Improving Dialog Safety using Socially Aware Contrastive Learning
The paper discusses the potential risks of conversational AI systems generating unsafe, toxic, or unethical content. Previous attempts to train these systems with adversarial datasets have been inadequate, as the…
Continue reading