arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2610.22293cs.LGcs.AI
收藏

面向持续学习:通过子空间调优缓解遗忘

Towards Continual Learning: Mitigating Forgetting With Subspace Tuning

Arnav Chavan, Kusha Sareen, Rishabh Tiwari, Abhranil Chandra, Johan Obando-Ceron, Kurt Keutzer, Scott Niekum, Inderjit S Dhillon, Devvrit Khatri

AI总结:

本研究针对大语言模型持续学习的灾难性遗忘问题,提出子空间调优(SST)方法,通过限制权重更新至底部奇异子空间,在保护已有能力的同时提升目标任务性能,推进了学习-遗忘帕累托前沿。

摘要:

大语言模型(LLMs)通过微调获取新技能,但会破坏已有的能力:灾难性遗忘仍是LLMs持续学习的核心障碍。近期研究将遗忘与微调移动权重的幅度及方向关联起来。强化学习、在线蒸馏等在线策略方法会避免更新权重矩阵的主奇异方向,因此遗忘更少。相比之下,监督微调(SFT)将更新集中在这些方向,易受灾难性遗忘影响。本研究提出子空间调优(SST),将每个权重更新限制在底部奇异子空间,消除主方向上的变化。SST与优化器无关,可与SFT、自蒸馏等微调方案良好组合。它结合了紧密的子空间保护、密集梯度估计和简单的刷新机制,比现有正交微调方法实现更强的学习效果。通过这些机制,保护多达99%的权重子空间可大幅提升保留能力,同时保持目标任务的强性能。该效果随模型和数据规模增强,在大模型场景下几乎消除了遗忘。SST还使模型接近其基础分布,比基准方案诱导的KL漂移少2-4倍。总体而言,SST推进了学习-遗忘帕累托前沿,使其成为微调预训练语言模型的实用默认方案。

英文摘要:

Large language models (LLMs) acquire new skills through fine-tuning, but it can destroy capabilities they already have: catastrophic forgetting remains a central obstacle to continual learning in LLMs. Recent works connect forgetting to how far, and along which directions, fine-tuning moves the weights. On-policy methods such as reinforcement learning and on-policy distillation avoid updating the principal singular directions of weight matrices and thus forget less. In contrast, SFT concentrates updates along these directions and is susceptible to catastrophic forgetting. In this work, we introduce Subspace Tuning (SST), which restricts each weight update to the bottom singular subspace, eliminating changes along the principal directions. SST is optimizer-agnostic and composes well with fine-tuning recipes including SFT and self-distillation. SST combines tight subspace protection, dense gradient estimation, and a simple refresh mechanism, enabling stronger learning than prior orthogonal fine-tuning methods. With these mechanisms, protecting as much as 99\% of the weight subspace substantially improves retention while preserving strong target-task performance. This behavior strengthens with model and data size and nearly eliminates forgetting in the large-model regime. SST also keeps the model close to its base distribution, inducing $2$--$4\times$ less KL drift than the baseline recipes. Overall, SST advances the learning--forgetting Pareto frontier, making it a practical default for fine-tuning pre-trained language models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑