重新利用对抗扰动进行持续学习：从防御到主动对齐
Repurposing Adversarial Perturbations for Continual Learning: From Defense to Active Alignment
AI总结：
提出AdvCL框架，通过将对抗扰动重新用作几何控制信号，结合三个即插即用模块（Intra-Smooth、Proto-Clip、Inter-Align），在持续学习中同时提升标准性能、鲁棒性、降低遗忘并增强迁移。
摘要：
在动态环境中，大型语言模型需要不断适应新任务，但持续学习常常遭受遗忘、有限的迁移以及对对抗扰动的脆弱性。为了解决这个问题，我们提出了AdvCL，它将对抗扰动重新用作稳定的持续适应的几何控制信号。AdvCL结合了三个即插即用模块：Intra-Smooth通过小的对抗扰动促进局部平滑性；Proto-Clip使用相似性裁剪以防止过度对齐到当前任务原型；Inter-Align则通过对齐到先前任务原型的方向性对齐来减少表示间隙。实验表明，在标准性能和鲁棒性方面均有一致的提升，同时具有更低的遗忘和更强的迁移。我们进一步通过量化Intra-Smooth对扰动设置的敏感性以及Inter-Align对任务相似性和几何距离的影响，分析了关键机制。总之，这些模块在组合时提供互补增益，每个模块也可以单独集成到各种持续学习范式中，包括回放、正则化和动态架构，从而为持续学习提供了一种几何控制机制。
英文摘要：
In dynamic environments, large language models need to keep adapting to new tasks, but continual learning often suffers from forgetting, limited transfer, and vulnerability to adversarial perturbations. To address this, we present AdvCL, which repurposes adversarial perturbations as a geometric control signal for stable continual adaptation. AdvCL combines three plug-in modules: Intra-Smooth promotes local smoothness via small adversarial perturbations; Proto-Clip uses similarity clipping to prevent excessive alignment to current task prototype; and Inter-Align applies directional alignment toward previous task prototype to reduce representational gaps. Experiments show consistent gains in both standard performance and robustness, with lower forgetting and stronger transfer. We further analyze key mechanisms by quantifying the sensitivity of Intra-Smooth to perturbation settings and the effect of Inter-Align on task similarity and geometric distance. In summary, the modules provide complementary gains when combined, and each can also be integrated individually into diverse CL paradigms, including replay, regularization, and dynamic architectures, thereby offering a geometric control mechanism for continual learning.
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。
