Please fill the form before April 10th to express your interest in participating. We have full funding for up to 15 external participants. You will be contacted by the organisers as soon as possible to confirm participation.

In this hands-on event we will collectively build research-level benchmarks to stay ahead of the rapidly advancing mathematical reasoning capabilities of AI models. The specific focus lies on topics surrounding nonlinear algebra. Participants will work with the ScienceBench platform (https://math.sciencebench.ai), formulating research level problems and running them against frontier LLMs.

Our goal is for you to learn how to use AI in your own math research, this event is organized for math researchers by math researchers. We believe that the scientific community should lead the discoveries around the capabilities of AI models.

The event will consist of practice sessions on prompting and benchmark design, followed by collaborative working sessions in which participants formulate, refine, and test their problems. We will also discuss the current landscape of AI capabilities in mathematics and discuss what makes a problem genuinely challenging for the current models.