Counterfactual, balanced benchmark for code vulnerability detection (CWE-20: Improper Input Validation) with graph representations (CPG), token mappings, node-level metadata, and GNN-ready inputs.Developed as part of the paper VISION: Robust and Interpretable Code Vulnerability Detection Leveraging Counterfactual Augmentation (AIES 2025).Full implementation available at 👉 GitHub: David-Egea/VISION.
CWE-20-CFA is a balanced… See the full description on the dataset page:
https://huggingface.co/datasets/David-Egea/CWE-20-CFA.
