A study of Transformer representation evolution decomposes learned additive updates into parallel and perpendicular components, finding that pretrained models preserve their current direction more than they redirect it. The research frames this evolution as a functional geometry across pretrained models.
Transformer representations evolve through learned additive transformations that either preserve their current direction or redirect it. We study this evolution as a functional geometry, decomposing learned updates into parallel and perpendicular components. Across pretrained models, we find substan
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.