Jimyeung Seo

M.S. Student in Computer Science and Engineering, Konkuk University

Graph Neural Networks, Knowledge Distillation, Recommender Systems

Portrait of Jimyeung Seo

I am a master’s student in Computer Science and Engineering at Konkuk University, supervised by Professor Byungkook Oh.

I study how models can better find relevant information by considering relationships among data alongside user context. To this end, I explore graph-based representations of these relationships and knowledge distillation methods that transfer image-derived knowledge to text-based models. I apply these approaches to recommender systems and question answering over knowledge graphs.

Research interests

  • Graph Neural Networks
  • Knowledge Distillation
  • Recommender Systems

Education

  1. Mar 2025 to PresentEducation

    M.S. in Computer Science and Engineering (in progress)

    Konkuk University, Seoul, South Korea

    Supervisor: Byungkook Oh

  2. Mar 2019 to Feb 2025Education

    B.S. in Computer Science and Engineering

    Konkuk University, Seoul, South Korea

    GPA: 3.96 / 4.5, Major GPA: 4.0 / 4.5

Publications

2026

  1. Internalizing Negation-Gated Logical Rules into LLMs for Document-Level Relation Extraction

    Hye-Yoon Baek, Sangjun Ji, Jimyeung Seo, Hae-Yoon Koo, Xiongnan Jin, Byungkook Oh

    Findings of the Association for Computational Linguistics: EMNLP

  2. Diversifying Differentiable Graph Retrieval with Topic-Adaptive Multi-Intent Learning

    Dong-Cheon Lee, Ji-Yeon Park, Hye-Yoon Baek, Jimyeung Seo, Seyeong Kim, Byungkook Oh

    The ACM Web Conference (WWW ’26)

  3. Image-Guided Debiasing Distillation with Preference Alignment Across Multi-News Histories

    Jimyeung Seo, Eun-Yeong Jo, Hye-Yoon Baek, Dongcheon Lee, Xiongnan Jin, Byungkook Oh

    ACM International Conference on Web Search and Data Mining (WSDM ’26)

2025

  1. Interaction-Grounded Semantic Graph Refinement for LLM-based Recommendation

    Wooseok Jeong, Young-Jin Kim, Hae-Yoon Koo, Jimyeung Seo, Jinho Choi, Byungkook Oh

    IEEE Access

  2. Relation-Faceted Graph Pooling with LLM Guidance for Dynamic Span-Aware Information Extraction

    Hye-Yoon Baek, Jinho Choi, Jimyeung Seo, Xiongnan Jin, Dongcheon Lee, Byungkook Oh

    ACM International Conference on Information and Knowledge Management (CIKM ’25)

Patent applications

  1. Multimodal Data Distillation Method, and Triple-based News Refinement and Recommendation Method

    Jimyeung Seo, Byungkook Oh

    KR 10-2025-016170, Filed:

Projects

Development of Multi-Turn Question Answering Technology Based on a Manufacturing-Domain Knowledge Graph

May 2026 to Present

Supported by LG Electronics Production Engineering Research Institute.

Project Member (Graph Retrieval & Demo Development)

  • Developed Python/Neo4j graph retrieval for manufacturing text-to-SQL, assembling tables, columns, formulas, and business rules into table-centered context while retaining mandatory rules.
  • Configured vLLM serving and built a Streamlit demo, connecting graph retrieval with model inference for interactive question answering and project demonstrations.
  • Knowledge Graph
  • Multi-Turn QA

Graph-Based Threat Tracking Analysis

Oct 2025 to Present

Supported by LIG Nex1.

Project Member

  • Developed a graph-based model for early fake news detection from partially observed propagation graphs.
  • Designed predictive propagation modeling to guide information flow among observed reactions.
  • Graph Analysis

Recommendation System Model Design Consulting and Case Study

Sep 2025 to Present

Part of an Agency for Defense Development (ADD) program led by Hanwha Systems.

Project Leader (Recommender Design Consulting)

  • Led working-level recommender design consulting, analyzing command-and-control entities, relationships, and state changes over time. Proposed profile-aware retrieval within permitted information scopes and ranking based on user roles and interests.
  • Reviewed personalization case studies and designed evaluation criteria for user representation, query analysis, retrieval, and ranking, including candidate coverage, ranking quality, and sparse histories or incomplete graph information.
  • Recommender Systems

Graph-Structured Interaction for Improving Reliability and Controllability of Multimodal Generative Models

Sep 2024 to Present

Supported by the National Research Foundation of Korea, funded by the Ministry of Science and ICT.

Project Member

  • Developed image-guided knowledge distillation into text-only models for news recommendation.
  • Designed triple-level supervised fine-tuning and preference alignment based on factuality and consistency across news histories.
  • Multimodal Generative Models
  • Graph-Structured Interaction

Awards and Scholarships

  1. 2026.06.05

    2026 AI서울테크연구지원사업 Scholarship Recipient (Master’s Program)

    서울미래인재재단

  2. 2025.12.17

    Kakao X KIISE AI Agent Contest: Excellence Award

    Kakao, Korean Institute of Information Scientists and Engineers

  3. 2025.12.17

    Graduate KU우수연구장학 Scholarship Recipient (Fall 2025)

    Konkuk University

  4. 2025

    Dean’s List

    Konkuk University

  5. 2023.08.17

    1st Konkuk University Hackathon: Excellence Award

    Ministry of Science and ICT, Institute of Information & Communications Technology Planning & Evaluation, Konkuk University

Skills & languages

Programming & tools
C/C++, Python, PyTorch, Git, Docker
Languages
Korean (Native), English (OPIc IH)