3 ms·Universal and Transferable Attacks on Aligned Language Models3 points by fgfm 3y agofgfm 3y agoStudy on adversarial attacks on LLMs to steer their objective into misalignment.