返回探索

Jev 真实决策评测基准公开:900 条决策由 Opus 5 评分

该基准包含 900 条真实 Jev 决策,由 Opus 5 评分,一致率 85.8/100,分歧几乎都出现在 Jev 报告低置信度的地方,任何人都可对自己的模型复现。

Open weight builders will need a public yardstick. One went out today: 900 real Jev decisions graded by Opus 5, 85.8 of 100 agree, and the misses sit almost entirely where Jev reported low confidence. Anyone can rerun it against their own model:

· 0 次赞在 X 打开