I'm a Ph.D. student and Research Assistant at University of Wisconsin-Madison Computer Sciences department. I am fortunate to be advised by prof. Sharon Yixuan Li, and I have also worked with prof. Hyunwoo J. Kim as a Master's student in the Computer Science and Engineering department at Korea University.
My research goal is to make AI Agents reliable in real-world applications. Reliability in agentic AI systems can be examined from multiple perspectives, e.g., architectural design, quality of reasoning, autonomous decision-making, and inter-agent communication. At present, my primary focus is on the reliability of interactions within multi-agent systems, where coordination, debate, and collective reasoning play a central role. Beyond this, I have explored a broad range of domains—spanning computer vision, natural language processing, knowledge graph reasoning, reinforcement learning, efficient deep learning, and financial machine learning—aiming to bring diverse perspectives to the study of dependable AI agents.
Current Research Interests
Reliable and Safe AI Agents
Multi-Agent Systems
Contact : hyeongkyu.choi [at] wisc [dot] edu
[05.01.2024] Our paper on Persona In-Context Learning (PICLe) has been accepted for presentation at ICML 2024!
[04.14.2024] Will be working as an Applied Scientist intern at Amazon during the summer!
[09.21.2023] Our paper on Neural Tree Search (NuTrea) has been accepted for presentation at NeurIPS 2023!
[07.05.2023] Our paper on Recurrent DETR has been accepted at IEEE Access!
[04.02.2023] Will be joining professor Sharon Li's group at University of Wisconsin-Madison this fall in pursuit of a MS/PhD degree in Computer Science
[02.28.2023] Our paper on Meta Loss Transformer (MELTR) has been accepted for presentation at CVPR 2023!
[02.24.2023] Will be giving a short talk/tutorial at KCVS 2023 Workshop on TokenMixup. [tutorial code]
[11.19.2022] Our paper on Question Answering Transformer (QAT) for knowledge graphs has been accepted for presentation at AAAI 2023!
[09.15.2022] Our paper on TokenMixup has been accepted for presentation at NeurIPS 2022!
[03.02.2022] Our paper on Cross-path Consistency has been accepted for poster presentation at CVPR 2022!
Selected Publications
Mitigating Selection Bias with Node Pruning and Auxiliary Options
Hyeong Kyu Choi, Weijie Xu, Chi Xue, Stephanie Eckman, Chandan K. Reddy
Annual Meeting of the Association for Computational Linguistics (ACL Main), 2025.
How Contaminated Is Your Benchmark? Measuring Dataset Leakage in Large Language Models with Kernel Divergence
Hyeong Kyu Choi*, Maxim Khanov*, Hongxin Wei, Yixuan Li
International Conference on Machine Learning (ICML), 2025
PICLe: Eliciting Diverse Behaviors from Large Language Models with Persona In-Context Learning
Hyeong Kyu Choi, Yixuan Li
International Conference on Machine Learning (ICML), 2024
NuTrea: Neural Tree Search for Context-guided Multi-hop KGQA
Hyeong Kyu Choi, Seunghun Lee, Jaewon Chu, Hyunwoo J. Kim
Advances in Neural Information Processing Systems (NeurIPS), 2023
TokenMixup: Efficient Attention-guided Token-level Data Augmentation for Transformers,
Hyeong Kyu Choi*, Joonmyung Choi*, Hyunwoo J. Kim
Advances in Neural Information Processing Systems (NeurIPS), 2022