About me

Hi, I’m Xiwen (Christina) Wei, a Ph.D. student in the Department of Electrical and Computer Engineering at The University of Texas at Austin (UT Austin), expecting to graduate in Spring 2027. My research interests include multimodal foundation models, LLM agents, and continual learning. I previously interned at AMD and Zillow.

Before joining UT Austin, I received my B.S. degrees in Electrical Engineering from University of Michigan, Ann Arbor, and in Electrical and Computer Engineering from Shanghai Jiao Tong University.

I’m seeking full-time opportunities starting in 2027 in machine learning, with a focus on LLM agents and multimodal foundation models. Please feel free to reach out!

News

• [May 2026] Happy to join Zillow as an Applied Scientist Intern, working on LLM-based user simulation for conversational agents!
• [October 2025] MoDE was accepted by NeurIPS 2025. See you in San Diego!
• [May 2025] Happy to join AMD RAD Team as a PhD Research Associate Intern, working on fine-tuning large-scale unified multimodal foundation models!
• [November 2024] Online-LoRA was accepted by WACV 2025. See you in Tucson!
• [October 2024] Two papers were accepted by 2nd Workshop on Regulatable ML and Continual FoMo Workshop @NeurIPS2024. See you in Vancouver!

Selected Projects

LLM Agents and Post-Training

Turing²: Scaling Turing-Style Evaluation to Optimize Human Simulation in Dialogue
Internship at Zillow, 2026
Developed a human-grounded framework for co-evolving user simulators and LLM-based evaluators, using human judgments and rationales to improve dialogue simulation.

Learning to Clarify Ambiguous Requirements in Coding Agents
Developed GRPO-based reinforcement learning methods to train coding agents to clarify ambiguous requirements through multi-turn interaction. Studied reward design and evaluated coding success under bounded clarification budgets.

Simulating Rumor Spreading in Social Networks using LLM Agents
Tianrui Hu, Dimitrios Liakopoulos, Xiwen Wei, Radu Marculescu, Neeraja J. Yadwadkar
AAAI Workshop on WMAC, 2025
[Paper] [Code]

Multimodal Foundation Models and Continual Learning

Mitigating Intra- and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models
Xiwen Wei, Mustafa Munir, Radu Marculescu
NeurIPS, 2025
[Paper] [Code]

Pareto LoRA: Mitigating Modality Imbalance in Unified Multimodal Models via Pareto-Optimal Gradient Integration
Xiwen Wei, Mark Nutter, Madhusudhanan Srinivasan, Radu Marculescu
arXiv, 2026
[Paper]

Online-LoRA: Task-free Online Continual Learning via Low Rank Adaptation
Xiwen Wei, Guihong Li, Radu Marculescu
WACV, 2025
[Paper] [Code]

Fuel Gauge: Estimating Chain-of-Thought Length Ahead of Time in Large Multimodal Models
Yuedong Yang, Xiwen Wei, Mustafa Munir, Radu Marculescu
CVPR, 2026
[Paper]

Fairness Implications of Machine Unlearning: Bias Risks in Removing NSFW Content from Text-to-Image Models
Xiwen Wei, Guihong Li, Radu Marculescu
NeurIPS Workshop on Regulatable ML, 2024
[Paper]