Data for the Paper "Multi-Taxonomy Vulnerability Classification with Hierarchically Finetuned Language Models"

This repository contains the data and experiment artifacts used in the paper "Multi-Taxonomy Vulnerability Classification with Hierarchically Finetuned Language Models" (accepted for publication at DIMVA 2026). The repository is organized as follows: config/ – Configuration files used for running the experiments. data/ – The dataset used in the study. flat_classifiers/ – Logs, checkpoints, and configuration files for experiments with flat classifiers. hierarchical_classifiers/ – Logs, checkpoints, and configuration files for experiments with hierarchical classifiers. temporal_splits/ – Data and configurations for experiments evaluating temporal splits. text_generation_models/ – Predictions and prompts used for experiments with text generation models as classifiers. Each fine-tuning experiment folder (e.g., within flat_classifiers, hierarchical_classifiers, or temporal_splits) follows a consistent structure that includes configuration files, TensorBoard logs, and test metrics. To save space on Zenodo, checkpoints and other large files have been omitted.