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。

LLM擅长生成文本,但如果应用只需要决策而不需要文本呢?JEV是一种通用决策模型,直接返回结构化输出,无需生成token来决定下一步。
LLMs are great at generating text. But what if your application doesn't need text? It just needs a decision. That's where JEV gets interesting. Instead of generating tokens to decide what to do next, JEV is designed as a general-purpose decision model that returns structured