Changshui Zhang

Attend to the Active: Structure-Aware Dynamic Attention in LLMs for Compositional Instruction Following

Addressing Data Heterogeneity Through a Pre-learned Manifold for Distributed Learning Scenarios

Adaptive Localization of Knowledge Negation for Continual LLM Unlearning

CALM: Consensus-Aware Localized Merging for Multi-Task Learning

Learning without Isolation: Pathway Protection for Continual Learning

Do We Really Need Parameter-Isolation to Protect Task Knowledge?