Two papers accepted in International Conference on Machine Learning (ICML) 2025

—

published in

Improving Continual Learning Performance and Efficiency with Auxiliary Classifiers
No Task Left Behind: Isotropic Model Merging with Common and Task-Specific Subspaces