School of Computing Science

The novelty of this ECML/PKDD 2026 paper lies in its mutual head distillation framework, which balances client-specific personalization and global generalization by coupling a global semantic head with lightweight personalized heads in federated learning.


First published: 28 May 2026