This project reproduces the results from paper: Fool Your (Vision and) Language Model With Embarrassingly Simple Permutations.
The analysis and main experiements can be found in analysis.ipynb, and the main utilitary functions we use are coded in LLMs_attack.ipynb.