About Me
I am a Ph.D. candidate in Artificial Intelligence at KAIST, where I am advised by Prof. Eunho Yang in the Machine Learning and Intelligence Lab. I expect to graduate in February 2027.
My research focuses on RL-based post-training for robust reasoning in language and multimodal models. I study self-verification, instruction-following robustness, and multimodal reasoning with discrete diffusion. Previously, I worked on text-guided image editing and generative-model manipulation, and completed research internships at Sony Research and NAVER Cloud.
Publications
Reasoning Model is Stubborn: Diagnosing Instruction Overriding in Reasoning Models
Doohyuk Jang*, Yoonjeon Kim*, Chanjae Park, Hyun Ryu, Eunho Yang
Verifying Meta-Awareness via Predictive Rewards in Reasoning Models
Yoonjeon Kim*, Doohyuk Jang*, Eunho Yang
Preserve or Modify? Context-Aware Evaluation for Balancing Preservation and Modification in Text-Guided Image Editing
Yoonjeon Kim*, Soohyun Ryu*, Yeonsung Jung, Hyunkoo Lee, Joowon Kim, June Yong Yang, Jaeryong Hwang, Eunho Yang
Learning Input-Agnostic Manipulation Directions in StyleGAN with Text Guidance
Yoonjeon Kim, Hyunsu Kim, Junho Kim, Yunjey Choi, Eunho Yang
Sequential Targeting: A Continual Learning Approach for Data Imbalance in Text Classification
Joel Jang, Yoonjeon Kim, Kyoungho Choi, Sungho Suh
Preprints
Efficient Reinforcement for Visual-Textual Thinking with Discrete Diffusion Model
Yoonjeon Kim, Yuhta Takida, Chieh-Hsin Lai, Eunho Yang, Yuki Mitsufuji
Work Experience
Sony Research
Research Intern · Tokyo, Japan
- Developed RL post-training methods for multimodal discrete diffusion models, targeting visual-textual reasoning and visual question answering.
- Worked with Yuhta Takida and Chieh-Hsin (Jesse) Lai on multimodal reasoning and reinforcement learning.
NAVER Cloud
Machine Learning Research Intern · Bundang, South Korea
- Researched dynamic Neural Radiance Fields for monocular-video 3D reconstruction.
- Investigated diffusion-based NeRF model distillation for efficient 3D scene representation and generation.
Selected Projects
Invited Talks
AI EXPO KOREA · KAIST AI Tech Seminar
Presenter · COEX, Seoul
Education
Service
Reviewer for ICML, NeurIPS, ICLR, and CVPR. For a complete record, download my CV.