Projects / Multimodal Reasoning & Document AI

DocGraph Copilot

Jan 2025 - Present · Korea University

EMNLP 2025CVPR 2026

Description

Developed a novel multimodal chunking methodology for real-world industrial data. It is designed to chunk documents by converting their layout information into a hierarchical structure.

Role

Research, Design, Development

Achievements

Accepted to EMNLP 2025 (Main) and CVPR 2026.