I am currently a Student Researcher at the School of Computing, National University of Singapore. I am supervised by Prof. Mong-Li Lee and Prof. Wynne Hsu at the Center for Trusted Internet and Community (CTIC). I also work with Dr. Hao Fei, Dr. Shengqiong Wu, and Dr. Bobo Li. I am grateful for their guidance, support, and the many discussions that continue to shape my research.
My long-term goal is to build AI with an evolving understanding of the world and the initiative to act within it. I am interested in how AI can accumulate knowledge from individual experience, exercise judgment in unfamiliar situations, and pursue meaningful goals under uncertainty. I see learning, judgment, agency, and collaboration as distinct but interacting capacities, and seek to understand how they can develop together.
An encounter can become a memory, a new explanation, or a different way of approaching a problem. I am interested in how AI consolidates experience into knowledge and reasoning strategies that remain open to revision. Different histories should be able to support different perspectives and expertise, while allowing systems to recognize when familiar lessons no longer apply.
Acting in a changing world requires sensitivity to both new evidence and the limits of existing beliefs. I study how AI can distinguish observations from assumptions, assess conflicting information, and reconsider its interpretation of a situation. A central challenge is knowing when experience offers a useful guide and when a surprising observation calls for a new explanation.
Agency involves forming purposes, recognizing possibilities, and taking initiative before every uncertainty is resolved. Real-world goals do not come with predefined problems; an agent must continually work out what is worth solving, pursuing, or changing as the world unfolds. I aim to develop AI that can propose goals, seek opportunities, and sustain or redirect its efforts. This raises questions of WHY → WHEN → WHAT → WHETHER → HOW. Confidence to explore and judgment about when to pause develop together.
Individual histories create differences in what people and AI know, expect, and intend. I am interested in how interaction makes these differences intelligible, supports the exchange of experience, and enables coordinated action. Collaboration should allow participants to question assumptions, negotiate goals, and learn from one another while preserving the distinct knowledge each brings.
Professional expertise is more than knowing facts: it involves knowing how to investigate a problem, weigh evidence, and decide what to do next.
We argue that AI should develop such expertise by coupling fluid generation with crystallized reasoning: explicit, executable structures that humans and AI can jointly refine through sustained interaction. New possibilities can be explored, tested, and consolidated into reusable methods, allowing different histories of interaction to shape distinct approaches to judgment within shared professional standards.
Exchanging fluent messages is not the same as coordinating knowledge, intentions, and actions.
We argue that AI communication should be designed around the internal representations and behavioral dependencies that make collaboration possible, rather than inherited from human conversation. The deeper challenge is to build and train models for collaboration from the outset, so that communication updates shared understanding, preserves responsibilities, and connects joint intentions to coordinated action.
Technical report · Coming soon
An open research framework for long-term human-AI collaboration. AICO combines personal and expert alignment with relationship context and dialogue strategies, using feedback to guide how AI responds and cooperates over time.
Pengcheng Zhou, X. Chen, X. Liu, X. Zhang, Y. Zhang, X. Yang, X. Xu, S. Li
Manuscript under review
P. Guo*, Pengcheng Zhou*, Y. Jian, S. Chen, Z. Yang, L. Zhou
Findings of EMNLP 2026
Pengcheng Zhou, P. Guo, S. Chen, M. Zhao, Z. Yang, L. Zhou
arXiv preprint
Pengcheng Zhou, H. Li, Z. Nie, J. Chen, Q. Gong, W. Zhang, C. Yu
arXiv preprint
Pengcheng Zhou, Z. Fang, Z. Yang, Z. Zhou, L. Zhou
IEEE Transactions on Information Forensics and Security
Z. Fang*, Pengcheng Zhou*, Z. Yang, Z. Zhou, L. Zhou
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
* Equal contribution.
Current
M.Sc. in Computer Engineering
Jun 2024
B.Eng. in Internet of Things Engineering
Jun 2024
B.Sc. in Internet of Things, First Class Honours
Joint programme with BUPT; dual degrees awarded.
Sep 2025 - Present
School of Computing
Advised by Prof. Mong-Li Lee and Prof. Wynne Hsu.
Jul 2023 - 2026
Pervasive Human-Computer Interaction Laboratory
Advised by Prof. Yuanchun Shi and Prof. Chun Yu.
Jul 2023 - Jul 2025
Mobile Internet Security Technology National Project Laboratory
Advised by Prof. Zhongliang Yang.
Mar 2022 - Dec 2022
Pattern Recognition and Intelligent Systems Laboratory
Advised by Prof. Junli Yang and Prof. Ming Wu.
PC Member: AAAI 2027. Reviewer: ACM Multimedia 2026, NeurIPS 2026, AAAI 2026, ICASSP 2025.
Email: zhoupengcheng0219@gmail.com