natural_reasoning facebook
NaturalReasoning is a large-scale dataset for general reasoning tasks. It consists of high-quality challenging reasoning questions backtranslated from pretraining corpora DCLM and FineMath. The questions have been deduplicated and decontaminated from popular reasoning benchmarks including MATH, GPQA, MMLU-Pro, MMLU-STEM. For each question, we extract the reference final answer from the original document from the pretraining corpora if possible. We also provide a model-generated response from… See the full description on the dataset page: https://huggingface.co/datasets/facebook/natural_reasoning.
- 类型
- dataset
- 许可
- cc-by-nc-4.0
- 语言
- en
- 下载量
- 2,905
- 点赞
- 580
- 访问
- public
- 文件
- 0
标签
- 数学推理
- 推理
- 预训练语料
- 指令微调
- 评测基准
- 大语言模型
- 文本数据集
摘要
NaturalReasoning 是由 Meta 发布的大规模通用推理数据集，包含从预训练语料 DCLM 和 FineMath 反向翻译生成的约110万条高质量挑战性推理问题，并与 MATH、GPQA 等主流推理基准做了去重和去污染处理。该数据集提供参考答案及 Llama3.3-70B 生成的回答，训练 Llama3.1-8B 时展现出优于其他数据集的扩展效果，适用于训练和强化大语言模型的推理能力。
README
--- license: cc-by-nc-4.0 task_categories: - text-generation language: - en pretty_name: Natural Reasoning size_categories: - 1M<n<10M --- [NaturalReasoning](https://arxiv.org/abs/2502.13124) 是一个用于通用推理任务的大规模数据集。它包含从预训练语料库 [DCLM](https://github.com/mlfoundations/dclm) 和 [FineMath](https://huggingfac…