论文标题
（适当的）定义可以调和鲁棒性和准确性
Robustness and Accuracy Could Be Reconcilable by (Proper) Definition
论文作者
论文摘要
在对抗文献中，鲁棒性和准确性之间的权衡得到了广泛的研究。尽管仍然有争议，但普遍的观点是，从经验或理论上，这种权衡是固有的。因此，我们在对抗训练中挖掘了这种权衡的起源，发现它可能源于不当定义的可靠误差，该错误施加了局部不变性的诱导偏见 - 对平稳性的过度校正。鉴于此，我们主张采用局部模棱两可来描述强大模型的理想行为，从而导致自符合的强大错误分数。根据定义，得分有助于鲁棒性与准确性之间的对帐，同时仍通过稳健优化处理最坏情况的不确定性。通过简单地将KL差异替换为距离指标的变体，得分可以有效地最小化。从经验上讲，我们的模型在AutoAttack下的强大稳定板上达到了最高的性能。此外，得分提供了指导性见解，以解释在健壮模型上观察到的过度拟合现象和语义输入梯度。代码可在https://github.com/p2333/score上找到。
The trade-off between robustness and accuracy has been widely studied in the adversarial literature. Although still controversial, the prevailing view is that this trade-off is inherent, either empirically or theoretically. Thus, we dig for the origin of this trade-off in adversarial training and find that it may stem from the improperly defined robust error, which imposes an inductive bias of local invariance -- an overcorrection towards smoothness. Given this, we advocate employing local equivariance to describe the ideal behavior of a robust model, leading to a self-consistent robust error named SCORE. By definition, SCORE facilitates the reconciliation between robustness and accuracy, while still handling the worst-case uncertainty via robust optimization. By simply substituting KL divergence with variants of distance metrics, SCORE can be efficiently minimized. Empirically, our models achieve top-rank performance on RobustBench under AutoAttack. Besides, SCORE provides instructive insights for explaining the overfitting phenomenon and semantic input gradients observed on robust models. Code is available at https://github.com/P2333/SCORE.
