The optimized data of DiverseVul-Derived

DiverseVul-Derived-Opt: This dataset is a re-constructed version based on an improved "identical function" detection logic used in the generation process of the DiverseVul-Derived dataset. Specifically, train.jsonl, valid.jsonl, and test.jsonl correspond to the training, validation, and testing sets after data evolution, respectively, while statisticInfo.json contains the corresponding statistical information. If this dataset has been helpful for your research, please consider citing our paper. @article{zhang2025real, title={Real-World Code Vulnerability Detection Framework: From Data Preprocessing to Multi-Feature Fusion Detection}, author={Zhang, Jixian and Du, Qingfeng and Li, Sheng and Lu, Zhongda and He, Ting and Liu, Chengwei}, journal={IEEE Transactions on Dependable and Secure Computing}, year={2025}, publisher={IEEE}}