New paper led by my awesome grad student Yurii Kvasiuk. We get precise constraints on primordial non-Gaussianity from halos by machine learning two different fields. We can show optimality of the method under some conditions, and robustness to baryonic feedback. arxiv.org/abs/2410.01007

Thrilled to share our new work with Moritz Münchmeyer, Jordan Krywonos & Matt Johnson, accepted to #ML4PhysicalSciences at #NeurIPS! Our GNN reconstructs dark matter & electron density directly from galaxy catalogs. Now on arxiv: arxiv.org/abs/2411.02496

1/5 Our theoretical physics benchmark for AI, TPBench, is out! Check out the website tpbench.org for scores, paper and examples. Collaboration with Fred Sala Maja Rudolph TSC Yurii Kvasiuk Tianyi ZHIQI GAO UW-Madison Physics datascience@uw. Some thoughts.

Update to our theoretical physics reasoning benchmark, including more models. OpenAI is still ahead of the competition, especially at research level difficulty. More detailed results are on our website tpbench.org.