PrimeVul

PrimeVul+ is a dataset that contains high-confidence repair pairs constructed from real-world vulnerabilities, used to evaluate the effectiveness of secure code generation frameworks.

Papers using PrimeVul (9)

Llama-based source code vulnerability detection: Prompt engineering vs Fine tuning2025 · 2 citesK-ASTRO: Structure-Aware Adaptation of LLMs for Code Vulnerability Detection2022 · 3 citesVulKey: Automated Vulnerability Repair Guided by Domain-Specific Repair Patterns2026Learning to Generate Secure Code via Token-Level Rewards2026VulInstruct: Teaching LLMs Root-Cause Reasoning for Vulnerability Detection via Security Specifications2025Boosting Vulnerability Detection of LLMs via Curriculum Preference Optimization with Synthetic Reasoning Data2025Vulnerability Detection with Code Language Models: How Far Are We?2024 · 13 citesCleanVul: Automatic Function-Level Vulnerability Detection in Code Commits Using LLM Heuristics2024 · 2 citesANVIL: Anomaly-based Vulnerability Identification without Labelled Training Data2024