nyoncore·2 年前·議論Isn't it obvious that since LLM are trained to predict the next word they do better than to predict the previous one?