[D] Why do we train language models with next word prediction instead of some kind of reinforcement learning-like setup? Submitted by blazejd t3_yzzxa2 on November 20, 2022 at 9:30 AM in MachineLearning 34 comments 18