Yuejin Xie

About Me

I am a master’s student at Tsinghua University, advised by Prof. Yujiu Yang. I am also a research intern at Shanghai AI Lab.

My research focuses on the safety of LLM-based agents: how agents perceive risk, use tools, and act in long-horizon environments. I build benchmarks, evaluations, and guardrails for agent safety, proactive risk awareness, and alignment.

Research Interests

• Agent safety: trajectory-level evaluation, diagnosis, and guardrails
• Tool-use safety: safe interaction between language-model agents and external tools
• Proactive risk awareness: recognizing risks before harm happens
• Alignment: scalable safety training and deployment for interactive agents

Education

• Tsinghua University, M.S. student in Electronic Information, Sep. 2026 - present. Advised by Prof. Yujiu Yang.
• Huazhong University of Science and Technology, B.E. in Electronic Engineering, Sep. 2022 - Jun. 2026. GPA: 4.51/5.00; Rank: 1/30 (Advanced Class).

Internship

• Research Intern, Shanghai AI Lab, Shanghai, China, Dec. 2025 - present.

News

• [Sep. 2026] Started my master’s program at Tsinghua University.
• [Sep. 2026] ClashBench is now available on arXiv.
• [Sep. 2026] PaSBench-Video was accepted to the EMNLP 2026 Main Conference.
• [Jun. 2026] PaSBench-Video is now available on arXiv.
• [May 2026] AgentDoG 1.5 is now available on arXiv.
• [Apr. 2026] ATBench is now available on arXiv.
• [Mar. 2026] Code2Math is now available on arXiv.
• [Nov. 2025] ToolSafety was published at EMNLP 2025.

Publications