We show that Kalman filters can be reparameterized for efficient parallel training and introduces GAUSS, a more expressive yet equally scalable state-space layer that outperforms …
We present a unified, algorithm-agnostic theory that defines forgetting as self-inconsistency in a learner’s predictive distribution—quantifying loss of predictive information—and …
Technical report detailing our method that won the 1X World Model Challenge 2025.
Introduction World models equip agents (e.g., humanoid robots) with internal simulators of their environments. By “imagining” the consequences of their actions, agents can plan, …