Tianxing He

Tianxing He is an Assistant Professor at Tsinghua University's Institute for Interdisciplinary Information Sciences (IIIS), also known as the Yao Class, which he joined in September 2024. He also works part-time at Shanghai Qi Zhi Institute where he recruits for research-oriented positions and video game development roles. His academic journey includes a postdoc at the University of Washington under Yulia Tsvetkov, a PhD at MIT supervised by James Glass, and bachelor's and master's degrees at Shanghai Jiao Tong University in the ACM honored class under Prof. Kai Yu.

Dr. He's research spans AI safety and AI-driven simulation with particular focus on large language models. His work examines LLM capabilities in cryptography (AICrypto benchmark), security vulnerabilities (jailbreaking techniques), multi-agent system integrity, and socially impactful applications like vaccine hesitancy simulation (VACSIM framework). He has developed visualization tools like MiniAgents-Pro using Unity for generative agent simulation and created watermarking techniques such as SemStamp and k-SemStamp for detecting machine-generated text.

His publication record shows consistent contributions to top-tier conferences including ACL, EMNLP, NeurIPS, and ICLR, with recent work focusing on evaluating LLM capabilities across diverse domains. The research demonstrates sophisticated understanding of both the technical aspects of language models and their broader societal implications, particularly in security and ethical considerations.

Scientific Awards:

• The Best Paper Award at the Efficient Natural Language and Speech Processing Workshop (NeurIPS ENLSP 2022) for "PCFG-based Natural Language Interface Improves Generalization for Controlled Text Generation"

Dr. He actively mentors students and recruits research interns, PhD candidates, and post-bachelor researchers. He plans to teach an undergraduate NLP course in Fall 2025 and organizes the IIIS AI Safety Lunch, a monthly gathering in Beijing for those working on AI safety. His recruitment efforts target students interested in LLM social simulation, AI+game development, and AI safety research, emphasizing the need for strong self-motivation and research passion.

He leads research initiatives including the IIIS AI Safety Lunch series and has developed multiple visualization and simulation frameworks. His work sits at the intersection of AI safety, language modeling, and practical applications, with team collaborations spanning multiple institutions including MIT, University of Washington, and Shanghai Jiao Tong University.