Universal and Transferable Adversarial Attacks on Aligned Language Modelsarxiv.org1 point·simon_acca··0 commentsOpen articleSaveView on HN