Evaluating Fairness in Self-supervised and Supervised Models for Sequential Data

Self-supervised learning (SSL) has become the de facto training paradigm of large models where pre-training is followed by supervised fine-tuning using domain-specific data and labels.