Personal academic website of Minghui Chen, a Predoctoral Staff Associate in the Decision, Risk, and Operations division at Columbia Business School, fortunately advised by Prof. Hongseok Namkoong. Exploring reinforcement learning with LLM agents.
Read More ↓Minghui (pronounced: Ming-hway) Chen is a Predoctoral Staff Associate in the Decision, Risk, and Operations division at Columbia Business School, fortunately advised by Prof. Hongseok Namkoong. Exploring reinforcement learning with LLM agents.
His research interests focus on Reliable AI and Deep Learning Phenomena. He is particularly interested in developing trustworthy AI systems, exploring LLM agents, and applying reinforcement learning to real-world applications.
He received his M.Sc. in Computer Science from Southern University of Science and Technology, China, and B.Sc. in Software Engineering from Sun Yat-sen University, China.
Professional experience and positions.
Decision, Risk, and Operations division at Columbia Business School, working with Prof. Hongseok Namkoong on reinforcement learning with LLM agents.
Nanyang Technological University, working with Prof. Xiaoxiao Li and Prof. Han Yu on federated learning and multi-agent.
Vector Institute, working with Prof. Xiaoxiao Li on federated learning.
University of British Columbia, working with Prof. Xiaoxiao Li and Prof. Zehua Wang on federated learning.
ICLR 2026
Summary A local learning principle for optimizing deep compound AI systems that mitigates exploding and vanishing textual gradients in long-horizon workflows.
ICLR 2025
Summary We systematically explore the potential and challenges of incorporating textual gradient into Federated Learning, introducing FedTextGrad - a novel FL paradigm for optimizing LLMs.
NeurIPS 2024
Summary An innovative model interpolation-based local training technique that enhances local training across different clients through regularized model interpolation, acting as a catalyst for seamless adaptation of pre-trained models in federated learning.
MICCAI 2023
Summary A novel federated model soup method that optimizes the trade-off between local and global performance through selective interpolation of model parameters, alleviating overfitting and seeking flat minima for improved generalization.
AAAI 2022
Summary A novel data augmentation method that generates diverse on-manifold samples through multi-source vicinal transfer to improve model robustness against image corruptions.
NeurIPS 2021
Summary The first comprehensive study establishing five ReID benchmarks for learning corruption invariant representations, providing insights on robustness of transformer vs CNN models and cross-dataset generalization.
Conference and Journal Reviewer
Serving as a reviewer for top-tier conferences and journals in machine learning, computer vision, and medical imaging.
Period: 2022 - Present
Period: 2023 - Present