Topic
#continual-learning
2 articles exploring continual-learning. Expert insights and analysis from our editorial team.
Showing 1–2 of 2 articles
Articles
Newest first
Models & Research
Learning, Fast and Slow: What arXiv 2605.12484 Proposes for LLMs That Adapt Continually
Fast-Slow Training splits LLM updates into prompt fast weights and parametric slow weights, cutting KL drift by 70% and lifting sample efficiency by 3×, keeping plasticity.
Models & Research
JumpLoRA's Sparse Adapters Break the Assumption That Continual Fine-Tuning Requires Full-Rank LoRA Stacks
JumpLoRA adds learnable JumpReLU gates to LoRA blocks for 87-95% sparse adapters with near-zero cross-task overlap. The work exposes that PEFT has no router for continual.