Natural and Universal Adversarial Attacks on Prompt-based Language Models

Prompt-based learning models are vulnerable to natural and universal adversarial attacks, impacting model predictions.