Penglin Cai

I am Penglin Cai (蔡鹏霖), a senior undergraduate majoring in Artificial Intelligence at Yuanpei College, Peking University. I am a member of the Tong Class, an honorary pilot class in Artificial Intelligence founded by Prof. Song-Chun Zhu.

I will begin my Ph.D. studies since Fall 2025 supervised by Prof. Zongqing Lu, under the guidance of whom I have been honored to work on research since July 2022.

My research interests lie in Reinforcement Learning, Embodied AI, as well as the general agents acquiring skills and accomplishing tasks in the open world.

Email  /  CV  /  Scholar  /  Github

profile photo

Publications and Preprints

I am interested in Reinforcement Learning, Embodied AI, as well as the general agents acquiring skills and accomplishing tasks in the open world. My past and on-going research problems vary from creative agents, offline model-based reinforcement learning, to applying pre-trained VLA models to dexterous robotic hands with new modalities.

Creative Agents: Empowering Agents with Imagination for Creative Tasks
Penglin Cai*, Chi Zhang*, Yuhui Fu, Haoqi Yuan, Zongqing Lu
The 41st Conference on Uncertainty in Artificial Intelligence (UAI), Poster, 2025
arXiv / talk / project page / code / blog / bibtex

Creative tasks are challenging for open-ended agents, where the agent should give novel and diverse task solutions. We propose creative agents with the ability of imagination and introduce several variants in implementation. We benchmark creative tasks in the challenging open-world game Minecraft and propose novel evaluation metrics utilizing GPT-4V. Creative agents are the first AI agents accomplishing diverse building creation in Minecraft survival mode.

Plan4MC: Skill Reinforcement Learning and Planning for Open-World Minecraft Tasks
Haoqi Yuan, Chi Zhang, Hongcheng Wang, Feiyang Xie, Penglin Cai, Hao Dong, Zongqing Lu
NeurIPS Foundation Models for Decision Making Workshop, 2023
workshop paper / arXiv / project page / code / bibtex / blog

Plan4MC is a multi-task agent in the open world Minecraft, solving long-horizon tasks via planning over basic skills. It acquires three types of fine-grained basic skills through reinforcement learning without demonstrations. With a skill graph pre-generated by the Large Language Model, the skill search algorithm plans and interactively selects policies to solve complicated tasks. Plan4MC accomplishes 40 diverse tasks in Minecraft and unlocks Iron Pickaxe in the Tech Tree.

Invited Talks

Academic Services

  • Reviewer, Multi-Agent Systems in the Era of Foundation Models: Opportunities, Challenges and Futures workshop @ ICML 2025
  • Reviewer, Open-World Agents workshop @ NeurIPS 2024
  • Reviewer, Foundation Models for Decision Making workshop @ NeurIPS 2023

Awards and Scholarships

  • Outstanding Graduate of Yuanpei College, Peking University, Summer 2025
  • Zheng Geru Outstanding Student Scholarship, Peking University, 2023-2024 Academic Year
  • Merit Student, Peking University, 2023-2024 Academic Year
  • The Third Prize of Peking University Scholarship, Peking University, 2022-2023 Academic Year
  • Award for Academic Excellents, Peking University, 2022-2023 Academic Year
  • The First Prize of Freshman Scholarship, Peking University, Fall 2021

The template of this website is borrowed here.