Pan Chen
Toronto, Ontario, Canada
Last updated: October 2, 2026
Research Interests
Large language models; in-context learning and reasoning; multimodal learning; AI for science; human–AI interaction.
Education
University of Toronto
2022 – Present
PhD in Computer Science; Vector Institute affiliation
Expected graduation: 06/2028
Supervisor: Alán Aspuru-Guzik. Committee: Nicholas Papernot and Michael Liut.
University of Toronto
2018 – 2022
BSc in Computer Science and Statistics. Research advisor: Joseph Jay Williams.
Research & Industry Experience
Applied Scientist II (Intern), Amazon
2026
- Led a research project in the Alexa+ team to improve the visual capabilities of multimodal large language models (MLLMs) through supervised fine-tuning (SFT) and GRPO, improving accuracy by 6.5 percentage points while maintaining inference latency.
- Research manuscript in preparation.
Doctoral Research, University of Toronto & Vector Institute
2022 – Present
- Research in-context learning and scientific reasoning with large language models.
- Developed Schema-Activated In-Context Learning (SA-ICL) to transfer structured reasoning templates to new questions, achieving performance gains of up to 36.19% on GPQA chemistry and physics questions with a single high-quality demonstration.
- Co-authored research on language model-assisted reading and writing (TreeReader, TreeWriter) and computer vision for high-throughput materials synthesis.
- Contributed to the review paper Molecular Knowledge Representations in the Era of Artificial Intelligence, covering molecular and chemical knowledge representations for AI.
Research Engineer, Dynamic Graphics Project Lab, University of Toronto
2022 – 2023
- Led software development and machine learning design for an adaptive experimentation platform deployed in more than 25 courses, in collaboration with CMU and UNC; our team won the XPRIZE Digital Learning Challenge.
Research Scholar, Data Sciences Institute, University of Toronto
2022
- Built a platform for collecting data from third-party websites and running data analysis through a web interface. Advisor: Michael Liut.
Software Developer Co-op, Infrastructures for Information (i4i)
2020 – 2021
Teaching Experience
University of Toronto, Computer Science & Statistics
2021 – Present
Instructor (2025 – Present); Teaching Assistant (2021 – 2024)
- Instructor for CSC148 (Introduction to Computer Science), CSC207 (Software Design), and CSC309 (Programming on the Web).
- Developed peer-feedback and grading-consistency tools for Software Design and a software-vulnerability assignment for Programming on the Web.
- Received the inaugural DCS Graduate Teaching Award (2026), recognizing exemplary teaching.
Selected Publications & Manuscripts
When Should A Model Trust Its Own Examples? Asymmetry in Self-Generated In-Context Learning
P. Chen, M. Wang, S. Chen, et al.
Manuscript; planned submission to ACL Rolling Review (October 2026).
Schema for In-Context Learning
P. Chen, S. Chen, M. Wang, et al.
arXiv:2510.13905, 2025 (revised 2026).
Molecular Knowledge Representations in the Era of Artificial Intelligence
Z. Zhang, J. Bai, Y. Nakamura, A. Wang, S. X. Leong, S. Zhang, P. Chen, et al.
Accepted; forthcoming. Preprint: ChemRxiv, 2026.
Computer Vision for High-Throughput Materials Synthesis: A Tutorial for Experimentalists
M. A. Gaidimas, A. Mandal, P. Chen, et al.
Digital Discovery, 5(2), 510–522, 2026.
TreeWriter: AI-Assisted Hierarchical Planning and Writing for Long-Form Documents
Z. Zhang, F. Du, X. Liu, P. Chen, et al.
arXiv:2601.12740, 2026.
TreeReader: A Hierarchical Academic Paper Reader Powered by Language Models
Z. Zhang, P. Chen, F. Du, et al.
IEEE VL/HCC, 2025.
Platform-based Adaptive Experimental Research in Education: Lessons Learned from The Digital Learning Challenge
I. Musabirov, M. Reza, H. Song, S. Moore, P. Chen, et al.
International Learning Analytics and Knowledge Conference (LAK), 2025.
Does the Medium Matter? An Exploration of Voice-Interaction for Self-Explanations
A. Zavaleta Bernuy, N. Sibia, P. Chen, et al.
ACM Designing Interactive Systems (DIS), 86–101, 2024.
Opportunities for Adaptive Experiments to Enable Continuous Improvement in Computer Science Education
I. Musabirov, A. Zavaleta Bernuy, P. Chen, et al.
Western Canadian Conference on Computing Education (WCCCE), 2024.
Behavioral Consequences of Reminder Emails on Students' Academic Performance: A Real-World Deployment
R. Ye, P. Chen, Y. Mao, et al.
ACM SIGITE, 16–22, 2022.
Parameter Efficient Dendritic-Tree Neurons Outperform Perceptrons
Z. Han, E. Gorobets, P. Chen.
arXiv:2207.00708; presented at the ICML Dynamic Neural Networks Workshop, 2022.
Statistical Consequences of Dueling Bandits
N. Saxena, P. Chen, E. Liu.
EDM Workshop on Reinforcement Learning for Education; arXiv:2111.00870, 2021.
Technical Skills
Research engineering: Python, PyTorch, TensorFlow, Hugging Face Transformers.
LLM training & evaluation: Multimodal post-training (SFT, GRPO), in-context learning, benchmarking.
Agentic coding & harness engineering: Codex/Claude Code workflows, repository instructions, tool integration, and automated validation.
Awards, Research Funding & Training
- Vector Research Grant; Department of Computer Science Graduate Program Award.
- Doctoral research supported by the Google–NSERC Industrial Research Chair.
- Carnegie Mellon University Summer School, full scholarship (2023); mentors and collaborators: John Stamper and Steven Moore.