{
 "id": "HYUKI5/MegaScience__MegaScience",
 "owner": "HYUKI5",
 "name": "MegaScience__MegaScience",
 "downloads": 0,
 "likes": 0,
 "last_modified": "2026-11-16T11:46:16.000Z",
 "stated_licence": "cc-by-nc-sa-4.0",
 "stated_from": "tag",
 "licence_name": null,
 "licence_link": null,
 "file_licence": null,
 "licence_files": [],
 "sources": [
  {
   "id": "MegaScience/MegaScience",
   "licence": "cc-by-nc-sa-4.0",
   "found": true,
   "evidence": "https://huggingface.co/datasets/MegaScience/MegaScience/blob/8df5586005374acba25aecc4f5469ce30fec605c/README.md"
  }
 ],
 "gated": false,
 "tasks": [
  "text-generation"
 ],
 "citation": "@article{fan2025megascience,\n  title={MegaScience: Pushing the Frontiers of Post-Training Datasets for Science Reasoning},\n  author={Fan, Run-Ze and Wang, Zengzhi and Liu, Pengfei},\n  year={2025},\n  journal={arXiv preprint arXiv:2507.16812},\n  url={https://arxiv.org/abs/2507.16812}\n}\n\n<details>\n<summary>原文（英語）</summary>\n\n# [MegaScience: Pushing the Frontiers of Post-Training Datasets for Science Reasoning](https://arxiv.org/abs/2507.16812)\n\n**Code:** https://github.com/GAIR-NLP/MegaScience\n**Project Page:** https://huggingface.co/MegaScience\n\nMegaScience is a large-scale mixture of high-quality open-source datasets consisting of 1.25 million instances. We first collect multiple public datasets, then conduct comprehensive ablation studies across different data selection methods to identify the optimal approach for each dataset, thereby contributing high-quality subsets. Furthermore, we annotate step-by-step solutions for all datasets except TextbookReasoning.\n\nWe train Llama3.1, Qwen2.5, and Qwen3 series base models on MegaScience, which outperform the official instruct models in average scientific reasoning performance, successfully advancing the frontiers of the open-source community in the science domain. We find that MegaScience exhibits greater effectiveness for larger and stronger models, suggesting a scaling benefit for scientific instruction tuning.\n![](main_figure.png)\n\n## Sample Usage\n\nTo use the MegaScience dataset, you can clone the repository with Git LFS:",
 "signal": "restrictive",
 "reason": "The declared licence (cc-by-nc-sa-4.0) is classified as having use restrictions.",
 "revision": "f7f6623530d9579ec64ab94cb919df873ff54334",
 "stated_evidence": "https://huggingface.co/datasets/HYUKI5/MegaScience__MegaScience/blob/f7f6623530d9579ec64ab94cb919df873ff54334/README.md",
 "file_evidence": "https://huggingface.co/datasets/HYUKI5/MegaScience__MegaScience/tree/f7f6623530d9579ec64ab94cb919df873ff54334",
 "card_nc": false
}