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。

用户给 Jev 两个相同 diff(标为 A 和 B)并询问哪个更好,发现参数顺序和标签都会影响评分,A 稳定在 0.95 左右,交换顺序后结果变化。
(2/5) hit some wild results. argument order and labels both move the score i gave jev 2 identical diffs (labelled A and B) and asked which is better. the results were shocking: fork A at 0.95 (five runs: 0.95, 0.96, 0.95, 0.96, 0.96) put B first with the same names and A is