The data and the PyTorch implementation for the models and experiments in the paper "Language Model Decoding as Likelihood–Utility Alignment". - View it on GitHub