GPT-5.6 Solves ~10% of 3,300 Open Math Problems
burny_tech · x · 2026-08-25
An experiment using GPT-5.6 Sol xhigh on 3,300 open mathematical problems resulted in roughly 170 counterexamples and 170 proofs, achieving a completion rate of around 10%. This demonstrates the current capabilities and limitations of frontier models in complex logical reasoning and mathematical proof.
Related event: GPT-5.6 Tackles 3,300 Open Math Problems, Completing About 10%(2 posts)→
More from Research
- Study analyzes AI disclosure statements in 13k+ papers, finds disconnect — danish037 · 2026-08-25
- Research Finds AI Use Policies Remain Vague and Under-Specified — danish037 · 2026-08-25
- Image orientation carries semantics in natural photos but not pathology images — iScienceLuvr · 2026-08-25
- Chris Manning notes SMT as joint optimization of memory-bottlenecked Transformers and RNNs — chrmanning · 2026-08-25
- Ant Group proposes a Densing Law for billion-scale user representation learning — antgroup · 2026-08-25
- WorldToken: Time-First Sequence Modeling for Robotic Imitation Learning — Chunkai Yang · 2026-08-25