A Python package for predicting large-scale adversarial risk in Large Language Models under Best-of-N sampling - View it on GitHub