Researchers already improve on OpenAI's recent math results, verified in Lean
Eliv_nurotic · reddit · 2026-10-08
A Reddit post reports that researchers have already made significant improvements on OpenAI's recent math results, with the work verified in the Lean proof assistant and open-sourced on GitHub (CrocSwap/integer-mult-bounds). The author cites this as a rebuttal to the "we don't learn anything from these slop proofs" criticism of AI-generated proofs.
Related event: Researchers Improve OpenAI's Math Proof, Verified in Lean(2 posts)→
More from Research
- Are URM and Universal Transformers the forgotten architecture beating standard LLMs? — moschles · 2026-10-09
- Researcher: Use AI to Rewrite Machine-Generated Math Proofs Into Human-Readable Forms — jd_pressman · 2026-10-09
- AutoScientist's two-agent checklist loop auto-audits every training example — sarahookr · 2026-10-09
- NeurIPS GenAI4Health oral: retrieval-based medical fact-checking fails in ways bigger models can't fix — mdredze · 2026-10-09
- Bigger models, more reasoning, better sources won't fix medical fact-checking, researchers say — mdredze · 2026-10-09
- DEX best abstract: top LLMs catch many physician diagnostic errors, but big gaps remain — mdredze · 2026-10-09