Towards Stable and Robust AdderNets (Supplementary Material) Minjing Dong 1,2, Yunhe Wang

As shown in Table 1, the stability of adversarial robustness is evaluated under different inference settings. We first show whether the shuffle of test set influences the performance. In the main body, we mainly focus on the comparison with CNN since we want to highlight the natural robustness of AdderNet compared to CNN under the same setting. We further evaluate the performance of A WN on CNNs. As shown in Table 3, with the involvement of A WN, CNN obtains slight better adversarial robustness.