Reproduce “SoK: Assessing the Evolution of LLM Capability in Software Vulnerability Research” 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

Amine Lbath, Manan Suri, Aurelien Delaitre, Vadim Okun, Massih-Reza Amini, Dinesh Manocha

The authors map vulnerability-agent claims across four tasks, assessing autonomy, correctness, access, and evidence. They find agents completed vulnerability work on real software without human intervention before restricted frontier systems of 2026; open-weight models later reached comparable autonomy for discovery and validation. CVE disclosures more than doubled in 2026, but records do not establish LLM causation or lifecycle acceleration.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.