Reproduce “CLEAR: Causal Context-Based Agentic Reasoning for Vulnerability Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sungju Yun, Sijune Hwang, Yeonjoon Lee, Kyungtae Kang, Sungbin Park
CLEAR represents vulnerability histories as chains of entrypoints, conditions, root causes, and patch intent in a vulnerability knowledge graph. It retrieves related chains and uses Claim, Critic, and Judge agents to assess whether target code matches them. On PrimeVul, gpt-4o-mini achieves 19.17% pair-correct prediction versus VulTrial’s 8.31%, but often labels both pair members benign; this is a benchmark result, not proof of repository-wide effectiveness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.