Mathematical modeling involves the cognitively demanding process of translating real-life situations into mathematical notations and is a fundamental skill for students pursuing science, technology, engineering, and mathematics (STEM). Cultivating mathematical modeling through collaborative learning can be particularly effective, but orchestrating such collaborative learning tasks requires significant efforts from teachers to oversee group discussions. In addition, students may not receive effective training from the group discussions, due to reasons such as mental barriers (e.g., anxiety) in peer interactions.
Through MathVC, the project aims to increase the opportunities for students to receive effective mathematics education. The project will also use the MathVC platform as a lens to understand the opportunities and risks of Generative AI techniques and provide insights for future researchers and educators.
An example is shown below.
Updates
- 04/2025: We will be organizing the Math EdVenture Camp at GMU during July 7-11, 2025. Go Register Here!
- 02/2025: We started a pilot study inviting middle-school students from Virginia to provide feedback on the MathVC prototype!
- 02/2025: The MathVC system prototype paper is accepted to the AAAI Workshop on AI4Edu!
- 04/2024: ArXiv version is online. Please check out our preprint.
Publications
(^ denotes equal contributions)
@article{yue2024mathvc,
title={Mathvc: An llm-simulated multi-character virtual classroom for mathematics education},
author={Yue, Murong and Lyu, Wenhan and Mifdal, Wijdane and Suh, Jennifer and Zhang, Yixuan and Yao, Ziyu},
journal={arXiv preprint arXiv:2404.06711},
year={2024}
}
@article{kumar2025can,
title={Can LLMs Simulate Personas with Reversed Performance? A Benchmark for Counterfactual Instruction Following},
author={Kumar, Sai Adith Senthil and Yan, Hao and Perepa, Saipavan and Yue, Murong and Yao, Ziyu},
journal={arXiv preprint arXiv:2504.06460},
year={2025}
}
@article{yue2025efficient,
title={Efficient but Vulnerable: Benchmarking and Defending LLM Batch Prompting Attack},
author={Yue, Murong and Yao, Ziyu},
journal={arXiv preprint arXiv:2503.15551},
year={2025}
}
Presentations
Outreach Activities
Team Members
Faculty
Dr. Ziyu Yao (PI)
George Mason University, Computer Science
George Mason University, Math Education
William & Mary, Computer Science
Advisory Board
Coming Soon!PhD Student Leads
George Mason University, Computer Science
William & Mary, Computer Science
Other Student Researchers
Wijdane Mifdal
Undergraduate Student at GMU CS
Sai Adith Senthil Kumar
Undergraduate Student at GMU CS
Hao Yan
PhD Student at GMU CS
Gallery
Coming soon!Acknowledge
We were grateful to receive support from the National Science Foundation (2418580, 2418582), Microsoft Accelerating Foundation Models Research program (https://www.microsoft.com/en-us/research/collaboration/accelerating-foundation-models-research/), Department of Computer Science at GMU, and William & Mary for conducting this research. We also appreciate computing resources from the Office of Research Computing (https://orc.gmu.edu) and the assistance from the Institutional Review Board (IRB) at GMU and William & Mary. Undergraduate students in the team have also received sponsorship from the Office of Student Creative Activities and Research (OSCAR) at GMU.
We thank Dr. Anthony Eamonn Kelly (GMU, Educational Psychology) and Dr. Melissa A. Gallagher (University of Houston, Department of Curriculum & Instruction) for early discussions about the idea, and students in the GMU NLP group (https://nlp.cs.gmu.edu/) for their thoughtful comments.