Timezone: »

Robustness and Accuracy Could Be Reconcilable by (Proper) Definition
Tianyu Pang · Min Lin · Xiao Yang · Jun Zhu · Shuicheng Yan

Wed Jul 20 10:40 AM -- 10:45 AM (PDT) @ Room 309

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.

Author Information

Tianyu Pang (Sea AI Lab)
Min Lin (Sea AI Lab)
Xiao Yang (Tsinghua University, Tsinghua University)
Jun Zhu (Tsinghua University)
Shuicheng Yan (Sea AI Labs)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors