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。

Jev 对争议 #241 的裁决为 Award Recipient,无不确定性标记,四次运行结果一致至 ±0.01。由于 Jev 不输出文本,作者指出理由由代码生成:问题为人类而写,答案为数字,代码掌握依据。
Jev's decision on dispute #241 was Award Recipient, with no uncertainty flags. Four runs gave the same numbers to ±0.01. Jev writes no text, so who writes the justification? The code does. Every question was written for humans, every answer is a number, and the code knows