Llama 3.1 Nemotron 70B Instruct
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.
Benchmark results
| Benchmark | Score | Tags | Source |
|---|---|---|---|
| ARC-C | 69.2% | self-reported llm-stats | link → |
| GSM8k | 91.4% | self-reported llm-stats | link → |
| GSM8K Chat | 81.9% | self-reported llm-stats | link → |
| HellaSwag | 85.6% | self-reported llm-stats | link → |
| Instruct HumanEval | 73.8% | self-reported llm-stats | link → |
| MMLU | 80.2% | self-reported llm-stats | link → |
| MMLU Chat | 80.6% | self-reported llm-stats | link → |
| MT-Bench | 0.09 | self-reported llm-stats | link → |
| TruthfulQA | 58.6% | self-reported llm-stats | link → |
| Winogrande | 84.5% | self-reported llm-stats | link → |
| XLSum English | 31.6% | self-reported llm-stats | link → |