发表机构
Yunnan University; Yunnan Normal University（云南大学; 云南师范大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对现有对抗蒸馏未充分利用扰动邻域内教师偏好监督的问题，提出CGARD方法，联合优化学生对抗样本与教师协同样本，在CIFAR-10/100上的实验显示其鲁棒性优于强基线。
AI 中文摘要
对抗蒸馏将鲁棒性从高容量教师模型迁移至紧凑的学生模型。现有对抗蒸馏方法主要利用教师模型对干净样本或对抗样本的预测来监督学生学习，但在对抗蒸馏中，扰动邻域内的教师偏好监督仍未得到充分探索。因此，我们提出协同引导对抗鲁棒蒸馏（CGARD），该方法在同一扰动邻域内联合优化不同的学生对抗样本与教师协同样本，其中教师协同样本被约束为在教师模型下产生的交叉熵损失不大于干净输入的损失。CGARD将教师的协同引导与对抗监督相结合，以提升鲁棒知识迁移效果。在CIFAR-10和CIFAR-100数据集上的实验，包括白盒评估及额外的黑盒迁移评估，均表明该方法相较于强对抗蒸馏基线实现了持续的鲁棒性提升。
英文摘要
Adversarial distillation transfers robustness from high-capacity teachers to compact students. Existing adversarial distillation methods mainly use teacher predictions on clean or adversarial examples to supervise student learning. However, teacher-favorable supervision within the perturbation neighborhood remains underexplored in adversarial distillation. We therefore propose Collaboratively Guided Adversarial Robust Distillation (CGARD), which jointly optimizes distinct student-adversarial and teacher-collaborative examples within the same perturbation neighborhood. The teacher-collaborative example is constrained to incur no greater cross-entropy loss under the teacher than the clean input. CGARD combines collaborative teacher guidance with adversarial teacher supervision to improve robust knowledge transfer. Experiments on CIFAR-10 and CIFAR-100, including white-box evaluation and additional black-box transfer evaluation, demonstrate consistent robustness improvements over strong adversarial distillation baselines.
