AI RESEARCH

Exemplar-Free Continual Learning for State Space Models

arXiv CS.LG

ArXi:2505.18604v2 Announce Type: replace State-Space Models (SSMs) excel at capturing long-range dependencies with structured recurrence, making them well-suited for sequence modeling. However, their evolving internal states pose challenges in adapting them under Continual Learning (CL). This is particularly difficult in exemplar-free settings, where the absence of prior data leaves updates to the dynamic SSM states unconstrained, resulting in catastrophic forgetting.