3 ms·Universal and Transferable Adversarial Attacks on Aligned Language Models2 points by ent101 3y ago