About Me

Hi, I am Jihwan Park, a Postdoctoral Researcher at KAIST supervised by Prof. Hyunwoo J. Kim. I received my Ph.D. and B.S. in Computer Science from Korea University.

My primary research centers on dense visual prediction—object detection and human-object interaction detection—with an emphasis on efficient and transferable learning for practical deployment, spanning weakly-supervised, zero-shot, and model-agnostic settings. Recently, I have been interested in model merging as a way to combine specialized models into a single generalist one without costly retraining, and in vision-language-action (VLA) models, where perception, language understanding, and action generation are integrated to build generalizable embodied AI systems.

Currently, my research interests include:

• Dense prediction for real-world vision systems
• Efficient and transferable adaptation on foundation model
• Vision-language-action models and embodied AI

I am currently exploring next-position opportunities in academia and industry. Please feel free to reach out at jseven7071[AT]kaist.ac.kr.

Education

• Ph.D. in Computer Science, Korea University
  March 2021 - February 2026
  Supervised by Prof. Hyunwoo J. Kim

• B.S. in Computer Science, Korea University
  March 2017 - February 2021

Work Experiences

• Postdoctoral Researcher, KAIST
  Daejeon, South Korea
  March 2026 - Present
  Supervised by Hyunwoo J. Kim

• Research Assistant, Korea University
  Seoul, South Korea
  September 2022 - June 2023

• Research Intern, Kakao Brain
  Pangyo, South Korea
  December 2021 - June 2022
  Supervised by Byungseok Roh

News

2026

• May New preprint: Mitigating State Aliasing in Vision-Language-Action Models via Inverse Dynamics Learning is now available on arXiv.
• March RegFormer has been accepted to CVPR 2026.
• January TransMiter will be presented as an oral at AAAI 2026.

Selected Publications

• [CVPR 2026] RegFormer: Transferable Relational Grounding for Efficient Weakly-Supervised Human-Object Interaction Detection Code
  Jihwan Park, Chanhyeong Yang, Jinyoung Park, Taehoon Song, Hyunwoo J. Kim

• [AAAI 2026 Oral] Transferable Model-agnostic Vision-Language Model Adaptation for Efficient Weak-to-Strong Generalization Code
  Jihwan Park, Taehoon Song, Sanghyeok Lee, Miso Choi, Hyunwoo J. Kim

• [NeurIPS 2025] Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection Code
  Chanhyeong Yang, Taehoon Song, Jihwan Park, Hyunwoo J. Kim

• [CVPR 2024] Groupwise Query Specialization and Quality-aware Multi-assignment for Transformer-based Visual Relationship Detection Code
  Jongha Kim*, Jihwan Park*, Jinyoung Park*, Jinyoung Kim, Sehyung Kim, Hyunwoo J. Kim

• [CVPR 2022] Consistency Learning via Decoding Path Augmentation for Transformers in Human Object Interaction Detection Code
  Jihwan Park, SeungJun Lee, Hwan Heo, Hyeong Kyu Choi, Hyunwoo J. Kim

See the Publications page for the full list.