A logit-space regularization penalty (ALR) during adversarial training and fine-tuning improves robustness across sequences of evolving attacks and reduces drops in previously defended attacks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Adapting to Evolving Adversaries with Regularized Continual Robust Training
A logit-space regularization penalty (ALR) during adversarial training and fine-tuning improves robustness across sequences of evolving attacks and reduces drops in previously defended attacks.