I am a Ph.D. student in the Department of Computer Science at the University of Maryland, College Park, advised by Prof. Heng Huang.

Before joining UMD, I received my M.S. degree from Shanghai Jiao Tong University, where I was advised by Prof. Xiaolin Huang. I received my B.E. degree in Automation from Xi’an Jiaotong University.

My research interests include machine learning, LLM agents, AI safety, and efficient LLM systems, with a focus on building safer and more efficient AI systems. I am always open to academic discussions and collaborations; please feel free to reach out via email.

🔥 News

• 2026.08: I started my Ph.D. study in Computer Science at the University of Maryland, College Park.
• 2026.07: 🎉 Our paper “InjecMEM: Memory Injection Attack on LLM Agent Memory Systems” was accepted to COLM 2026!
• 2026.06: 🎉 Our paper “Stochastic Optimal Control Sampling for Diffusion Inverse Problems” was accepted to ECCV 2026!
• 2026.01: 🎉 Our paper “RAIN-Merging: A Gradient-Free Method to Enhance Instruction Following in Large Reasoning Models with Preserved Thinking Format” was accepted to ICLR 2026 as an Oral presentation!
• 2025.05: 🎉 Our paper “Primphormer: Efficient Graph Transformers with Primal Representations” was accepted to ICML 2025!
• 2025.01: 🎉 My first first-author paper, “Simulating Training Dynamics to Reconstruct Training Data from Deep Neural Networks,” was accepted to ICLR 2025!

📝 Selected Publications

COLM 2026

InjecMEM: Memory Injection Attack on LLM Agent Memory Systems

Hanling Tian, Gengyu Zhang, Zeyang Sha, Jingying Wang, Yuhang Liu, Zhehao Huang, Kun Yang, Xiaolin Huang

Conference on Language Modeling (COLM), 2026

OpenReview   arXiv   Code

We introduce InjecMEM, a memory injection attack against LLM agents with persistent memory. A single malicious interaction can poison the memory system and influence the agent’s responses to future benign queries.

ICLR 2025

Simulating Training Dynamics to Reconstruct Training Data from Deep Neural Networks

Hanling Tian, Yuhang Liu, Mingzhen He, Zhengbao He, Zhehao Huang, Ruikai Yang, Xiaolin Huang

International Conference on Learning Representations (ICLR), 2025

OpenReview   Code   Slides   Poster

We propose SimuDy, which reconstructs training data from trained deep neural networks by explicitly simulating the training dynamics from model initialization to the final trained parameters.

📚 Other Publications

• Stochastic Optimal Control Sampling for Diffusion Inverse Problems
  Jie Zhang, Youmei Qiu, Hanling Tian, Jingyuan Zhang, Xiang Yin, Xiaolin Huang.
  European Conference on Computer Vision (ECCV), 2026.

• RAIN-Merging: A Gradient-Free Method to Enhance Instruction Following in Large Reasoning Models with Preserved Thinking Format
  Zhehao Huang, Yuhang Liu, Baijiong Lin, Yixin Lou, Zhengbao He, Hanling Tian, Tao Li, Xiaolin Huang.
  International Conference on Learning Representations (ICLR), 2026. Oral.

• Primphormer: Efficient Graph Transformers with Primal Representations
  Mingzhen He, Ruikai Yang, Hanling Tian, Youmei Qiu, Xiaolin Huang.
  International Conference on Machine Learning (ICML), 2025.

📖 Education

• 2026 – Present, Ph.D. in Computer Science, University of Maryland, College Park
  Advisor: Prof. Heng Huang

• 2023 – 2026, M.S., Shanghai Jiao Tong University
  Institute of Image Processing and Pattern Recognition
  Advisor: Prof. Xiaolin Huang

• 2019 – 2023, B.E. in Automation, Xi’an Jiaotong University

💻 Experience

• Microsoft Research Asia (MSRA) — Research Intern, DKI Group
  Research on multimodal learning, layout understanding, and generative models.

• Ant Group — Research Intern, AI Safety
  Research on the safety and security of LLM agents and memory systems.

📝 Academic Service

• Reviewer: ICML 2026, COLM 2026, NeurIPS 2026, MICCAI 2026 Workshop AMPLIFAI
• Program Committee: AAAI 2027

🌎 Visitors