The model saw the same case and the same 19 labels; only their positions changed. Yet models frequently changed answers!
只换选项顺序、病例和标签一字不动，模型答案就变，说明分数里混着大量与医学无关的形式噪声。这条可以直接迁移到任何榜单：看到排名差几个百分点，先问是不是换个 prompt、换个选项顺序就翻盘，别急着当能力差距读。
1 Matching Annotations
- Last 7 days
-
-