A large language model customized by NVIDIA to improve the helpfulness of LLM generated responses. It is a fine-tuned version of Llama 3.1 70B Instruct. The model was trained using RLHF (REINFORCE) with HelpSteer2-Preference prompts.
发布日期2024年10月1日
参数规模70B
上下文长度—
许可证Llama 3.1 Community License
知识截止2023年12月1日
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| GSM8k | math reasoning | 91.4 | 来源 |
| HellaSwag | reasoning | 85.6 | 来源 |
| Winogrande | reasoning language | 84.5 | 来源 |
| GSM8K Chat | math reasoning | 81.9 | 来源 |
| MMLU Chat | language reasoning math general | 80.6 | 来源 |
| MMLU | general reasoning language math | 80.2 | 来源 |
| Instruct HumanEval | general | 73.8 | 来源 |
| ARC-C | reasoning general | 69.2 | 来源 |
| TruthfulQA | general reasoning legal healthcare finance | 58.6 | 来源 |
| XLSum English | summarization language | 31.6 | 来源 |
| MT-Bench | communication reasoning general roleplay | 9.0 | 来源 |
Pricing
API 价格对比
暂无 API 价格。