维基百科链接竞速:高选项数决策演示
让模型在数百到数千个链接中选择下一步,用游戏方式展示决策速度与约束输出的效果。原帖附有 4 分 04 秒视频。

作者重建了 @nutlope 的 Jev 论文地图流程,用 Jev 将 1000 篇 AI 论文分类为 24 个主题,并让 Opus 5 对标签进行评分。
1,000 AI papers sorted into 24 topics for $0.0585. Then Opus 5 graded the labels. @nutlope's Jev paper map went viral, but the pipeline never shipped and the eval was "still running". So I rebuilt both and opened them. The first judge run came back empty: Opus spent its whole