Discriminative Finetuning of Generative Large Language Models without Reward Models and preference | Xiaol.x | Podwise