About Me

I am an undergraduate studying Computer Engineering and Information Statistics at Duksung Women’s University, with expected graduation in February 2027. Most recently, I was a research intern at ETRI, working on knowledge graphs and multi-hop GraphRAG. Previously, I conducted research on LLM reasoning and document understanding at DMKD Lab under the supervision of Prof. Jehyeok Rew, and on low-light human pose estimation in the Vision and Multimodal Intelligence Lab at the University of Mississippi under the supervision of Prof. Hawk Wang. I also worked at BCG, analyzing data for a global battery manufacturer. Here is my CV.

My research spans document understanding, LLM reasoning, and retrieval-augmented generation. Building on my work in document structure analysis (CoDU) and efficient prompting (SPTC), my recent work on multi-hop GraphRAG over knowledge graphs at ETRI focuses on retrieving and connecting evidence for multi-hop question answering.

News

Scroll for earlier updates.

Work in Progress

Multi-Hop GraphRAG over Knowledge Graphs

Developed path-preserving beam search and evidence reranking for multi-hop question answering during my research internship at ETRI.

Publications

UDAPose Figure 1: comparison of enhancement-based, image translation-based, and proposed low-light pose estimation approaches

UDAPose: Unsupervised Domain Adaptation for Low-Light Human Pose Estimation [paper][code]

Haopeng Chen, Yihao Ai, Kabeen Kim, Robby T. Tan, Yixin Chen, Bo Wang

CVPR 2026

ConFuse Figure 1: overview of context-aware LLM and GBDT prediction fusion

ConFuse: Context-Aware Fusion of LLMs and GBDTs for Sleep Prediction from Lifelog Data [code]

Kabeen Kim*, Yena Kim*, Minjeong Seo*

* Equal contribution.

ICTC 2026

CoDU Figure 2: bounding box detection and refinement, OCR extraction, and reading order rearrangement pipeline

CoDU: Multi-Stage Approach for Complex Document Structure Understanding

Kabeen Kim, Minhye Lee, Haein Seo, Jehyeok Rew

KDBC 2025

SPTC Figure 1: comparison of Tree-of-Thought on the left and SPTC on the right

SPTC (Single-Pass Tree Chain-of-Thought): An Efficient Prompting Scheme for Small Language Models [paper]

Kabeen Kim, Jiye Park, Jehyeok Rew

KCC 2025

Education

Work Experience

Electronics and Telecommunications Research Institute

Research Intern · Knowledge Graph & Multi-Hop GraphRAG | Daejeon, South Korea

Jul 2026 – Aug 2026

Boston Consulting Group

Research Analyst Intern · Data Analysis for a Global Battery Manufacturer | Seoul, South Korea

Mar 2026 – May 2026

DMKD Lab · Duksung Women’s University

Undergraduate Researcher · LLM Reasoning & Document Understanding | Seoul, South Korea

Advisor: Prof. Jehyeok Rew

Mar 2025 – Feb 2026

Vision and Multimodal Intelligence Lab · University of Mississippi

Undergraduate Researcher · Low-Light Human Pose Estimation | Mississippi, United States

Advisor: Prof. Hawk Wang

Aug 2024 – Feb 2025

Awards & Scholarships

  • 2025.10: 3rd Place, Samsung AI Challenge — Visually-Rich Document Understanding track (210 teams).
  • 2024: Mirae Asset Overseas Exchange Scholarship, Mirae Asset Park Hyeon Joo Foundation.
  • 2024: Excellence Award, TAVE 13th Data Analysis Project.
  • 2024: 2nd Prize, DS Global Vision Makers Innovation Idea Competition.
  • 2021–2022: Academic Excellence Scholarship, Duksung Women’s University (3 awards).

Outreach

  • Volunteer Coding Instructor, CJ UNIT (8th cohort) · Mar–Jun 2023. Developed software curricula and delivered remote coding instruction, including LEGO-based programming activities, to local middle-school students.