ROSE: Robust Selective Fine-tuning for Pre-trained Language Models
This work addresses the problem of adversarial robustness for users of pre-trained language models, offering an incremental enhancement to existing fine-tuning methods.
The paper tackles the vulnerability of pre-trained language models to adversarial attacks by introducing ROSE, a robust selective fine-tuning method that filters out unrobust parameter updates, achieving significant improvements in adversarial robustness on various NLP tasks.
Even though the large-scale language models have achieved excellent performances, they suffer from various adversarial attacks. A large body of defense methods has been proposed. However, they are still limited due to redundant attack search spaces and the inability to defend against various types of attacks. In this work, we present a novel fine-tuning approach called \textbf{RO}bust \textbf{SE}letive fine-tuning (\textbf{ROSE}) to address this issue. ROSE conducts selective updates when adapting pre-trained models to downstream tasks, filtering out invaluable and unrobust updates of parameters. Specifically, we propose two strategies: the first-order and second-order ROSE for selecting target robust parameters. The experimental results show that ROSE achieves significant improvements in adversarial robustness on various downstream NLP tasks, and the ensemble method even surpasses both variants above. Furthermore, ROSE can be easily incorporated into existing fine-tuning methods to improve their adversarial robustness further. The empirical analysis confirms that ROSE eliminates unrobust spurious updates during fine-tuning, leading to solutions corresponding to flatter and wider optima than the conventional method. Code is available at \url{https://github.com/jiangllan/ROSE}.