Discover a low-cost method to estimate target task performance from unsupervised pretext tasks in semi/self-supervised learning for improved AI models.
Boost semi-supervised text classification with adversarial training to improve model robustness, generalization, and use of unlabeled data effectively.