Improving the robustness and accuracy of biomedical language models through adversarial training

Milad Moradi, Matthias Samwald

Abstract: Highlights•We investigated the robustness of deep biomedical language models against adversarial noise.•We conducted extensive adversarial training experiments on various biomedical NLP tasks.•The deep biomedical language models achieved state-of-the-art results after adversarial training.•We improved the robustness and accuracy of the biomedical language models.