I am a master’s student in Computer Science at the Institute of Computing Technology, Chinese Academy of Sciences (ICT, CAS). My research focuses on multimodal large language models, AI agents, and reinforcement learning.
I am particularly interested in training agents to solve long-horizon tasks and improving the reliability and alignment of multimodal models. I study how reinforcement learning, better task environments, and robust evaluation can help models reason and act more effectively.
I am currently a research intern at JD.com, working on multimodal safety models and long-horizon code agents.
Research Interests
Multimodal Large Language Models
Long-Horizon AI Agents
Reinforcement Learning
Education
M.S. in Computer ScienceInstitute of Computing Technology, CAS
B.E. in Software EngineeringBeijing University of Posts and Telecommunications
Multimodal safety foundation models and long-horizon code agents
Developing a multimodal safety model end to end: 1.2M+ data construction, model distillation, SFT, GRPO post-training, evaluation, and deployment.
Building an automated pipeline from real-world CVEs to Agent-RL tasks, including Docker environments, benchmarks, reward design, and trajectory generation.
2025.09 – Present
Institute of Computing Technology, CAS
M.S. in Computer Science and Technology · GPA 3.83/4.0
2021.09 – 2025.06
Beijing University of Posts and Telecommunications
B.E. in Software Engineering · GPA 3.76/4.0 · Rank 5/170
Honors & Awards
Beijing Outstanding Graduate
Second Prize, National College Student Information Security Contest
Third Prize (Beijing), Chinese Collegiate Computing Competition
Second Prize, Asia-Pacific Mathematical Contest in Modeling (APMCM)
National Third Prize, Chinese Mathematics Competitions
University scholarships and Outstanding Student awards, BUPT
Beyond Research
Beyond research, I enjoy playing the violin, table tennis, swimming, and debate. These interests give me space to stay curious, focused, and connected with people outside the lab. I am always happy to meet like-minded friends—whether you would like to exchange research ideas, play a game of table tennis, make music together, or simply have an interesting conversation, please feel free to reach out.