ContextualAI/CLAIR_and_APO

Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment

Jupyter NotebookMIT