历史专项测试
历史数据存档。查看近期模型,请返回模型能力评测。
知识与推理 · MMLU-Pro
记录保留原模型配置与数据提供方;计算机知识分项衡量知识,不是代码项目完成率。
262 条 · 按总正确率降序缺测记为 —
| 模型/配置 | 总正确率 | 数据提供方 |
|---|---|---|
| 65.1% | Self-Reported | |
| 65.0% | Self-Reported | |
| 64.4% | TIGER-Lab | |
| 64.1% | TIGER-Lab | |
| 63.9% | TIGER-Lab | |
| 63.7% | TIGER-Lab | |
| 63.7% | Self-Reported | |
| 63.6% | TIGER-Lab | |
| 63.2% | Self-Reported | |
| 63.1% | TIGER-Lab | |
| 63.1% | Self-Reported | |
| 62.8% | TIGER-Lab | |
| 62.8% | TIGER-Lab | |
| 62.4% | TIGER-Lab | |
| 62.1% | TIGER-Lab | |
| 61.8% | Self-Reported | |
| 61.7% | Self-Reported | |
| 61.6% | Self-Reported | |
| 60.6% | Self-Reported | |
| 60.5% | Self-Reported |
来源、版本与使用范围
数据版本:4623b7c68d2276218d487464b0608dc978f4b301
来源许可证:Apache-2.0。模型自身许可证与成绩文件许可证分别核对。本页未运行模型评测;不同来源或配置记录不合并,价格未提供。
固定版本数据文件