AI × Mathematics 2026 was a week-long interdisciplinary residency for students and researchers interested in how artificial intelligence can support genuine mathematical discovery. The programme introduced the emerging AI pipeline for mathematics—conjecture generation, auto-formalisation, and automated theorem proving—showing how modern neuro-symbolic systems blend deep learning with formal reasoning to propose, refine, and prove nontrivial mathematical results.
Participants learned how graph neural networks, transformers, and proof-search engines interact within current reasoning systems, gaining hands-on understanding of the tools now shaping research in number theory, arithmetic and algebraic geometry, and related fields. The workshop also explored the physical and geometric principles underlying learning itself: energy-based models, optimal-transport flows, and equivariant architectures inspired by ideas from physics. These perspectives revealed how insights from physics and geometry can guide the design of interpretable and generalisable AI systems tailored for mathematical discovery.
Abstract 
