magic. The repo for paper: Exploiting the Index Gradients for Optimization-Based Jailbreaking on Large Language Models.
