一篇新论文认为，微调机制，特别是可训练参数子空间，是评估持续学习方法的一个关键变量。研究人员发现，像EWC、LwF、SI和GEM等标准持续学习方法的相对性能排名，会根据所选的微调深度而发生显著变化。更深的适应机制与遗忘增加有关，这表明当前的评估协议可能在不同的微调设置下不够稳健。 AI
影响 强调了在持续学习研究中需要有意识地进行机制评估的协议，这可能会影响未来方法的基准测试方式。
排序理由 在arXiv上发表的学术论文，讨论了一种新颖的持续学习评估方法。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2604.21927v2 Announce Type: replace Abstract: Continual learning (CL) studies how models acquire tasks sequentially while retaining previously learned knowledge. Despite substantial progress in benchmarking CL methods, comparative evaluations typically keep the fine-tuning …
Continual learning (CL) studies how models acquire tasks sequentially while retaining previously learned knowledge. Despite substantial progress in benchmarking CL methods, comparative evaluations typically keep the fine-tuning regime fixed. In this paper, we argue that the fine-…
Continual learning (CL) studies how models acquire tasks sequentially while retaining previously learned knowledge. Despite substantial progress in benchmarking CL methods, comparative evaluations typically keep the fine-tuning regime fixed. In this paper, we argue that the fine-…