FGSM adversarial fine-tuning avoids catastrophic overfitting at standard eps=4 and 8 and matches PGD robustness within 1.4% at a quarter of the training time.
Minimally distorted adversarial examples with a fast adaptive boundary attack
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Are Fast Methods Stable in Adversarially Robust Transfer Learning?
FGSM adversarial fine-tuning avoids catastrophic overfitting at standard eps=4 and 8 and matches PGD robustness within 1.4% at a quarter of the training time.