解决长尾数据下对抗训练的不平衡问题，自适应调整扰动提升鲁棒性。
Taming the Long Tail: Rebalancing Adversarial Training via Adaptive Perturbation
- 提出自适应扰动机制，动态调整对抗训练中的扰动强度。
- 在长尾数据集上显著提升模型鲁棒性和类别平衡性。
- 无需修改模型结构，可直接接入现有对抗训练流程。
深度神经网络对对抗样本（即微小扰动）极为敏感，可能大幅降低性能。尽管对抗训练是主流防御策略，但多数研究聚焦于平衡数据集，忽略了真实世界中长尾数据带来的挑战。我们理论分析发现，对抗样本中的扰动会改变训练分布，进而引发两个关键问题：(i) 由类别不平衡导致的偏斜训练目标；(ii) 对抗分布演化不稳定。进一步表明，扰动可同时缓解对抗脆弱性与类别不平衡。基于此，我们提出 RobustLT 框架，通过在对抗训练过程中自适应调整扰动，实现稳健优化。大量实验表明，RobustLT 在长尾数据集上持续提升对抗鲁棒性与类别平衡性。代码已开源：https://github.com/zhang-lilin/RobustLT。
原文摘要 · Abstract (English)
Deep neural networks are highly vulnerable to adversarial examples, i.e.,small perturbations that can significantly degrade model performance. While adversarial training has become the primary defense strategy, most studies focus on balanced datasets, overlooking the challenges posed by real-world long-tail data. Motivated by the fact that perturbations in adversarial examples inherently alter the training distribution, we theoretically investigate their impact. We first revisit adversarial training for long-tail data and identify two key limitations: (i) a skewed training objective caused by class imbalance, and (ii) unstable evolution of adversarial distributions. Furthermore, we show that perturbations can simultaneously address both adversarial vulnerability and class imbalance. Based on these insights, we propose RobustLT, a plug-and-play framework that adaptively adjusts perturbations during adversarial training. Extensive experiments demonstrate that RobustLT consistently enhances adversarial robustness and class-balance on long-tailed datasets. The code is available at \href{https://github.com/zhang-lilin/RobustLT}{https://github.com/zhang-lilin/RobustLT}.
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。