Jev 与文本生成模型的取舍
Jev 不生成自由文本。Diogo 用一段对比视频说明并行决策模型与逐字生成模型在任务形态上的差别。

帖子指出传统分类需要为每组固定类别创建新数据集并训练,而 Jev 绕过这一过程,无需创建数据集或重新训练即可支持新的类别集合。
This creativity and flexibility are what we need. Many think they can do normal classification but that requires creating a new dataset for every fixed set of classes. Jev bypass that process. We don’t have to create a dataset and train whenever we create a new set of classes