两篇新的研究论文介绍了改进AI模型持续学习能力的新颖方法。E$^2$-LoRA 专注于在领先的秩中集中和排序知识，以释放未来任务的容量，并采用动态秩分配策略。Janus-LoRA 通过使用梯度校正来强制执行正交性以及解耦的边距损失来进行特征分离，从而解决稳定-塑性权衡问题，旨在防止灾难性遗忘并增强学习。 AI
影响 持续学习方面的这些进步可能带来更高效、更有能力的AI系统，这些系统可以在不忘记先前知识的情况下学习新信息。
排序理由 两篇在arXiv上发表的学术论文，介绍了持续学习的新方法。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2605.27482v1 Announce Type: cross Abstract: While orthogonal subspace methods try to mitigate task interference in Continual Learning (CL), they often suffer from energy diffusion across the basis, hindering knowledge compaction and exhausting capacity for future tasks. We …
arXiv:2605.28495v1 Announce Type: new Abstract: Low-Rank Adaptation (LoRA) has emerged as a promising paradigm for Continual Learning. It independently updates its low-rank factors ($A$ and $B$), creating a composite update to the full weight matrix through their interaction. To …
Low-Rank Adaptation (LoRA) has emerged as a promising paradigm for Continual Learning. It independently updates its low-rank factors ($A$ and $B$), creating a composite update to the full weight matrix through their interaction. To prevent catastrophic forgetting, this update sho…