CleanVul

Dataset Card for CleanVul CleanVul is a function-level vulnerability dataset derived from vulnerability-fixing commits. It provides paired before/after function snapshots and metadata produced by an LLM-assisted heuristic pipeline (VulSifter) to identify which changed functions are likely to correspond to vulnerability

Papers using CleanVul (2)

Improving LLM Reasoning for Vulnerability Detection via Group Relative Policy Optimization2025CleanVul: Automatic Function-Level Vulnerability Detection in Code Commits Using LLM Heuristics2024 · 2 cites