Hulin Wang
Hulin Wang
About
Projects
Experience
Honors
Publications
News
Contact
CV
Light
Dark
Automatic
LLM Agents
Autonomous Vulnerability Patching Agent — DARPA AIxCC
Led the patching component of Team Shellphish’s cyber reasoning system, which placed
5th worldwide
in DARPA’s AI Cyber Challenge.
>95% patch correctness
in the final scored rounds; 30+ C/Java vulnerabilities patched in projects such as curl, NGINX, and Apache ZooKeeper.
AIxCC
Multi-Agent Triage for Static-Analysis Rules — Amazon
Automated triage of production static-analysis findings at
90% accuracy
, cutting validation time for new rules from
10–15 days to 2–3 days
at ~4 minutes and $0.50 per finding.
BuildBench: Benchmarking LLM Agents on Compiling Real-World Open-Source Software
A benchmark and agent for compiling diverse, real-world C/C++ open-source projects with LLMs.
Zehua Zhang
,
Ati Priya Bajaj
,
Divij Handa
,
Siyu Liu
,
Arvind S Raj
,
Hongkai Chen
,
Hulin Wang
,
Yibo Liu
,
Zion Leonahenahe Basque
,
Souradip Nath
,
Vishal Juneja
,
Nikhil Chapre
,
Tiffany Bao
,
Yan Shoshitaishvili
,
Adam Doupé
,
Chitta Baral
,
Ruoyu Wang
PDF
Cite
×