WHAT NEW IN X10.31 SAT

X 上的 AI 合集档案 · 22 个合集 · 62,413 条

jevPOST

Jev-as-judge:可放进循环里的实时裁判

作者预测 Jev 做裁判的用例会爆发,列出 agent 漂移、卡死检测、进度完整性、工具调用风险、上下文相关性等可连续运行的判断场景。

Josh Rosen
@JoshARosen2026.09.20收录原推发布于 2026.09.19
Josh Rosen @JoshARosen原推发布于 2026.09.19
在 X 上打开原推

Look for an explosion of Jev-as-judge use cases. Jev-as-judge is the near realtime counterpart to LLM-as-judge that’s fast and cheap enough to run continuously inside the loop: - agent drift - worker stuck detection - progress and completeness - tool call risk - context relevance - output quality - human intervention needed

x.com/josharosen/status/2101286067739541819采集于 2026.09.20 10:16