jevIMAGE
用 Jev 做 agent 坏习惯检测的分类器
作者做反馈工具检测 agent 的坏做法，把 Jev 当作分类模型来判定输出是否违规。
julian
@crucianijm2026.09.30收录原推发布于 2026.09.30
julian @crucianijm原推发布于 2026.09.30
for the past couple weeks, I've been playing around with improving agent output through feedback tools that detect bad practices while working on this, Jev and a lot of competing classifier models came out, a natural fit for this problem
x.com/crucianijm/status/2105026835021279480采集于 2026.09.30 10:06