M.S. in Kim Jaechul Graduate School of AI
KAIST · MLILAB · Advisor: Prof. Eunho Yang
Mar 2026 — Expected 2028
- Research on visual reasoning in multimodal large language models.
Mar. 2026
🎓 Started M.S. at KAIST AI and joined MLILAB.
About Me
Yoonsik Park
M.S. Student · KAIST AI · MLILAB
I am an M.S. student in the Kim Jaechul Graduate School of AI at KAIST, working in MLILAB advised by Prof. Eunho Yang.
My current research focuses on visual reasoning in multimodal large language models, with interests in faithfulness, hallucination mitigation, and robust multimodal learning.
Previously, I received a B.S. in Data Science from Hanyang University, graduating early with Summa Cum Laude (Ranked 1st in Dept. of Data Science).
News
Jan. 2026
📄 BiasEdit accepted to WWW 2026 (Oral Presentation). My first Paper!
Dec. 2025
🔬 Joined NAVER Place AI as a research intern.
Aug. 2025
🏅 Graduated early in 7 semesters from Hanyang University with Summa Cum Laude (Ranked 1st in the Department of Data Science).
Thanks, Hanyang! 🦁
Thanks, Hanyang! 🦁
Jun. 2025
🧪 Joined KAIST MLILAB as an undergraduate research intern.
Feb. 2024
🌱 Joined BAIKLAB at Hanyang University as an undergraduate research intern.
Education
B.S. in Data Science
Hanyang University (Seoul)
Mar 2022 — Aug 2025
- Cumulative GPA: 4.35/4.5 · Major GPA: 4.38/4.5.
- Early graduation with Summa Cum Laude (Ranked 1st in Dept. of Data Science).
Experiences
AI Research Intern
NAVER Place AI
Dec 2025 — Jan 2026
- Developed passage-level models for review relevance and informativeness assessment using an LLM-driven pipeline with 150K+ labeled samples.
- Built an efficient model on a lightweight 0.8B-parameter embedding backbone, achieving 0.90+ F1 for relevance and 0.94+ F1 for informativeness.
Undergraduate Research Intern
KAIST · MLILAB · Advisor: Prof. Eunho Yang
Jun 2025 — Aug 2025
- Investigated methods to mitigate hallucinations in multimodal large language models.
- Focused on improving visual faithfulness so generated text remains grounded in image evidence.
Undergraduate Research Intern
Hanyang University · BAIKLAB · Advisor: Prof. Sungyong Baik
Feb 2024 — Jun 2025
- Conducted research on debiasing vision models trained on web-sourced image data.
- Extended the study to vision-language models such as CLIP, focusing on bias-robust soft prompts and worst-group robustness.
Publication
Harnessing Bias for Debiasing: Annotation-Free Prompt Learning for Robustness in CLIP
BiasEdit: A Training-Free Bias-Detect-and-Edit Framework for Learning Fair Visual Classifiers
Projects
HBD (Harnessing Bias for Debiasing)
Developed an annotation-free CLIP prompt-learning framework that discovers pseudo-groups and combines bias-conflict and performance signals to improve worst-group robustness.
Notify: LLM-based Personal Schedule Assistant
Built a mobile app that predicts personalized departure times by combining schedules, traffic signals, and persona-aware LLM reasoning.
The Motivational Organizer & Mentor (M.O.M.)
Built a persona-driven conversational agent based on Gemma 2 for empathetic, nudge-based productivity support.
F.D.A.: Fine Dining Assistant
Built a RAG-based fine-dining chatbot that simulates professional service personas to deliver personalized, experiential fine-dining previews.
Cooling with Nature
Developed a hybrid ARIMAX-Transformer time-series model to forecast data center cooling consumption and optimize locations for energy efficiency.