Continual Representation Learning (CoRe) moves PEFT from weight-level updates to representation-space interventions, solving catastrophic forgetting in dynamic environments.
arXiv · March 13, 2026 · 2603.11201
Why it matters
Instead of fine-tuning model weights (which drifts and erases old knowledge), CoRe learns low-rank linear transformations of hidden states. This provides explicit control over representation drift, allowing models to learn new tasks without degrading performance on previously learned ones, a major hurdle for production AI agents.
From the abstract
The world is inherently dynamic, and continual learning aims to enable models to adapt to ever-evolving data streams. While pre-trained models have shown powerful performance in continual learning, they still require finetuning to adapt effectively to downstream tasks. However, prevailing Parameter-Efficient Fine-Tuning (PEFT) methods operate through empirical, black-box optimization at the weight level. These approaches lack explicit control over representation drift, leading to sensitivity to