Did you know,  Geometric inductive bias beats scale!

Smaller geometric models can outperform larger generative models when symmetries (invariance, equivariance,) and structure (shape of data, topological features, ) are correctly encoded as priors.

tinyurl.com/38md53zz