Regularized Best-of-N Sampling with Minimum Bayes Risk Objective for Language Model Alignment.

Yuu Jinnai, Tetsuro Morimura, Kaito Ariu, Kenshi Abe

Browse the full NAACL paper archive.