[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"readme:graphgen":3},"\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002FInternScience\u002Fgraphgen\u002FHEAD\u002Fassets\u002Flogo.png\" \u002F>\n\u003C\u002Fp>\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopen-sciencelab\u002FGraphGen\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopen-sciencelab\u002FGraphGen.svg\" alt=\"stars\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopen-sciencelab\u002FGraphGen\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002Fopen-sciencelab\u002FGraphGen.svg\" alt=\"forks\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopen-sciencelab\u002FGraphGen\u002Fissues\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues-raw\u002Fopen-sciencelab\u002FGraphGen\" alt=\"open issues\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopen-sciencelab\u002FGraphGen\u002Fissues\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues-closed-raw\u002Fopen-sciencelab\u002FGraphGen\" alt=\"issue resolution\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fchenzihong.gitbook.io\u002Fgraphgen-cookbook\u002F\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-latest-blue\" alt=\"documentation\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fpypi.org\u002Fproject\u002Fgraphg\u002F\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fgraphg.svg?style=flat&logo=pypi&logoColor=white\" alt=\"pypi\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fcdn.vansin.top\u002Finternlm\u002Fdou.jpg\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fwechat-brightgreen?logo=wechat&logoColor=white\" alt=\"wechat\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.20416\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPaper-arXiv-white\" alt=\"arXiv\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fpapers\u002F2505.20416\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPaper-on%20HF-white?logo=huggingface&logoColor=yellow\" alt=\"Hugging Face\" \u002F>\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fchenzihong\u002FGraphGen\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDemo-on%20HF-blue?logo=huggingface&logoColor=yellow\" alt=\"Hugging Face\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fmodelscope.cn\u002Fstudios\u002Fchenzihong\u002FGraphGen\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%96%20Demo-on%20MS-green\" alt=\"Model Scope\" \u002F>\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>GraphGen: Enhancing Supervised Fine-Tuning for LLMs with Knowledge-Driven Synthetic Data Generation\u003C\u002Fp>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FInternScience\u002Fgraphgen\u002Fblob\u002FHEAD\u002FREADME.md\" rel=\"nofollow ugc noopener\">English\u003C\u002Fa> | \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FInternScience\u002Fgraphgen\u002Fblob\u002FHEAD\u002FREADME_zh.md\" rel=\"nofollow ugc noopener\">中文\u003C\u002Fa>\u003C\u002Fp>\n\u003Cdetails>\n\u003Csummary>\u003Cb>📚 Table of Contents\u003C\u002Fb>\u003C\u002Fsummary>\u003Cul>\n\u003Cli>📝 \u003Ca href=\"#-what-is-graphgen\" rel=\"nofollow ugc noopener\">What is GraphGen?\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>📌 \u003Ca href=\"#-latest-updates\" rel=\"nofollow ugc noopener\">Latest Updates\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>⚙️ \u003Ca href=\"#-support-list\" rel=\"nofollow ugc noopener\">Support List\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>🚀 \u003Ca href=\"#-quick-start\" rel=\"nofollow ugc noopener\">Quick Start\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>🏗️ \u003Ca href=\"#-system-architecture\" rel=\"nofollow ugc noopener\">System Architecture\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>🍀 \u003Ca href=\"#-acknowledgements\" rel=\"nofollow ugc noopener\">Acknowledgements\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>📚 \u003Ca href=\"#-citation\" rel=\"nofollow ugc noopener\">Citation\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>📜 \u003Ca href=\"#-license\" rel=\"nofollow ugc noopener\">License\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>📅 \u003Ca href=\"#-star-history\" rel=\"nofollow ugc noopener\">Star History\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fdetails>\u003Ch2>📝 What is GraphGen?\u003C\u002Fh2>\n\u003Cp>GraphGen is a framework for synthetic data generation guided by knowledge graphs. Please check the \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.20416\" rel=\"nofollow ugc noopener\">\u003Cstrong>paper\u003C\u002Fstrong>\u003C\u002Fa> and \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopen-sciencelab\u002FGraphGen\u002Fissues\u002F17\" rel=\"nofollow ugc noopener\">best practice\u003C\u002Fa>.\u003C\u002Fp>\n\u003Cp>It begins by constructing a fine-grained knowledge graph from the source text，then identifies knowledge gaps in LLMs using the expected calibration error metric, prioritizing the generation of QA pairs that target high-value, long-tail knowledge.\nFurthermore, GraphGen incorporates multi-hop neighborhood sampling to capture complex relational information and employs style-controlled generation to diversify the resulting QA data.\u003C\u002Fp>\n\u003Cp>After data generation, you can use \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhiyouga\u002FLLaMA-Factory\" rel=\"nofollow ugc noopener\">LLaMA-Factory\u003C\u002Fa> and \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FInternLM\u002Fxtuner\" rel=\"nofollow ugc noopener\">xtuner\u003C\u002Fa> to finetune your LLMs.\u003C\u002Fp>\n\u003Ch2>📌 Latest Updates\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>🎉 \u003Cstrong>2026.04.13\u003C\u002Fstrong>: The paper based on GraphGen, \u003Cem>Knowledge-to-Verification: Exploring RLVR for LLMs in Knowledge-Intensive Domains\u003C\u002Fem>, has been accepted to the \u003Cstrong>ACL 2026\u003C\u002Fstrong> Main Conference! Congratulations! [\u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.18261\" rel=\"nofollow ugc noopener\">arXiv\u003C\u002Fa>][\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FSeedScientist\u002FK2V\" rel=\"nofollow ugc noopener\">Code\u003C\u002Fa>]\u003C\u002Fli>\n\u003Cli>\u003Cstrong>2026.02.04\u003C\u002Fstrong>: We support HuggingFace Datasets as input data source for data generation now.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>2026.01.15\u003C\u002Fstrong>: \u003Cstrong>LLM benchmark synthesis\u003C\u002Fstrong> now supports single\u002Fmultiple-choice & fill-in-the-blank & true-or-false—ideal for education 🌟🌟\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cdetails>\n\u003Csummary>History\u003C\u002Fsummary>\u003Cul>\n\u003Cli>\u003Cstrong>2025.12.26\u003C\u002Fstrong>: Knowledge graph evaluation metrics about accuracy (entity\u002Frelation), consistency (conflict detection), structural robustness (noise, connectivity, degree distribution)\u003C\u002Fli>\n\u003Cli>\u003Cstrong>2025.12.16\u003C\u002Fstrong>: Added \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ffacebook\u002Frocksdb\" rel=\"nofollow ugc noopener\">rocksdb\u003C\u002Fa> for key-value storage backend and \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fkuzudb\u002Fkuzu\" rel=\"nofollow ugc noopener\">kuzudb\u003C\u002Fa> for g\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fdetails>",1785715924119]