WHAT NEW IN X10.29 THU

X 上的 AI 合集档案 · 22 个合集 · 60,636 条

jevPOST

Jev 不训练怎么保证有效?只能拿自己数据测

作者回应常见疑问:Jev 是通用分类器,无法预先保证效果,必须用自己的数据和标准实测;好处是会给置信度分数且速度快。

Matija Sosic
@MatijaSosic2026.09.18收录原推发布于 2026.09.18
Matija Sosic @MatijaSosic原推发布于 2026.09.18
在 X 上打开原推

Another good question on Jev that keeps popping up: "If I don't need to train it, how can I be sure it works?" The answer is: you can't. You need to test it on your data and your criteria and verify that it works for you. It's a one-size-fits-all classifier. Just like you can't know whether an LLM's output is correct, you can't know with Jev either. But at least you get a confidence score, and it's fast.

这条是引用着另一条写的:x.com/matijasosic/status/2100190746389135772

x.com/matijasosic/status/2100635155513635088采集于 2026.09.18 10:14