Loda

New methods tackle continual learning in LLMs by separating task-specific and shared knowledge

Two new research papers propose novel methods for continual learning in large language models, addressing the challenge of acquiring new knowledge without forgetting previous information. The first paper, "Split-on-Shar…