SEC-bench Pro, our benchmark for long-horizon software security tasks, is now available on arXiv.
Hwiwon Lee
Computer Science PhD student at UIUC
I build and evaluate AI agents for cybersecurity.
I work across software security, systems, and AI with my advisor, Lingming Zhang.
2026 →
Chronicle
Agentic Vulnerability Reasoning on COTS Binaries is now available on arXiv.
Recognized as a Microsoft Most Valuable Security Researcher for 2026 Q1.
Selected
Publications
2026
arXiv
SEC-bench Pro: Can Language Models Solve Long-Horizon Software Security Tasks?
PDF: SEC-bench Pro2026
arXiv
Agentic Vulnerability Reasoning on COTS Binaries
PDF: Agentic Vulnerability Reasoning on COTS Binaries2025
NeurIPS
SEC-bench: Automated Benchmarking of LLM Agents on Real-World Software Security Tasks
PDF: SEC-bench2024
IEEE S&P · Distinguished Paper