Llama 2-7B
Code source ouvertpar Meta · Sorti le 2024-01-01
43.0
score moyen
N/A
Prix d'entrée
N/A
Prix de sortie
N/A
Fenêtre de contexte
text
Type
Tested on 11 benchmarks with 43.0% average. Top scores: TriviaQA (73.7%), LAMBADA (73.3%), HellaSwag (69.6%).
Scores de benchmark
| Benchmark | Catégorie | Score | Bar |
|---|---|---|---|
| TriviaQA | knowledge | 73.7 | |
| LAMBADA | knowledge | 73.3 | |
| HellaSwag | knowledge | 69.6 | |
| PIQA | knowledge | 57.6 | |
| OpenBookQA | knowledge | 44.8 | |
| Winogrande | knowledge | 38.4 | |
| ARC AI2 | knowledge | 27.9 | |
| MMLU | knowledge | 27.7 | |
| ScienceQA | knowledge | 24.1 | |
| BBH | reasoning | 18.9 | |
| GSM8K | math | 16.7 |