Output-space adversarial training improved clean-data performance and adversarial robustness of two bird sound classifiers across seven soundscape test sets, and stabilized prototype-based explanations.
Impact of Adversarial Training on Robustness and Generalizability of Language Models
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
Adversarial training is widely acknowledged as the most effective defense against adversarial attacks. However, it is also well established that achieving both robustness and generalization in adversarially trained models involves a trade-off. The goal of this work is to provide an in depth comparison of different approaches for adversarial training in language models. Specifically, we study the effect of pre-training data augmentation as well as training time input perturbations vs. embedding space perturbations on the robustness and generalization of transformer-based language models. Our findings suggest that better robustness can be achieved by pre-training data augmentation or by training with input space perturbation. However, training with embedding space perturbation significantly improves generalization. A linguistic correlation analysis of neurons of the learned models reveals that the improved generalization is due to 'more specialized' neurons. To the best of our knowledge, this is the first work to carry out a deep qualitative analysis of different methods of generating adversarial examples in adversarial training of language models.
citation-role summary
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Adversarial Training Improves Generalization Under Distribution Shifts in Bioacoustics
Output-space adversarial training improved clean-data performance and adversarial robustness of two bird sound classifiers across seven soundscape test sets, and stabilized prototype-based explanations.