Llama Evaluation Suite
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Large Language Model Reasoning and Understanding | Llama-3.2-3B-Instruct Evaluation Suite (test val) | MMLU Pro37.3 | 25 | |
| Zero-shot Reasoning and Language Modeling | LLaMA-3-8B Evaluation Suite ARC-C, HellaSwag, WinoGrande, PIQA, ARC-E, BoolQ, GSM8K, WikiText-2 (test) | ARC-C Accuracy53.4 | 13 | |
| Language Understanding | Llama-3.1-70B Evaluation Suite MMLU, WinoGrande, HellaSwag, ARC-Easy, ARC-Challenge | MMLU78.58 | 13 | |
| Language Understanding and Code Generation | Llama 1B Evaluation Suite (ARC, HellaSwag, MMLU, TruthfulQA, WinoGrande, Humaneval) 3.2 | ARC39.33 | 6 |
Showing 4 of 4 rows