An LLM can Fool Itself: A Prompt-Based Adversarial Attack (ICLR 2024) - View it on GitHub