VibeMathedMath problems solved with AI

The pure semisimplicity conjecture

A ring RR is right pure semisimple if every right RR-module is a direct sum of finitely generated modules. Every ring of finite representation type is both left and right pure semisimple. The pure semisimplicity conjecture asserts the converse from one side: every right pure semisimple ring is of finite representation type, equivalently every right pure semisimple ring is also left pure semisimple. It was known for commutative rings (Warfield), Artin algebras (Auslander) and rings with a polynomial identity (Herzog), and Herzog and Simson reduced it to hereditary triangular matrix rings over two division rings. Is it true?

Result
Disproved(see note)
Status
Candidate (review pending)
AI contribution
AI co-developed
Method
Construction
Field
Ring theory; representation theory of rings
Posed by
—
Year posed
—
Years open
—
Solved
2026-10-29
Model
GPT-6 Astra, GPT-6.1 Sol, Claude Opus 5.5
Vendor
OpenAI; Anthropic
Collaborators
Junpeng Ren
Verification
Lean-checked, statement unaudited
Publication
Preprint
Significance
40 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Over Q\mathbb Q with q=1/2q=1/2 (v1; in v2 any field element qq of infinite multiplicative order), quantum chains of division rings give an extension H⊆TH\subseteq T whose bimodule HTT{}_HT_T has right dimension sequence (1,2,2,… )(1,2,2,\dots) and dim⁡HT=∞\dim{}_HT=\infty. The triangular ring R=(HT0T)R=\begin{pmatrix}H&T\\0&T\end{pmatrix} is right pure semisimple but neither left pure semisimple nor of finite representation type, so the conjecture fails. Version 2 also gives pairwise Morita inequivalent counterexamples and hereditary artinian rings of every Coxeter type I2(m)\mathbb I_2(m), m≥4m\ge4.

What the AI did

From the paper's declaration of AI use (v1): GPT-6 Astra contributed to the exploration that led to the criterion used and to the proof of the main theorem, in particular the construction of the division rings TT and HrH_r. The author verified the proofs and completed the Lean 4 formalization and the editing with the assistance of GPT-6.1 Sol and Claude Opus 5.5. Version 2 adds Claude Fable 5.1 for the finite-order (Coxeter type) results, which are not part of this entry.

Verification

No independent mathematician has checked this yet. Lean-checked on this site's ladder, not Lean-verified. Checked here on 15 November 2026: arXiv 2610.39883 v1 and v2 (the TeX sources), including the conjecture as stated in the introduction and the declarations of AI use and formal verification. The Lean repository oasisss6666/pure-semisimplicity-lean at commit e4905ea (Lean v4.34.0 with Mathlib, 161 files, about 15,900 lines) formalizes the v1 instance (Q\mathbb Q, q=1/2q=1/2). The headline pure_semisimplicity_conjecture_fails and its definitions were read: right pure semisimple means every module over the opposite ring, in any universe, is isomorphic to a direct sum of finitely generated modules; finite representation type means finitely many isomorphism classes of finitely generated indecomposable right and left modules (the artinian clause is omitted, which only makes the negation stronger); indecomposable is nontrivial with no nontrivial complemented submodules. The statement is the negation of the conjecture as posed. A grep of all files found no sorry, admit, axiom declarations, native_decide, implemented_by, extern or unsafe. The author's log reports 832 theorems on propext, Classical.choice and Quot.sound only; not rebuilt here, and the repository has no CI. No Palomar or Formal Conjectures entry was found. Generalov and Tursunbaev (arXiv 2611.09749, 6 November 2026) independently give another counterexample, a week after this one.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion