<ul><li>Continual Learning (CL) aims to enable neural networks to acquire new knowledge while retaining existing knowledge.</li><li>Adapting PTMs before the core CL process (ACL) is a novel framework proposed to enhance plasticity in neural networks.</li><li>ACL refines the PTM backbone through a plug-and-play adaptation phase before learning each new task, balancing stability and plasticity.</li><li>Extensive experiments show that ACL improves CL performance across benchmarks and integrated methods, providing a versatile solution for PTM-based CL.</li></ul>

Adapt before Continual Learning

Discover more