Sign error forces OpenAI to retract 3 math papers two days after publishing 722
机器之心 · wechat · 2026-10-08
Two days after publishing its openai/math repository with 722 manuscripts, OpenAI issued its first changelog: 3 papers retracted, 14 revised, 13 citations updated (722 → 719).
- All three retractions relate to the Hodge conjecture. One paper wrote +1 where its own conventions required −1, turning a count that should cancel to zero into a nonzero value, invalidating a classical theorem the construction relied on; two papers built on it.
- OpenAI stresses the proofs—not the conjectures—are withdrawn; originals remain accessible.
- 14 revisions span statistical physics, complex geometry, symplectic geometry and Navier-Stokes; one over-strong side claim on a Donaldson conjecture was narrowed.
- 300 of 719 main results (42%) now have Lean formalizations. Notably, retracted and substantively revised work falls almost entirely in the unformalized portion.
The corrections followed good practice—errors documented, dependency chains explained, versions preserved—but this changelog is unlikely to be the last.
More from Research
- Study finds female AI agents 'paid' less than male ones, extending gender bias to agents — TMWNN · 2026-10-09
- LightOnOCR-3 training: using OCR reference text to guide logical block grouping — IgorCarron · 2026-10-09
- No single layout model works everywhere: LightOnOCR-3 team on combining complementary sources — IgorCarron · 2026-10-09
- Text2Sim turns text prompts into editable physics-grounded videos and sim setups — erwincoumans · 2026-10-09
- Building Rome from a single image: new method reconstructs 3D scenes with geometry beyond the visible view — jonstephens85 · 2026-10-09
- PredActor unifies humanoid motion generation and control in one diffusion policy at 50Hz on Unitree G1 — carlosdponx · 2026-10-09