测试版
排行榜/Mistral 7B V0.1
Mistral AI logo

Mistral 7B V0.1

开源

来自 Mistral AI · 发布于 2023-09-20

41.6
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 16 benchmarks with 41.6% average. Top scores: TriviaQA (75.2%), HellaSwag (74.7%), OpenBookQA (73.1%).

基准测试类别分数Bar
TriviaQAknowledge75.2
HellaSwagknowledge74.7
OpenBookQAknowledge73.1
ARC AI2knowledge71.5
PIQAknowledge66.0
GSM8Kmath54.4
Winograndeknowledge50.6
MMLUknowledge50.0
BBHreasoning41.5
IFEvallanguage23.9
MMLU-PROknowledge22.4
BBH (HuggingFace)general22.0
ANLIknowledge20.6
MUSRreasoning10.7
GPQAknowledge5.6
MATH Level 5math3.0