发现大模型持续学习中性能下降实为任务对齐失效，非知识丢失。
Spurious Forgetting in Continual Learning of Language Models
- 提出'虚假遗忘'概念，指出性能下降多因任务对齐破坏。
- 实验显示新任务初期优化会打乱原有对齐关系。
- 冻结底层参数可显著提升四种场景下的持续学习效果。
大型语言模型在持续学习中出现令人困惑的现象：尽管训练充分，性能仍显著下降，引发对任务对齐与知识保留的质疑。本研究首次探讨'虚假遗忘'概念，提出此类性能下降往往反映任务对齐的削弱，而非真实知识损失。通过合成数据集的受控实验，我们研究了新任务初始训练阶段的模型表现动态，发现早期优化步骤会破坏先前建立的任务对齐。理论分析将这些变化归因于模型权重的正交更新，提供理解该行为的稳健框架。最终，我们提出一种冻结策略，固定模型底层参数，在四种持续学习场景中均取得显著性能提升。研究强调任务对齐与知识保留之间的关键区别，为更有效的持续学习策略铺平道路。
原文摘要 · Abstract (English)
Recent advancements in large language models (LLMs) reveal a perplexing phenomenon in continual learning: despite extensive training, models experience significant performance declines, raising questions about task alignment and underlying knowledge retention. This study first explores the concept of "spurious forgetting", proposing that such performance drops often reflect a decline in task alignment rather than true knowledge loss. Through controlled experiments with a synthesized dataset, we investigate the dynamics of model performance during the initial training phases of new tasks, discovering that early optimization steps can disrupt previously established task alignments. Our theoretical analysis connects these shifts to orthogonal updates in model weights, providing a robust framework for understanding this behavior. Ultimately, we introduce a Freezing strategy that fix the bottom layers of the model, leading to substantial improvements in four continual learning scenarios. Our findings underscore the critical distinction between task alignment and knowledge retention, paving the way for more effective strategies in continual learning.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。