Llama 2 7b Hf
开源来自 Meta · 发布于 2023-07-13
23.6
平均分
N/A
输入价格
N/A
输出价格
N/A
上下文窗口
text-generation
类型
Tested on 11 benchmarks with 23.6% average. Top scores: JSQuAD (79.9%), LLM-JP — Overall (37.2%), JNLI (36.1%).
基准测试分数
| 基准测试 | 类别 | 分数 | Bar |
|---|---|---|---|
| JSQuAD | language | 79.9 | |
| LLM-JP — Overall | language | 37.2 | |
| JNLI | language | 36.1 | |
| JMMLU | language | 28.6 | |
| JCommonsenseQA | language | 25.5 | |
| IFEval | language | 25.2 | |
| BBH (HuggingFace) | general | 10.3 | |
| MMLU-PRO | knowledge | 9.6 | |
| MUSR | reasoning | 3.8 | |
| GPQA | knowledge | 2.2 | |
| MATH Level 5 | math | 1.7 |