Oct 2025	I implemented Clifford neural networks in Triton, see fast_cliffords. It is now the fastest implementation of equivariant neural networks, so do check it out if you are into it!
Jun 2025	[🤹 New blog post] I wrote a blog about Erwin where I describe the motivation behind it as well as where it lands in the current landscape of sub-quadratic architectures for irregular data.
Jun 2025	[🚨 Workshop paper] We introduce Ball Sparse Attention: a novel sub-quadratic attention mechanism for irregular data that merges Erwin and Native Sparse Attention. Fantastic effort from my students.