Urgent.News

What's breaking now, across hundreds of outlets.

AI

OpenAI Withdraws 3 Math Papers

Article URL: https://github.com/openai/math/blob/main/history.md Comments URL: https://news.ycombinator.com/item?id=50003107 Points: 271 # Comments: 218

OpenAI has withdrawn three math papers due to significant errors in their content. The withdrawn papers are "Algebraicity of Weil classes on split abelian eightfolds" and two other papers that relied on the flawed work. The error was a sign error in a stabilization-trace cancellation argument, which invalidated the entire construction in the withdrawn paper as well as the two dependent papers.

To rectify the situation, OpenAI has revised 14 other papers, correcting proof errors, clarifying hypotheses and dependencies, and providing more transparent citations to archived manuscripts. Additionally, they have updated 13 other papers to cite the revised editions of companion papers, ensuring that their version dates reflect the changes made.

These revisions contribute to OpenAI's ongoing efforts to formalize top-line results. Currently, 300 out of the 719 top-line results have been formalized, representing approximately 42% of their total work. Furthermore, OpenAI has added six formalizations and five other additions that cover supporting results, further enhancing the accuracy and completeness of their mathematical work.

Rewritten by urgent.news; may contain errors — check the original.

Source: github.com

This story

This is one outlet's version. Read the fullest account.

Every outlet that reported this, and when →

Read the original at github.com →

Report a problem with this story

More in AI

Deterministic AI Agents in Mission-Critical Systems: Why Microfrontend Boundaries Matter

You've deployed an LLM agent to production to automate customer workflows. It's trained on your entire codebase, your entire state tree, your entire API surface. It works. Then it doesn't.

  • LLM agents in mission-critical systems risk hallucinations and shared state mutations
  • Microfrontend boundaries limit agent context to a single domain, reducing blast radius
  • Scoped agents provide predictability and easier testing in critical environments

More from Thursday 8 October →