Structure-based Reasoning
Parsing · Hierarchical & Knowledge Graphs
HiKEY ACL 2026PILAR EMNLP 2026M3DocDep CVPR 2026M.S. Thesis PNU 2023Neural-Symbolic Parsing HCLT 2022
Joongmin Shin
Senior Researcher, Human-Inspired AI Research, Korea University · Founder & Lead, KUDoc
Working with Prof. Heuiseok Lim and Prof. Jaehyung Seo
B.S. & M.S., Pusan National University (Advisor: Prof. Hyuk-Chul Kwon)
My research uses structure-based reasoning to address real-world and industrial problems, with a focus on making AI genuinely useful to people. My master’s research in parsing gave me this structural perspective, while my later work in human-inspired AI clarified what that perspective should ultimately serve. I have since developed a coherent research program spanning multimodal and agentic AI.
Looking ahead, I aim to study how AI systems should represent, select, update, and verify information so that they remain useful and reliable as they reason over complex multimodal inputs and engage in extended interactions. Over the longer term, I am interested in extending these principles to world models and embodied agents in robotics.
- 14 Publications Conferences, journals, and theses
- 6 First-author papers NeurIPS · ACL (Oral) · CVPR · EMNLP, main track at all four
- 5 Patents Korea · 2 registered, 3 pending
- 7 Awards Incl. 3 paper or research awards
Research Areas
My research uses structure-based reasoning to address real-world and industrial problems. Parsing, hierarchies, and knowledge graphs give multimodal and agentic systems the structure they need to represent, select, update, and verify information.
Multimodality
Documents · Video · World Models
MultiDocFusion EMNLP 2025M3DocDep CVPR 2026HiKEY ACL 2026PILAR EMNLP 2026StyleDFS EMNLP 2024Long-Video QA/RAG Ongoing
Agentic AI
Memory · Verification
News
- Sep 2026
ADMIT was accepted at NeurIPS 2026 (Main, first author).
- Aug 2026
PILAR was accepted at EMNLP 2026 (Findings, first author).
- Apr 2026
HiKEY was accepted at ACL 2026 (Main, Oral, first author).
- Feb 2026
M3DocDep was accepted at CVPR 2026 (Main, first author).
- Aug 2025
MultiDocFusion was accepted at EMNLP 2025 (Main, first author).
- Aug 2025
Appointed to the National Representative K-AI Research Team (with NC AI).
- Oct 2024
StyleDFS was accepted at EMNLP 2024 (Industry, co-first author).
Selected Publications
All publications →* denotes first or co-first author.
ADMIT: Support-Gated Memory-Write Admission for Document QA Agents
Support-gated admission control that decides what a document QA agent is allowed to commit to memory.
PILAR: A Page-Grounded Unified Evidence Representation via an Entity-Linked Assertion Graph for Open-Domain QA Agents over Multimodal Document Corpora
Entity-linked assertion graphs unify text, tables, and figures into a single page-grounded evidence representation for open-domain multimodal QA agents.
HiKEY: Hierarchical Multimodal Retrieval for Open-Domain Document Question Answering
Hierarchical retrieval for multimodal document QA with structured evidence assembly.
M3DocDep: Multi-modal, Multi-page, Multi-document Dependency Chunking with Large Vision-Language Models
LVLM-based dependency chunking that reconstructs cross-page structure for long-document retrieval and QA.
MultiDocFusion: Hierarchical and Multimodal Chunking Pipeline for Enhanced RAG on Long Industrial Documents
A hierarchical multimodal chunking pipeline that preserves layout and improves evidence composition in industrial RAG.
Intelligent Predictive Maintenance RAG Framework for Power Plants: Enhancing QA with StyleDFS and Domain Specific Instruction Tuning
Domain-specific RAG framework for scientific and industrial QA; led to two technology transfers.
Projects
All projects →Multimodal Reasoning & Document AI 5
Video Curation SystemDocGraph CopilotScience-Domain Korean LLM & RAG PipelineSynerpeace PilotSynapse
Foundation Models and Adaptation 6
National Representative K-AI Research TeamKo-GemmaKULLMRobot Process Automation – PLC Control-Code GenerationMi:deum K 1.0Exobrain WiseQA (ETRI)
Additional Projects 5
TTS Prosodic Boundary Prediction (KT)TTS Pronunciation Sequence (KT)Navigation TTS System (Navice)AI Online Judge for Coding EducationAI Mineral Location Prediction (KIGAM)