Hoyeon Chang [hojʌn tɕɑŋ]

prof_pic.jpg

E-mail: retapurayo (at) kaist.ac.kr

I am a Ph.D. student at KAIST Graduate School of AI, advised by Minjoon Seo and Seongjoon Oh. My research investigates the structural foundations of intelligence: specifically, how systematic compositionality and robust belief revision can be realized in connectionist networks. Currently, I am exploring the interface of category theory and the geometry of neural representations, aiming to uncover the structural constraints that govern generalization. I deeply value interdisciplinary insights and enjoy connecting ideas from philosophy, linguistics, cognitive science, and mathematics to build a comprehensive picture of intelligence.

Previously, I obtained a B.S. in Electrical & Computer Engineering and Biological Sciences from Seoul National University. Outside of research, I am a jazz pianist and occasionally play at jazz clubs in Seoul.

Education

Mar 2024 - Present Ph.D. in Artificial Intelligence
Korea Advanced Institute of Science and Technology (KAIST)
Advisor: Minjoon Seo and Seongjoon Oh
Sep 2022 - Feb 2024 M.S. in Artificial Intelligence
Korea Advanced Institute of Science and Technology (KAIST)
Advisor: Minjoon Seo
Thesis: On the Knowledge Acquisition in Language Model Pre-training
Mar 2018 - Aug 2022 B.S. in Biological Sciences and Electrical & Computer Engineering
Seoul National University
Graduated with Highest Honors (Summa Cum Laude)

Academic Experience

Jul 2024 - Present Founder & Lead
TMI (Thinking, Meaning, and Intelligence) Group
Founded and leading a monthly interdisciplinary academic group inspired by Norbert Wiener’s project on cybernetics, bringing together graduate students from philosophy, neuroscience, linguistics, computer science, physics, and education. The group explores fundamental questions about thinking, meaning, and intelligence through collaborative discussions and projects.
Feb 2022 - Aug 2022 Research Intern
KAIST LK Lab
Advisor: Minjoon Seo
Sep 2021 - Jan 2022 Research Intern
SNU LDI Lab
Advisor: Seung-won Hwang
Dec 2020 - Feb 2021 Research Intern
SNU CHANGLab
Advisor: Hyeshik Chang
May 2020 - Dec 2020 Undergraduate Researcher
SNU Undergraduate Research Program

Publications

  1. Dynamics Reveals Structure: Challenging the Linear Propagation Assumption
    Hoyeon Chang, Bálint Mucsányi, and Seong Joon Oh
    ICML 2026 (Spotlight)
    ICLR 2026 Workshop on Unifying Concept Representation Learning (Oral)
    arXiv Talk
  2. Characterizing Pattern Matching and Its Limits on Compositional Task Structures
    Hoyeon Chang*, Jinho Park*, Hanseul Cho*, Sohee Yang, Miyoung Ko, and 5 more authors
    ICLR 2026
    arXiv
  3. Latent Reasoning via Sentence Embedding Prediction
    Hyeonbin Hwang*, Byeongguk Jeon*, Seungone Kim, Jiyeon Kim, Hoyeon Chang, and 5 more authors
    COLM 2025 RAM 2 Workshop (Oral)
    arXiv
  4. Dynamic VLM-Guided Negative Prompting for Diffusion Models
    Hoyeon Chang*, Seungjin Kim*, and Yoonseok Choi*
    NeurIPS 2025 Workshop on Generative and Protective AI for Content Creation
    arXiv
  5. How Does Vision-Language Adaptation Impact the Safety of Vision Language Models?
    Seongyun Lee*, Geewook Kim*, Jiyeon Kim*, Hyunji Lee, Hoyeon Chang, and 2 more authors
    ICLR 2025
    arXiv
  6. How Do Large Language Models Acquire Factual Knowledge During Pretraining?
    Hoyeon Chang, Jinho Park, Seonghyeon Ye, Sohee Yang, Youngkyung Seo, and 2 more authors
    NeurIPS 2024
    arXiv
  7. Nonparametric Decoding for Generative Retrieval
    Hyunji Lee, JaeYoung Kim, Hoyeon Chang, Hanseok Oh, Sohee Yang, and 3 more authors
    ACL 2023 Findings
    arXiv

Languages

Korean Native
English Full academic proficiency (TOEIC 925/990)
Japanese Full academic proficiency (JLPT N1 164/180)
German Elementary proficiency
Spanish Elementary proficiency