OpenAI internal models crack open math problems where humans scored near zero

OpenAI researcher Steven Heidel shared an internal chart showing a significant jump in solve rates on open math problems where human effective scores were previously near zero.

2026-09-09 ~ 2026-09-09 · 2 related posts

• OpenAI's Internal Model Cracks Open Math Problems Where Humanity Scored Effectively Zero — stevenheidel · 2026-09-09

1 near-duplicate retellings: stevenheidel