Education
Peking University · School of Computer Science
PhD student
Wuhan University · School of National Cyber Security
Undergraduate student
PhD student · Peking University
Fengyuan Ran
冉丰源
Secure, adaptive, and intelligent systems.
I am a PhD student in the School of Computer Science at Peking University. My research lies at the intersection of multi-agent systems, agentic reinforcement learning, TrustWorthy AI, and AI for software engineering, with a focus on building trustworthy, adaptive, and collaborative intelligent systems.
Experience
JD.com · 京东
TGT 顶尖青年技术天才计划
Intern · Tech Genius Team
DiDi
Intern
Latest News
One paper on multi-agent systems accepted by NeurIPS 2026.
Two papers on federated graph learning and temporal point processes accepted by NeurIPS 2025.
Paper on VR security accepted by Ubicomp 2025.
Paper on LLM jailbreaks accepted by USENIX Security 2025.
Paper on temporal point processes accepted by AAAI 2025.
Honors & Awards
NeurIPS 2026 Top Reviewer
武汉大学十大杰出青年（学生）
全校 10 人 · 获奖比例 10/59,774 ≈ 0.017%
武汉大学三好学生标兵
全校 60 人 · 获奖比例 60/59,774 ≈ 0.1%
榜样珞珈
全校 10 人 · 本科生获奖比例 10/29,450 ≈ 0.034%
雷军卓越奖学金
奖金 10 万元 · 全校 10 人，其中本科生 4 人 · 本科生获奖比例 4/29,450 ≈ 0.014%
国家奖学金
全国获奖比例 0.2%
Selected publications
View all publications
RIGOR: Risk-Gated Topology Adaptation for Robust LLM Multi-Agent Reasoning
Fengyuan Ran first author · NeurIPS 2026 · CCF-A
MOTION: Multi-Sculpt Evolutionary Coarsening for Federated Continual Graph Learning
Fengyuan Ran co-first author · NeurIPS 2025 · CCF-A
TPP-SD: Accelerating Transformer Point Process Sampling with Speculative Decoding
Fengyuan Ran co-first author · NeurIPS 2025 · CCF-A
HT-Auth: Unobtrusive VR Headset Authentication via Subtle Head Tremors
Fengyuan Ran 2nd author · Ubicomp / IMWUT 2025 · CCF-A
PAPILLON: Efficient and Stealthy Fuzz Testing-Powered Jailbreaks for LLMs
Fengyuan Ran 4th author · USENIX Security Symposium 2025 · CCF-A
A Plug-and-Play Bregman ADMM Module for Inferring Event Branches in Temporal Point Processes
Fengyuan Ran 5th author · AAAI 2025 · CCF-A
The Readability Spectrum: Patterns, Issues, and Prompt Effects in LLM-Generated Code
Fengyuan Ran 2nd author · under review
Personal archiveA visual note
A little more
Research should be rigorous, useful, and open to conversation.
My work sits between machine learning, security, and intelligent systems. I enjoy turning hard technical questions into methods that are both principled and practical.
Start a conversation