Research Interests

• Multimodal Large Language Model

• Vision Language Model

• Spatio Temporal Visual Grounding

• Efficiency in Large Language Model Inference

Education

• M.S. Student (Mar 2025 - Present)

Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST)
Daejeon, Korea
Advisor : Prof. Yong Man Ro

• B.S. (Mar 2019 - Feb 2025)

Electrical Engineering, Hanyang University
Seoul, Korea

 Publications

• TARS: Tailness-aware Advantage ReScaling for Domain-specific Visual Recognition with LVLMs

Beomchan Park*, Seongho Kim*, and Yong Man Ro

Under Review

• Language-Guided Semantic Cues: Robust Visual Grounding with MLLMs in Crowded Scenes

Beomchan Park*, Seongho Kim*, Hyunjun Kim, Suneung Kim, and Yong Man Ro

Under Review

• GROUNDPRUNE: Efficient Visual Token Pruning without Losing Fine-Grained Visual Understanding

Hyunjun Kim, Sungjune Park, Hosu Lee, Beomchan Park, Seongho Kim, and Yong Man Ro

Under Review

• SelecT: Selection-based Spatio-Temporal Video Grounding with MLLMs

Hyunjun Kim*, Beomchan Park*, Seongho Kim*, and Yong Man Ro

Under Review

• DIP-R1: Deep Inspection and Perception with RL Looking Through and Understanding Complex Scenes 

Sungjune Park*, Hyunjun Kim*, Junho Kim, Seongho Kim, and Yong Man Ro

Under Review [paper]

• Robust Grounding with MLLMs against Occlusion and Small Objects via Language-Guided Semantic Cues

Beomchan Park*, Seongho Kim*, Hyunjun Kim, Sungjune Park, and Yong Man Ro

IEEE International Conference on Acoustics, Speech, and Signal Processing  (ICASSP), 2026

• GTA-Crime: A Synthetic Dataset and Generation Framework for Fatal Violence Detection with Adversarial Snippet-Level Domain Adaptation

Seongho Kim*, Sejong Ryu*, Hyoukjun You*, Je Hyeong Hong

IEEE International Conference on Image Processing (ICIP), 2025 [paper]