Mathematicians Unveil 50 High-Stakes Problems Designed for AI-Verifiable Solutions
skdh · x · 2026-10-02
At Epoch AI's request, mathematicians held workshops in six cities (London, Toronto, LA, NYC, Berkeley, Cambridge) to compile a new list of 50 pressing open math problems.
- Key constraint: each problem must have automatically checkable solutions, so AI-generated answers can be objectively verified.
- Context: AI models from various companies have recently been cutting through many of math's biggest challenges, and Epoch AI benchmarks this progress.
- Yang-Hui He of the London Institute for Mathematical Sciences submitted three problems spanning knot theory, algebra, topology, and number theory.
More from Research
- DyRAD introduces first off-path evaluation protocol for radar novel view synthesis — orlitany · 2026-10-02
- EMBL-EBI's saezlab open-sources Karenina, a framework for multi-dimensional biomedical AI evaluation — anshulkundaje · 2026-10-02
- Nature Biotech: calibrated metrics show deep learning perturbation models can beat baselines — VectorInst · 2026-10-02
- arXiv papers grew 17% longer over the past decade, and AI may stretch them further — luismbat · 2026-10-02
- Suffix Cache Reuse: fixing KV cache for agents that edit context in place — RulinShao · 2026-10-02
- Satirical dialogue skewers RL environments: nobody actually reads the thousands of tasks and rubrics — samiramanabi · 2026-10-02