Gain from Neighbors: Boosting Model Robustness in the Wild via Adversarial Perturbations Toward Neighboring Classes
Open paper
Benchmark
Model
Rank
Results
adversarial-robustness-on-imagenet-c
GFN (FAN)
–
mean Corruption Error (mCE): 39.8
image-classification-on-imagenet
GFN (FAN)
–
Top 1 Accuracy: 86.4%
Rank counts only results with a code link.