Adversarial Training for Robust and Generalizable Natural Language Processing

到目前为止，进展顺利。去年我们在两个国际会议上发表了两篇论文。按照原来的计划，我们探索了在对抗性训练中将扰动应用于网络的其他层，以及与多任务学习的结合。发表的论文如下：Lis Kanashiro Pereira，Kevin Deh，Fei Cheng，Masayuki Asahara，Ichiro Kobayashi。专注于注意力的对抗性训练，以进行稳健的时间推理。LREC 2022.Lis Kanashiro Pereira，小林一郎OCHADAI在SemEval-2022年任务2：多语种智商检测的对抗性培训。2022年半决赛。

The progress has been smooth so far. We were able to publish two papers at two international conferences last year. Following the original plan, we explored applying perturbations to other layers of the network during adversarial training, as well as the combination with multi-task learning.The published papers are the following:Lis Kanashiro Pereira, Kevin Duh, Fei Cheng, Masayuki Asahara, Ichiro Kobayashi . Attention-Focused Adversarial Training for Robust Temporal Reasoning. LREC 2022.Lis Kanashiro Pereira, Ichiro Kobayashi. OCHADAI at SemEval-2022 Task 2: Adversarial Training for Multilingual Idiomaticity Detection. SemEval 2022.