Jailbreaking Large Language Models via Multi-Task Embedding-based Prompt

MTEA: Paper Code Reproduction Overview This repository contains the official code and experimental results for our paper. The repository consists of three main components: - MTEA/: Implementation of our proposed method.- Baseline/: Source code of baseline methods used for comparison.- results/: Experimental results obtained in our experiments. It contains META's complete attack prompts and responses for each victim model. The datasets used in the paper are provided directly in the root directory.