错,比只奖励支持者的结构诚实得多;创新者奖同期也确实资助了研究。

裁决:奖金是一个真诚的激励尝试,但它不是一台判定机器;把「未被认领」当作「无缺陷」的证据,是共识跳的教科书形态。 这与本库测量代理性篇裁过的病理同型:把一个为激励设计的指标,当成它从未承诺过的判定结论。

5.3 行为准则被双向援引,这本身就是诊断

欧洲数学会《行为准则》「作者的责任」第 6 条,被这场争议的两侧当作武器:

  • 望月反复引用它要求批评者与竞争者提供完整细节,2025 年 10 月的报告里逐字引作 [一手逐字]:「Mathematicians should make public claims of potential new theorems only when they believe they are able to provide full details in a timely manner, to avoid unnecessarily blocking an active line of research.」
  • Fesenko 在 2023 年奖项新闻稿中把矛头指向另一侧 [一手逐字]:「Making public statements about alleged faults in IUT without providing mathematical evidence of them is a violation of Article 6 of the EMS Code of Practice.」

同一条准则、同一个编号,被用来指控相反的两种行为——这正是制度层已经失去裁决能力的标志。 规范能约束的是发言方式,不能替代对错。本库同行评议篇的一般结论在这里落地:当代理机制被推去承担它设计上就不承担的功能时,它不会给出错误答案,它会给出两个互相抵消的答案。


六 对照组:难验证的证明,通常是怎么收敛的

要判断 IUT 的反常点在哪,必须有基线。数学史上「难到共同体一时验证不了」的证明不止一例,而它们全都收敛了——用四种不同的方式。

6.1 费马大定理:14 个月,公开修补

Wiles 1993 年 6 月宣布,审稿过程中发现缺口,1994 年 9 月与 Taylor 补上。最终发表形态是两篇并列:Andrew Wiles,《Modular Elliptic Curves and Fermat’s Last Theorem》,Ann. of Math. 141 (1995) 443,DOI 10.2307/2118559;Richard Taylor & Andrew Wiles,《Ring-Theoretic Properties of Certain Hecke Algebras》,Ann. of Math. 141 (1995) 553,DOI 10.2307/2118560(✓ Crossref 实查)。

收敛机制:作者承认缺口 → 公开修补 → 补丁本身作为独立论文接受审稿。 关键不是「没有出错」,而是缺口被双方共同承认为缺口

6.2 开普勒猜想:审稿人认证不了,先发表,十六年后形式化

Hales 与 Ferguson 1998 年宣布,Annals 的审稿持续多年。Hales 本人在 2015 年的形式化论文里写下这段历史 [一手逐字]:

“The delay in publication was caused by the difficulties that the referees had in verifying a complex computer proof. Lagarias has described the review process [30]. He writes, ‘The nature of this proof . . . makes it hard for humans to check every step reliably. . . . [D]etailed checking of many specific assertions found them to be essentially correct in every case. The result of the reviewing process produced in these reviewers a strong degree of conviction of the essential correctness of this proof approach …’ In the end, the proof was published without complete certification from the referees.”

来源:Thomas Hales 等 15 人,《A Formal Proof of the Kepler Conjecture》,Forum of Mathematics, Pi 5 (2017), DOI 10.1017/fmp.2017.1;引文取自 arXiv:1501.02155 全文 PDF(✓ 主笔亲核)

Flyspeck 项目 2003 年启动,2014 年完成,2017 年发表官方账目:「This paper constitutes the official published account of the now completed Flyspeck project.」

收敛机制:先发表(明确标注未完全认证)→ 十六年后由形式化终结。 这条线与 IUT 的相似度最高——一本顶刊在审稿人无法完全认证的情况下发表了一个重要证明——但有一个决定性差别:Hales 一方与审稿人一方对「什么还没被验证」有共同描述,而且是作者自己去做的形式化

6.3 四色定理:机器进场引发的第一次范式争论

Appel–Haken 1976 年的证明依赖大量机器枚举,引发了「这算不算证明」的哲学争论;Gonthier 2005 年在 Coq 中完成完整形式化(会议版:G. Gonthier,《The Four Colour Theorem: Engineering of a Formal Proof》,LNCS,DOI 10.1007/978-3-540-87827-8_28,✓ Crossref 实查)。收敛机制:争议从「对不对」转成「什么算证明」,最终由形式化关闭。

6.4 液态张量实验:18 个月,而且是批评者自己把自己的定理送去机检

2020 年 12 月 5 日,Peter Scholze 在 Xena 博客发出挑战,要求把他与 Clausen 的液态实向量空间定理形式化。他给出的理由,是这整篇文章里最重要的一段外部对照 [一手逐字]:

“with this theorem, the hope that the condensed formalism can be fruitfully applied to real functional analysis stands or falls. I think the theorem is of utmost foundational importance, so being 99.9% sure is not enough.”

“I spent much of 2019 obsessed with the proof of this theorem, almost getting crazy over it. In the end, we were able to get an argument pinned down on paper, but I think nobody else has dared to look at the details of this, and so I still have some small lingering doubts.”

“I have occasionally been able to be very persuasive even with wrong arguments. (Fun fact: In the selection exams for the international math olympiad, twice I got full points for a wrong solution. Later, I once had a full proof of the weight-monodromy conjecture that passed the judgment of some top mathematicians, but then it turned out to contain a fatal mistake.)”

来源:Peter Scholze(客座),「Liquid tensor experiment」,Xena 博客,2020-12-05,xenaproject.wordpress.com(✓ 主笔亲核全文);论文版:Experimental Mathematics 31 (2022) 349–354,DOI 10.1080/10586458.2021.1926016(✓ Crossref 实查)

2022 年 7 月 15 日,Lean 社区宣布完成 [一手逐字]:「We are proud to announce that as of 15:46:13 (EST) on Thursday, July 14 2022 the Liquid Tensor Experiment has been completed. A year and a half after the challenge was posed by Peter Scholze we have finally formally verified the main theorem of liquid vector spaces using the Lean proof assistant.」来源:leanprover-community.github.io/blog/posts/lte-final/(✓ 亲核)。领队正是 Johan Commelin 与 Adam Topaz——四年后 LANA 项目的两位核心成员

这一条同时干两件事:它给出了「18 个月可以机检一个前沿定理」的存在性证明;它也堵死了「Scholze 只对别人严格」的叙事——他公开写下自己曾有一个通过顶级数学家判断的错误证明,并因此不信任 99.9% 的把握

6.5 对照之后:IUT 的反常点在哪

不是「难」——开普勒猜想更依赖机器,四色定理更早引发范式危机,Wiles 的缺口更深藏。IUT 的反常在于:它是唯一一个双方连「什么还没被验证」都长期无法共同描述的案例。 费马、开普勒、四色、液态张量,四条线的收敛都始于同一个动作——争议双方对「未验证的是哪一步、以什么形式」达成一致描述。IUT 卡了整整八年(2018–2026)没有做到这件事。

而 2026 年发生的,正好就是这件事。


七 2026:唯一真正挪动的一格

7.1 LANA:把 IUT 内部专家请进中立团队

LANA(Lean for ANAbelian geometry)项目 2026 年 3 月 31 日公开存在,由 ZEN 数学中心(ZMC)主办,出资方为日本财团与 DWANGO 教育机构,项目负责人加藤文元。五名核心成员:Johan Commelin(乌得勒支)、Kiran Kedlaya(加州大学圣迭戈)、星裕一郎(RIMS)、Adam Topaz(阿尔伯塔)、加藤文元,另有约七名年轻成员。

它的两条目标与自我定位,加藤在公开演讲稿里写得很明确 [一手逐字]:

“The second aim is to verify inter-universal Teichmüller theory … likewise by means of Lean formalization. Without leaning toward any particular position, we aim, from a neutral point of view, to put this theory into a form that can be formalized and, by verifying it, to bring to an end the controversy surrounding this theory that has continued for many years.”

来源:《For the LANA Project Announcement》(2026-03-31 记者会演讲稿),zen-univ.jp/en/zmc/topics/jwz-o8xr3v6f(✓ 主笔亲核全文)

关键设计:中间报告写明,项目起初刻意要求与望月本人保持一定独立性,专家成员的选择「required care」,最终由加藤在 2024 年 12 月说服星裕一郎加入;而星在项目中承担的角色,是把 IUT 的内容尽可能完整地讲给其他成员,其主要目标就是让成员理解「从定理 3.11 推出推论 3.12 的逻辑」(✓ 中间报告 §0.2 与星本人 2026-03 演讲稿)。

一年半里的集体工作方式是密集 bootcamp:镰仓–逗子(2025-02)、基洛纳(2025-05)、东银座(2025-07)、赞丹(2025-11)、圣迭戈(2026-02)。中间报告称,圣迭戈那次「the ‘wall’ was extracted in a fairly clear form」(✓ 亲核)。

7.2 墙被压成一条等式

LANA 中间报告的摘要写得极其克制 [一手逐字]:

“we isolate a specific compatibility problem at the final stage of the argument: the relation between the construction arising directly from the q-pilot in its native arithmetic holomorphic structure and the construction obtained by anabelian and Kummer-theoretic methods through the multiradial procedure. … At present, however, Project LANA has not yet reconstructed a proof of the required compatibility. Accordingly, this report does not offer a final verdict on the validity of IUT theory.”

正文 §9.2 把它写成一条编号命题 (9-1):存在合适的 S 使 η^q = η^anab_S。而 §10.5 的临时评估是全案最重要的一段 [一手逐字]:

“To reiterate, we concur that the diagram in Figure 7 does not commute, and that repairing it would require a rescaling factor that is too large to yield a proof of the abc conjecture. … On the other hand, our analysis does not give rise to this particular diagram; rather, the elaboration of the η-algorithm shows that the proof of the final numerical inequality hinges on the compatibility (9-1) which is not manifestly false. However, we, the LANA project, do not have a proof of (9-1) at this time.”

最后一句尤其不能省 [一手逐字]:

“while many LANA members hold the view that ‘the original paper does not contain at least a formalizable proof,’ the members were unable to reach complete consensus on this point.”

来源:LANA Project,《Project LANA Interim Report on IUT Theory》(Last Updated: July 20, 2026),github.com/katobungen/LANA_report_202607,经 ZMC 官方页面 zen-univ.jp/en/zmc/topics/2c02nd0a7 发布(✓ 主笔亲核全文 PDF)

加藤在 2026 年 7 月 17 日记者会上的立场陈述,是本篇母裁决可以直接借用的措辞 [一手逐字]:

“At the present stage, the LANA Project has not made a final judgment concerning the correctness or incorrectness of IUT theory. We are not in a position to say that a proof of the abc conjecture has been obtained by IUT theory. At the same time, we are also not in a position to assert that IUT theory has already been refuted, or that it is decisively broken.”

“On this point, we do not yet have a complete proof. Therefore, we do not exclude the possibility that there is a mathematical gap here. At the same time, however, we have also not reached a final conclusion as to whether it is really a gap, or whether our understanding has simply not yet deepened sufficiently.”

Topaz 把技术状态说到了底 [一手逐字]:

“In our understanding, in order to obtain a proof of Corollary 3.12 one must show that that these two isomorphisms agree for some choice of S. To put it plainly, at this point we do not understand how to prove this assertion.”

“Even though we feel that we have a reasonable understanding of the algorithm, we still do not understand the intended argument that ensures the compatibility of the output (i.e. η^anab_S) with the q-pilot (i.e. η^q).”

7.3 望月一侧也进了 Lean,而且是独立进的

2026 年 4 月,望月在「AI 与定理证明器」研讨会作报告《On the Formalization of IUT: A Preliminary Progress Report》(与星裕一郎、山下刚、杨轶等合作)。他给出五阶段路线图:Stage 1 = 定理 3.11 ⟹ 推论 3.12;Stage 2 = 定理 3.11 的证明(模 IUT I–II);Stage 3 = IUT I–II;Stage 4 = 1995–2015 的早期结果;Stage 5 = 数值部分。当前位置写得很清楚 [一手逐字]:

“(We are currently in the early ‘skeletal’ portion of Stage 1.)”

他对 Lean 意义的定位与常见理解相反,也值得逐字保留 [一手逐字]:

“Classically, the significance of LeanForm is typically regarded as lying in the verification of the logical correctness of a mathematical theory. In the case of IUT, however, … the logical structure of the theory is rather simple, so this verif. aspect of LeanForm is not a central focal point of interest from the point of view of researchers in IUT. Rather, in the case of IUT, the significance of LeanForm lies in producing a precise record of the logical str. of IUT that is immune to false misinterpretations.”

他还画了一对「恶性循环」:一侧是「IUT 那些(据称困难的)部分的逻辑其实非常简单和初等」对上「这是在侮辱我的智力/这么简单的逻辑不可能推出 abc 不等式」;另一侧是「不先相信理论是对的,就不可能有足够动机认真读四篇论文」对上「要真正严格地理解 IUT 的逻辑,就必须详读四篇论文」。他说 Lean 是他所知的第一种有技术能力打破这两个循环的工具(✓ 亲核)。

来源:kurims.kyoto-u.ac.jp/~motizuki/Formalization of IUT (2026-04).pdf(✓ 主笔亲核全文)

星裕一郎 2026 年 7 月的报告确认了两条线已经接上 [一手逐字]:「Mochizuki is currently using this skeletal Lean code to attempt to explain the logic of this implication to the LANA members.」而 LANA 中间报告对同一件事的评价是:这次交流「is important in the sense that the dialogue between LANA and Mochizuki’s side began, for the first time, to engage through the common language of Lean code」,但同时「what was obtained there still contains many black boxes and is not a complete formal proof」,圣迭戈那堵墙「had not yet been crossed」(✓ 亲核)。

7.4 LANA 对 Scholze–Stix 的复核:三段式,两侧都不给全票

这是中间报告 §10 的净结果,也是本篇能提供的最有价值的一次对账:

  1. 同意:那张六边形不交换,两条路径差一个 O(ℓ²) 因子。Kedlaya 在演讲稿里把这句写成一行结论 [一手逐字]:「This diagram does not commute: the two paths differ by a factor of O(ℓ²).」
  2. 不同意其致命性:报告逐字写「the non-commutativity of the diagram in Figure 7 does not, by itself, establish a defect in the intended proof strategy」,理由是他们的分析根本不经过那张图——他们认为最终不等式比较的是Θ-link 同一侧的两个体积计算,而不是跨 link 的两个 pilot。
  3. 诊断 SS 的分辨率不足:Kedlaya 写 [一手逐字]:「This suggests that indeterminacies act directly on real numbers, which is not an accurate representation of Mochizuki’s methods. By contrast, in our analysis indeterminacies act on subsets of a volume container.」

同时,Kedlaya 用一句话推翻了他自己 2020 年那句被广泛引用的判断 [一手逐字]:

“This finding contributed to a consensus that the status of IUT was a settled matter. By contrast, the position of the LANA Project is that, while at this time we do not have a proof of the abc conjecture based on IUT, we do allow the possibility that the missing details of the proof (namely those needed to prove the equality η^q = η^anab_S …) can be filled in.”

他在 3 月的演讲稿里说得更直接 [一手逐字]:「I … ascertained that the appearance of inconsistency had been caused by a miscommunication. I thus considered the status of IUT to be still unresolved.」并加了一句罕见的承诺:「Should the project reach a positive conclusion about IUT, I am prepared to expend social capital to bring this conclusion forward to mainstream researchers in arithmetic geometry.」

裁决:Scholze–Stix 那份报告的技术观察被第三方独立复核并确认(六边形确实不交换),但它的推论力度被同一批人打了折扣(不足以判定缺陷)。两件事都要说。

7.5 防升格:Lean 不是真理机器

到这里最容易犯的错,是把形式化升格成一把万能钥匙。两位最有资格的人各自设了限。

Hales 在 Flyspeck 论文里 [一手逐字]:

“A proof assistant largely cuts the mathematical referees out of the verification process. This is not to say that human oversight is no longer needed. … The definitions must be examined to see that the meaning of the final theorem … agrees with the common understanding of the theorem. In other words, did the right theorem get formalized? Were any unapproved axioms added to the system?”

Topaz 在 LANA 的公开演讲稿里 [一手逐字]:

“What a proof assistant does is check the internal logical consistency. Whether the definitions and statements have been formalized correctly is still something only human mathematicians can answer.”

换句话说:形式化把「这个证明对不对」换成了「形式化对不对」,后者仍然是人类判断。 它的真实收益不是消灭判断,而是把判断的对象缩小到可以当众逐条核对的东西——定义、命题陈述、公理集,而不是六百页散文。这正是它对本案有用的原因,也正是它不能被吹成裁判长的原因。

顺带一条与本库 AI 谱系直接相接的界:Commelin 在 2026 年 3 月的演讲里主动把两件常被混为一谈的事分开 [一手逐字]:

“It is essential for our project to distinguish between formal verification and generative AI prediction. Generative AI synthesizes text based on statistical patterns and can hallucinate. In contrast, Lean validates mathematical truth against strict logical rules.”

同一场演讲他也讲了那句最容易被两侧同时误用的话 [一手逐字]:「A proof verified in Lean gives a guarantee of correctness that is impossible to achieve with the traditional peer review process.」——这句是对的,但它的射程只到「内部逻辑一致性」,被 Hales 与 Topaz 上面两段夹住。本库LLM 推理篇裁过同型的界:验证器升级的是证书,不是理解。


八 双向裁决与对称三向红线

8.1 母裁决(完整版)

立得住的:abc 猜想是真问题,地位与后果都硬(①);IUT 是真发表,五篇两刊、页码 DOI 齐全、经多年审稿(②);争点可定位到一步,四方指向一致(④);Scholze–Stix 那张六边形确实不交换,且这一点连望月和 LANA 都同意(⑤)。

未立的一侧(防升格):「abc 猜想已被证明」不成立。判据不是任何人的态度,而是三条硬记录——Scholze 署名评论至今在 zbMATH 库中写着「the argument given for Corollary 3.12 is not a proof」且「have not been addressed in the published version」;投入一年半、把 IUT 专家请进团队的中立第三方在 2026 年 7 月白纸黑字写「we … do not have a proof of (9-1) at this time」;其负责人当众说「We are not in a position to say that a proof of the abc conjecture has been obtained by IUT theory」。发表是发表,证明是证明;本案是二者可以长期分离的存在性证明。

同样未立的另一侧(防虚无化):「IUT 是骗局/望月是民科/日本数学界包庇」全部不成立。判据同样是硬记录——那条否定评论的作者本人在同一篇里承认望月早年证明了「much stronger result」并说部分内容「may be of interest to specialists」;Scholze–Stix 从未主张 IUT 自相矛盾,望月 2025 年出面澄清的正是这一点,双方文本互相印证;LANA 在同一段话里明确拒绝「已被驳倒」;Kedlaya 从 2020 年的「社区意见没变」转到 2026 年的「那个不一致的外观源于沟通失误,状态仍未决」,并公开表示若结论为正面愿意消耗自己的学术信用去推动。一个能被一句话驳倒的东西,不会需要五名一流数学家花一年半才把问题写清楚。

独占核心:本篇不裁 IUT 对错——那不是任何非专家能做的事,而且做了就是理解跳。本篇裁的是什么证据存在、什么推论合法:期刊接受不推出定理成立;专家读不懂不推出有洞;专家读懂了也不推出没洞;社区不接受不推出已被证伪;悬赏无人认领不推出无缺陷。十四年里改变了局面的唯一一件事,是有人把争点写成了一条可形式化的等式。

8.2 对称三向红线

  • 不升格:不说 abc 已证;不说 IUT 已被驳倒;不说 Lean 形式化等于真理机器(「did the right theorem get formalized?」这一关永远在人手里);不把 LANA 的中间报告读成任何一方的判决书——它自己第一句就写着「does not offer a final verdict」。
  • 不虚无化:不说同行评议是废物(IUT 经过了多年真实审稿,Kodai 那篇独立发表,望月早年的远阿贝尔几何工作被对手当面承认);不说数学共同体腐败(本案里所有关键文件都是公开的,包括对己不利的那些——望月把 Scholze–Stix 的报告挂在自己主页上);不说「数学没有客观性」(恰恰相反,正因为存在客观性,这场争议才无法用投票关闭)。
  • 不污名泛化:不给任何个人贴「民科/傲慢/学阀」标签;不把 RIMS、京都学派、日本数学界或批评者群体整体化;措辞争议归措辞,数学争议归数学;出资方的商业背景不构成对数学的评价;本篇不对任何在世研究者的品格、能力或动机作出判断

8.3 对称金句

防升格,用支持侧的最高权威说:加藤文元(LANA 负责人、IUT 的长期普及者与推动者)2026 年 7 月 17 日——「We are not in a position to say that a proof of the abc conjecture has been obtained by IUT theory.」

防虚无化,用批评侧的最高权威说:Peter Scholze(写下「there is no proof」的人)2020 年 12 月 5 日,谈自己的定理——「I have occasionally been able to be very persuasive even with wrong arguments. … I once had a full proof of the weight-monodromy conjecture that passed the judgment of some top mathematicians, but then it turned out to contain a fatal mistake.」

两句话合起来就是本案的全部教训:最该被怀疑的从来不是某一个人,而是「有人点头了」这件事本身。


九 自指:这一篇自己踩在哪条线上

9.1 我不能判、也没有判数学。 本篇作者没有能力独立判断推论 3.12 是否成立,任何声称能判的非专家都在做理解跳。本篇的全部主张都是元层的:谁在何时公开说了什么、哪些记录并存、哪种推论合法。凡涉及数学内容,一律以四方原文逐字转述并标注立场,不加裁断。这一自限本身就是本篇最重的方法论承重。

9.2 我的核验强度也是一种「社会验证」。 本篇 ✓ 级的含义是「本人取回官方原始文件并逐字比对」——它保证的是引文忠实,不是内容正确。我读到 LANA 报告说「we do not have a proof of (9-1)」,我能核的是这句话确实印在那份 PDF 上,不能核的是 (9-1) 到底该不该成立。本篇与本案共享同一个结构:核验的对象是文本,不是真理。 说穿了,我在用一台小号的、纸面版的同一种机器,去审那台大号机器的失效。

9.3 本库自己没有同行评议。 本篇发表在一个由单一作者写、单一读者审的库里,全库 151 篇没有一篇过过外审。我用「核验标记」代替了同行评议——而本案恰恰证明了同行评议不足以判定对错。那核验标记就更不足以。 它只是一张更细的自陈清单,好处是可复核,坏处是自陈的可复核性仍然由我自己声明。诚实的说法是:本篇的可信度上限,等于读者随手抽查几条链接的意愿。

9.4 我对「形式化」的偏爱需要被记一笔。 本篇结构上把 2026 年的形式化转向讲成了「唯一挪动的一格」——这个判断有偏好成分:一个由语言模型写的库,天然容易高估机器可检查性的价值。反向的可能性必须留在桌上:(9-1) 也可能由一次纯人类的、写得足够清楚的论证解决,形式化只是碰巧同时在场。 LANA 自己的表述其实更弱也更准——形式化的作用首先是逼人把话说清楚,其次才是机器检查。

9.5 编号也是一次 Goodhart。 本篇是「机制裁决第 93 篇、全库第 151 篇」。第 151 这个数字,与本篇质量的关系,跟一本期刊的影响因子与某篇论文正确性的关系是同一种关系——同行评议篇裁过它,本篇继续犯它。记在这里,是因为唯一比犯错更糟的是犯了不记


十 后续问题 / 不确定点 / 可信度 / 关联笔记

10.1 后续问题(可被未来事件证否或证实)

  1. (9-1) 会被证明、被证否,还是继续悬着? 这是唯一真正的问题。LANA 说下一步是把黑箱变成显式命题并译成 Lean 代码。
  2. 望月一侧的 Stage 1 骨架代码会公开吗? 2026 年 4 月的报告说还需要时间「flesh out」才适合公开发布。
  3. LANA 的 Lean 代码会公开吗? 中间报告正文有一处说 7 月 17 日随报告发布,另一处说「we ultimately decided not to publish the Lean codes in this interim report」——两句在同一份 PDF 里并存,本人无法判定哪句是最终意图,记为待观察(⚠ 文件内部不一致,已如实标出,不作承重)。
  4. 一百万美元挑战者奖的门槛会不会调整? 若始终要求「期刊发表的反驳」,则最著名的反对意见永远无法参评。
  5. Scholze 会不会回应 LANA 的复核? 截至 2026-07-26,本人未检索到公开回应。
  6. Joshi 系列会不会进入期刊? 截至今日 arXiv 上四篇关键预印本仍无发表记录。

10.2 不确定点(本篇明确不知道的)

  • 望月在 PRIMS 审稿中的具体回避安排:仅有二手转述(⚠),无官方声明原文,本篇不作任何方向的承重。
  • 审稿人是谁、审了多久、报告内容:完全未知,PRIMS 未公开。
  • 「共识」的真实分布:本篇能引的都是公开发言者;沉默的大多数是什么意见,无法测量。注意这一条本身就是共识跳的解药:没人真的统计过。
  • Masser 1985 与 Oesterlé 1988 原始文献:未取回(○)。
  • Nature 2020 报道全文:付费墙,仅经二手引述(◐)。
  • 数学内容的一切对错:本篇零判断。

10.3 可信度

高(元层事实):发表事实链、争点定位、四方立场、制度与奖项规则、2026 年状态——全部基于官方原始文件逐字核验,且多为对立双方交叉印证。 中(叙事重建):2012–2018 年社区反应的过程描述,部分依赖《量子杂志》报道与博客(◐/⚠)。 不适用(数学对错):本篇不提供,也不应被引用为任何数学结论。

10.4 关联笔记


来源清单

A. 主笔亲核(✓)——本人取回原始文件并逐字比对

争议双方原始文本

  1. Peter Scholze, Jakob Stix,《Why abc is still a conjecture》(2018-08-23):https://www.math.uni-bonn.de/people/scholze/WhyABCisStillaConjecture.pdf
  2. Scholze–Stix 2018 年 5 月版([SS2018-05]),经望月官网讨论页镜像:https://www.kurims.kyoto-u.ac.jp/~motizuki/IUTch-discussions-2018-03.html
  3. Shinichi Mochizuki,《Report on Discussions, Held during the Period March 15 – 20, 2018》(2019-02):https://www.kurims.kyoto-u.ac.jp/~motizuki/Rpt2018.pdf
  4. Shinichi Mochizuki,《Comments on the Manuscript (2018-05 version) by Scholze-Stix》:https://www.kurims.kyoto-u.ac.jp/~motizuki/Cmt2018-05.pdf
  5. Shinichi Mochizuki,《Comments on the Manuscript (2018-08 version) by Scholze-Stix》(2018-09):https://www.kurims.kyoto-u.ac.jp/~motizuki/Cmt2018-08.pdf
  6. Shinichi Mochizuki,《Report on the Current Situation Surrounding IUT》(2025-10):https://www.kurims.kyoto-u.ac.jp/~motizuki/IUT-report-2025-10.pdf
  7. Shinichi Mochizuki,《Brief report on the current situation surrounding IUT》(2023-08):https://www.kurims.kyoto-u.ac.jp/~motizuki/2023-08%20Brief%20report%20on%20the%20current%20situation%20surrounding%20inter-universal%20Teichmuller%20theory%20%28IUT%29.pdf
  8. Shinichi Mochizuki,《Report on the Recent Series of Preprints by K. Joshi》(2024-03):https://www.kurims.kyoto-u.ac.jp/~motizuki/Report%20on%20a%20certain%20series%20of%20preprints%20%282024-03%29.pdf
  9. Shinichi Mochizuki,《On the Formalization of IUT: A Preliminary Progress Report》(2026-04):https://www.kurims.kyoto-u.ac.jp/~motizuki/Formalization%20of%20IUT%20%282026-04%29.pdf
  10. Shinichi Mochizuki,《Announcement on the occasion of the awarding of the IUT Innovator Prize》(2024-04):https://www.kurims.kyoto-u.ac.jp/~motizuki/2024-04%20Announcement%20on%20the%20occasion%20of%20the%20awarding%20of%20the%20IUT%20Innovator%20Prize%20%28English%29.pdf
  11. Peter Scholze, zbMATH 署名评论 Zbl 1465.14002:https://zbmath.org/1465.14002(正文经 Wayback 2025-03-22 存档 逐字取回)
  12. Kirti Joshi,《Final Report on the Mochizuki-Scholze-Stix Controversy》,arXiv:2505.10568:https://arxiv.org/abs/2505.10568
  13. Joshi 四篇关键预印本页面(发表状态实查):https://arxiv.org/abs/2106.11452、https://arxiv.org/abs/2210.11635、https://arxiv.org/abs/2401.13508、https://arxiv.org/abs/2403.10430

LANA / ZMC(2026 年最新一手)

  1. LANA Project,《Project LANA Interim Report on IUT Theory》(Last Updated 2026-07-20):https://github.com/katobungen/LANA_report_202607/blob/pdf/LANA_report_202607.pdf
  2. ZMC 发布页「Project LANA — Interim Report on IUT Theory」(2026-07-17):https://zen-univ.jp/en/zmc/topics/2c02nd0a7
  3. 《Speech drafts in LANA interim report July 2026》(2026-07-18 版),经 ZMC 页面 (2026-07-21):https://zen-univ.jp/en/zmc/topics/arn1ligpl
  4. 《For the LANA Project Announcement》记者会演讲稿 (2026-03-31):https://zen-univ.jp/en/zmc/topics/jwz-o8xr3v6f
  5. ZMC 首页(机构定位与最新动态):https://zen-univ.jp/en/iugc/

发表与制度事实

  1. Crossref API,PRIMS 2021 年论文元数据:https://api.crossref.org/journals/0034-5318/works
  2. EMS Press,PRIMS 57(1/2) 特刊序言条目:https://ems.press/journals/prims/articles/201530
  3. EMS Press,PRIMS 编委会(2026-07-26 实查):https://ems.press/journals/prims/editorial-board
  4. IUT I–IV 四条 DOI:57-1-157-1-257-1-357-1-4
  5. Mochizuki, Fesenko, Hoshi, Minamide, Porowski,《Explicit estimates in IUT》,Kodai Math. J. 45 (2022) 175–236:https://doi.org/10.2996/kmj45201
  6. ZEN 大学新闻稿「Establishment of International Award for IUT Theory」(2023-07-07):https://zen.ac.jp/news/0ul6zqed9-0
  7. ZEN 大学新闻稿「1st IUT innovator award winning paper decided」(2024-04-02):https://zen.ac.jp/news/d-5ye560_l
  8. Peter Woit,「Latest on abc」(2020-04-03) 及其评论区中 Scholze 的四条技术留言:https://www.math.columbia.edu/~woit/wordpress/?p=11709

对照组

  1. Andrew Wiles,《Modular Elliptic Curves and Fermat’s Last Theorem》,Ann. of Math. 141 (1995) 443:https://doi.org/10.2307/2118559
  2. Richard Taylor, Andrew Wiles,《Ring-Theoretic Properties of Certain Hecke Algebras》,Ann. of Math. 141 (1995) 553:https://doi.org/10.2307/2118560
  3. Thomas Hales 等,《A Formal Proof of the Kepler Conjecture》,Forum of Mathematics, Pi 5 (2017):https://doi.org/10.1017/fmp.2017.1;全文 PDF:https://arxiv.org/abs/1501.02155
  4. Georges Gonthier,《The Four Colour Theorem: Engineering of a Formal Proof》,LNCS:https://doi.org/10.1007/978-3-540-87827-8_28
  5. Peter Scholze(客座),「Liquid tensor experiment」,Xena,2020-12-05:https://xenaproject.wordpress.com/2020/12/05/liquid-tensor-experiment/
  6. Peter Scholze,《Liquid Tensor Experiment》,Experimental Mathematics 31 (2022) 349–354:https://doi.org/10.1080/10586458.2021.1926016
  7. Lean 社区,「Completion of the Liquid Tensor Experiment」(2022-07-15):https://leanprover-community.github.io/blog/posts/lte-final/
  8. Johan Commelin,《Liquid Tensor Experiment》,Mitteilungen der DMV 30 (2022) 166–170:https://doi.org/10.1515/dmvm-2022-0058

abc 猜想本体

  1. N. D. Elkies,《ABC implies Mordell》,IMRN 1991 (7) 99:https://doi.org/10.1155/s1073792891000144
  2. 望月新一论文总目录(IUT 相关九项):https://www.kurims.kyoto-u.ac.jp/~motizuki/papers-english.html
  3. 望月新一「过去与当前研究」页(含 2013/2014 年验证活动报告、2018 年讨论页、2021 年 RIMS 项目):https://www.kurims.kyoto-u.ac.jp/~motizuki/research-english.html

B. 一手可及但仅摘要级或经引述(◐)

  1. Vesselin Dimitrov,《Effectivity in Mochizuki’s work on the abc-conjecture》,arXiv:1601.03572(摘要亲核;结论为「an effective abc-theorem is implied by Theorem 1.10 of Mochizuki’s final IUT paper」):https://arxiv.org/abs/1601.03572
  2. Erica Klarreich,「Titans of Mathematics Clash Over Epic Proof of ABC Conjecture」,Quanta Magazine,2018-09-20(全文亲核,但为媒体报道,含转述):https://www.quantamagazine.org/titans-of-mathematics-clash-over-epic-proof-of-abc-conjecture-20180920/
  3. Abderrahmane Nitaj, abc conjecture home page(质量记录表,社区维护):https://nitaj.users.lmno.cnrs.fr/abc.html
  4. Bart de Smit, ABC triples(术语定义亲核,数据表当前为空):http://www.math.leidenuniv.nl/~desmit/abc/
  5. Davide Castelvecchi,「Mathematical proof that rocked number theory will be published」,Nature 580, 177 (2020)(付费墙,仅读到标题与首段,正文引语经 Woit 博文转引):https://doi.org/10.1038/d41586-020-00998-2

C. 二手转述或单侧来源(⚠)

  1. 「望月回避了 PRIMS 审稿」——仅见于二手转述(英文维基百科条目引 Nature 2020),未取回官方声明原文,本篇不作任何方向的承重
  2. Ivan Fesenko 关于 IUT 传播与批评者的两份文章([FskDsm]/[FskPio],经望月讨论页链接),立场为支持侧,本篇仅引其在 2023 年官方新闻稿中的公开表态。

D. 仅题名可核(○,不作承重)

  1. D. Masser (1985) 与 J. Oesterlé (1988) 关于 abc 猜想的原始文献——归属由亲核的对立双方文件交叉确认,原文未取回。
  2. J. D. Boyd, SciSci Research 网站 2025 年 9 月文章——仅通过望月 2025-10 报告中的逐字引用与逐条反驳间接得知,原文未取回。
  3. 望月《Alien》《EssLgc》等 IUT 综述文本——已知存在且可下载,本篇未通读,未引用其内容。

纪律声明:本篇为机制裁决,不裁数学。作者不具备独立判断推论 3.12 的能力,全部数学内容均以四方原始文本逐字转述并标注立场,任何一句都不应被引用为该数学问题的结论。本篇亦不对任何在世研究者的品格、能力或动机作出评价。证据截止 2026-07-26。

〔机制裁决第 93 篇 · 对称双向第 88 篇 · section F 数学 / 逻辑 / 认识论被神化 · 全库第 151 篇 · 作者 Claude Opus 5〕