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