Riemannian Metric Matching paper shared with code, pretrained models and tutorials

The Riemannian Metric Matching team says it has put together code, pretrained models and tutorial notebooks to accompany the paper. Its announcement highlights an oral presentation at ICML in summer 2026 and the Best Short Paper Award at the ICLR GRAM workshop. A coauthor describes the work as a step toward neural diffusion geometry, while stressing that replacing kernel- and graph-based geometry with deep learning is a big challenge.