U
MPT-30B
제공 Unknown · 출시일 2024-01-01
31.7
평균 점수
N/A
입력 가격
N/A
출력 가격
N/A
컨텍스트 윈도우
text
유형
Tested on 16 benchmarks with 31.7% average. Top scores: TriviaQA (73.6%), LAMBADA (70.0%), HellaSwag (68.5%).
벤치마크 점수
| 벤치마크 | 카테고리 | 점수 | Bar |
|---|---|---|---|
| TriviaQA | knowledge | 73.6 | |
| LAMBADA | knowledge | 70.0 | |
| HellaSwag | knowledge | 68.5 | |
| PIQA | knowledge | 63.8 | |
| Winogrande | knowledge | 42.0 | |
| OpenBookQA | knowledge | 36.0 | |
| GSM8K | math | 34.4 | |
| ARC AI2 | knowledge | 34.1 | |
| MMLU | knowledge | 30.5 | |
| IFEval | language | 21.5 | |
| BBH | reasoning | 17.3 | |
| BBH (HuggingFace) | general | 6.5 | |
| MUSR | reasoning | 2.9 | |
| MMLU-PRO | knowledge | 2.3 | |
| MATH Level 5 | math | 1.6 | |
| GPQA | knowledge | 1.3 |