About Me

LLM Agents · Agent Safety · Generative Vision

I am a fourth-year undergraduate student majoring in Software Engineering at the School of Computer and Information Science, Southwest University. My research focuses on LLM agents, agent safety, and generative vision. I am interested in how agents reason, collaborate, and interact, and how to evaluate and improve the safety and reliability of these systems. I explore generative models for visual content creation.

I am currently supervised by Prof. Xianfeng Han (Google Scholar). I am also a research intern at The Chinese University of Hong Kong, Shenzhen, in the group led by Prof. Haizhou Li (IEEE Fellow) (Google Scholar). My internship research is guided by Dr. Xueyi Zhang (Google Scholar), a Postdoctoral Research Fellow at the National University of Singapore (NUS). I am actively seeking opportunities for MPhil programs starting in Fall 2027. Please feel free to contact me directly via email.

Research Intern @ CUHK-Shenzhen MPhil Fall 2027 Southwest University

Publications

Under ARR
EXAM framework: extremist agent modeling, injection strategies, and four social simulation scenarios

When Extremists Enter LLM Societies: Investigating Extreme Behavior in Multi-Agent Systems

Guo Chen, Huaijin Xie, Ziwen Li, Reed Li, Yu Lu, Haibo Shi, Bingbing Xu, Junjie Huang.

arXiv preprint · 2026

We introduce EXAM, a configurable framework for modeling and injecting extremist agents into LLM-based societies, and study their effects across four social simulation scenarios.

Under WWW
ForestBench framework: task selection, collaboration graph construction, and reference-forest evaluation

ForestBench: A Unified Graph Framework for Evaluating Multi-Agent Collaboration

Guo Chen, Ziwen Li, Reed Li, Yu Lu, Haibo Shi, Bingbing Xu, Junjie Huang.

arXiv preprint · 2026

We introduce ForestBench, a unified collaboration-graph benchmark with 844 queries from seven public datasets. It evaluates multi-agent collaboration against verified reference forests without additional LLM inference.

WISA
NGM-RAG paper thumbnail

NGM-RAG: Neural Graph Matching based Retrieval-Augmented Generation

Guo Chen, Ziwen Li, Maolin Zheng, Qi Cao, Junjie Huang, Tao Jia.

We propose NGM-RAG, a unified graph-based RAG framework that adaptively identifies key nodes and relations. It outperforms existing RAG methods on multi-hop QA and long-context summarization with lower retrieval and inference costs.

Manuscripts Under Review

Text-to-360-Degree Panoramas AAAI Phase 2
Agent Unlearning AAAI Phase 2
OpenClaw Safety WWW 2027 · Under Review

Honors and Awards

  • 2026 EMNLP 2027 Great Reviewer.
  • 2026 National First Prize, Chinese Collegiate Computing Design Competition, China.
  • 2026 Provincial-Level Project, Chongqing Municipal Undergraduate Innovation and Entrepreneurship Training Program, China.
  • 2025 Provincial First Prize, National Undergraduate Mathematical Contest in Modeling, China.
  • 2025 Honorable Mention, Mathematical Contest in Modeling / Interdisciplinary Contest in Modeling, International.
  • 2025 Finalist, Certified Cup International Mathematical Modeling Competition, China.
  • 2025 Second-Class Scholarship, Southwest University.
  • 2024 Third-Class Scholarship, Southwest University.

Education and Experience

B.Eng. in Software Engineering Southwest University, Chongqing, China · GPA: 3.67 / 5.0
Research Intern The Chinese University of Hong Kong, Shenzhen, Shenzhen, China
Industrial Agent Designer LongRangeOrder
Project Leader Chongqing CAS Automotive Software Innovation Center · Indoor autonomous navigation modeling with IsaacSim & ROS; Gazebo simulation and 2D SLAM with Gmapping.

Community Contributions

I actively contribute to the open-source community, including: