Large language models (LLMs) are increasingly being deployed as software engineering agents that autonomously contribute to repositories. A major benefit these agents present is their ability to find and patch security vulnerabilities in the codebases they oversee. To estimate the capability of agents in this domain, we introduce ZeroDayBench, a benchmark where LLM agents find and patch 22 novel critical vulnerabilities in open-source codebases. We focus our efforts on three popular frontier agentic LLMs: GPT-5.2, Claude Sonnet 4.5, and Grok 4.1. We find that frontier LLMs are not yet capable of autonomously solving our tasks and observe some behavioral patterns that suggest how these models can be improved in the domain of proactive cyberdefense.
@article{arxiv.2603.02297,
title = {ZeroDayBench: Evaluating LLM Agents on Unseen Zero-Day Vulnerabilities for Cyberdefense},
author = {Nancy Lau and Louis Sloot and Jyoutir Raj and Giuseppe Marco Boscardin and Evan Harris and Dylan Bowman and Mario Brajkovski and Jaideep Chawla and Dan Zhao},
journal= {arXiv preprint arXiv:2603.02297},
year = {2026}
}
Comments
Accepted to ICLR 2026 Workshop "Agents in the Wild"