works on
multi-task learning 1regularization 1reinforcement learning 1robotics 1world models 1

From the 1 of 1 linked paper with an AI index.

collaborators

1 paper

cs.LG2026

Temporally Centered SIGReg Improves Multi-Task LeWorldModel Learning: From Analysis to Method

Chang Liu, Fei Suo, Yanzhou Jin +3

The paper introduces a temporally centered version of the Sketched Isotropic Gaussian Regularizer (SIGReg) that regularizes residuals instead of latent marginals, improving stabili…