Ph.D. Candidate

Taewon Park, Ph.D. Candidate (Homepage, Google Scholar)

Research Interest
Taewon Park is a Ph.D. candidate at Kyungpook National University, advised by Prof. Hyun-Chul Kim. His research focuses on advancing the efficiency and generalization of AI models. Specifically, his works involve memory-augmented neural architectures, and modular or structured representation learning for AI, aiming to enable capabilities such as:

• Long-context reasoning

• Systematic or compositional generalization

Education

• Mar. 2019 - Present:  Integrated M.S. and Ph.D., Dept. of Artificial Intelligence, Kyungpook National University, Daegu, Republic of Korea 

• Mar. 2015 - Feb. 2019:  B.S., Electronics Engineering, Kyungpook National University, Daegu, Republic of Korea

Publications
[Conference]

• Discrete Dictionary-based Decomposition Layer for Structured Representation Learning, NeurIPS 2024

• Attention-based Iterative Decomposition for Tensor Product Representation, ICLR 2024

• Learning Associative Reasoning Towards Systematicity using Modular Networks, ICONIP 2022

• 4W1H Keyword Extraction-based Summarization Model, ICEIC 2021 

[Journal]

• Improving Systematic Generalization of Linear Transformer Using Normalization Layers and Orthogonality Loss Function, Mathematics 2024

• Distributed associative memory network with memory refreshing loss, Neural Networks 2021

Honors and Awards

• Financial Aid Award from NeurIPS, 2024

• KNU President's Excellence Award in Brain Korea Forum, 2021

Academic Service

• Reviewer at ICLR 2025