Decoding-time Realignment of Language Models
Decoding-time Realignment (DeRa) is a method that allows users to adjust how aligned a language model's output is during generation without retraining. It blends between an unaligned reference model and an aligned model at decoding time using a parameter called lambda (λ). This lets users explore different alignment strengths simply by changing λ, controlling the trade-off between following the original model and adopting desired behaviors.