Language Alignment via Nash-learning and Adaptive feedback

Ari Azarafrooz, Farshid Faal

Recent research has shown the potential of Nash Learning via Human Feedback for large language model alignment by incorporating the notion of a preference model in a minimax game s…