license: apache-2.0
language:
- en
- zh
GraphGen-Data
Data Description
GraphGen-Data is the dataset for verification in the paper "GraphGen: Enhancing Supervised Fine-Tuning for LLMs with Knowledge-Driven Synthetic Data Generation". It involves three domains:
- Agricultural(SeedEval)
- Medical(PQArefEval)
- General(HotpotEval)
GraphGen is a framework for synthetic data generation guided by knowledge graphs. We released our code in Github.
Source Data
SeedEval is adapted from SeedBench, a benchmark with 11 tasks related to seed knowledge. For this study, we selected Task QA–4 (covering one-shot and zero-shot scenarios) related to textual knowledge question answering. PQArefEval is derived from PQAref, from which we extracted 5,818 instances for our analysis. HotpotQA is a dataset for diverse, explainable multi-hop question answering, where questions require integrating information from multiple sources. We used the test set of HotpotQA as the new evaluation dataset, HotpotEval. Each dataset comprises two components: the QA test set and the corresponding source texts. The Corpus for SeedEval is provided by anonymous agricultural experts and cannot be made public due to confidentiality restrictions. The Corpus for PQArefEval and HotpotEval are constructed from the original references of PQAref and HotpotQA, respectively.
Citation
BibTeX:
@misc{chen2025graphgenenhancingsupervisedfinetuning,
title={GraphGen: Enhancing Supervised Fine-Tuning for LLMs with Knowledge-Driven Synthetic Data Generation},
author={Zihong Chen and Wanli Jiang and Jinzhe Li and Zhonghang Yuan and Huanjun Kong and Wanli Ouyang and Nanqing Dong},
year={2025},
eprint={2505.20416},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2505.20416},
}