prof_pic.JPG

Yuzhou Nie (聂宇舟)

Meta Superintelligence Lab. rucnyz@gmail.com.

1 Meta Way, Menlo Park, CA 94025

I am a Senior AI Research Scientist at Meta Superintelligence Lab, working on post-training. Side interest in inference system optimization.

I received my PhD in Computer Science from the University of California, Santa Barbara in 2026, advised by Prof. Wenbo Guo, working closely with Prof. Bo Li and Prof. Dawn Song. I received my bachelor’s degree in Statistics from Renmin University of China, where I was advised by Prof. Hongteng Xu.

Previously, I was a founding engineer at Virtue AI, leading large-scale post-training of cybersecurity-specialized LLMs.

Research

News

Jul 6, 2026 Joined Meta Superintelligence Lab as a Senior AI Research Scientist!
Jun 25, 2026 Selected for the Anthropic Fellows program.
Apr 30, 2026 Three papers accepted to ICML 2026: OpenSage, BlueCodeAgent, and THETA!
Feb 15, 2026 VulnLLM-R-7B reached #2 most-downloaded model on HuggingFace this month! Repo: ucsb-mlsec/VulnLLM-R.
Oct 28, 2025 Two papers accepted to NeurIPS 2025: SeCodePLT and OWL. See you in San Diego!
Aug 21, 2025 One paper (AgentVigil) accepted to EMNLP 2025!
Jul 12, 2025 One paper accepted to COLM 2025, see you in Montréal!
Jun 9, 2025 We will be giving a tutorial on multimodal generation evaluation at CVPR 2025 in Nashville!
Apr 23, 2025 We release the new agent framework OWL, achieving GAIA benchmark score of 69.09%, ranking #1 among open-source frameworks. Check the technical report here.
Apr 1, 2025 Joined Virtue AI as a founding engineer, working on cybersecurity-specialized LLMs.
Feb 13, 2025 Our team (ForesAIght) won the 1st place in safety track of Berkeley LLM Agents Hackathon! Check out our report, code for more details.
Sep 23, 2024 I received the UCSB Research Excellence Award!
Jul 21, 2024 I received the FAR AI Research Grant to support my research from 2024-2025!
Mar 21, 2024 I will join UCSB as a PhD student in Computer Science.
Jan 16, 2023 I will join Microsoft Research Asia as a Research Intern.