Mattimax/DATA-AI_full-reasoning

NaturalReasoning是一个用于通用推理任务的大规模数据集。它由来自预训练语料库[DCLM](https://github.com/mlfoundations/dclm)和[FineMath](https://huggingface.co/datasets/HuggingFaceTB/finemath)的高质量挑战性推理问题翻译而成，并且已经去重和去污染，不受其他流行推理基准测试（如MATH、GPQA、MMLU-Pro、MMLU-STEM）的干扰。对于每个问题，如果可能，我们从预训练语料库的原始文档中提取参考最终答案。我们还提供了DATA-AI_Chat_4_Thinking模型生成的响应。该数据集向研究界发布了一个包含110万条问题的子集，以促进对强大LLM推理器的训练研究。

NaturalReasoning is a large-scale dataset for general reasoning tasks. It consists of high-quality challenging reasoning questions backtranslated from pretraining corpora [DCLM](https://github.com/mlfoundations/dclm) and [FineMath](https://huggingface.co/datasets/HuggingFaceTB/finemath), and has been deduplicated and decontaminated from popular reasoning benchmarks including MATH, GPQA, MMLU-Pro, MMLU-STEM. Reference final answers are extracted from the original documents of the pretraining corpora when possible, and model-generated responses are provided. A subset of 1.1 million instances of the dataset has been released to the research community to foster research on training strong LLM reasoners.