Robust recipes for to align language models with human and AI preferences - View it on GitHub