Portrait of Kyungmin Min 민경민

Kyungmin Min 민경민

Ph.D. Student in Artificial Intelligence, Seoul National University

About

I am a Ph.D student at Seoul National University, advised by Prof. Kyomin Jung. My research lies at the intersection of multimodal LLMs and experience-driven memory for AI agents. I study how multimodal models can stay grounded in what they see, and how agents can learn from their own experience: what to store, how to retrieve and reuse it, and how memory helps them solve tasks more accurately and in fewer steps. Bringing these two threads together, I am extending agent memory beyond text to images and videos.

Agentic AI Multi-Agent System VLM/LLM

Education

  • Seoul National University — Ph.D. Student in Artificial Intelligence (Mar 2023 – Present)
    Advisor: Prof. Kyomin Jung
  • Sungkyunkwan University — B.S. in Computer Science and Engineering, Magna Cum Laude (Mar 2017 – Feb 2023)

Honors & Awards

  • Best Paper Award, The Eleventh Dialog System Technology Challenge (DSTC 11) — Track 5 (Task-oriented Conversational Modeling with Subjective Knowledge) Sep 2023

Work Experience

LG AI Research logo LG AI Research, EXAONE Lab
Research Intern
Mentor: Soyeon Kim
May 2026 – Aug 2026
Adobe Research logo Adobe Research
Collaborative Researcher
Host: David Seunghyun Yoon
May 2025 – May 2026

News

Jun 18, 2026 Our ReflectCAP: Detailed Image Captioning with Reflective Memory paper is accepted to ECCV 2026!
Apr 01, 2026 Our Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning paper is accepted to ACL 2026 Findings!
Aug 21, 2025 Our Fooling the LVLM Judges: Visual Biases in LVLM-Based Evaluation paper is accepted to EMNLP 2025!
Jan 23, 2025 Our Mitigating Hallucinations in Large Vision-Language Models via Summary-Guided Decoding paper is accepted to NAACL 2025 Findings, See you in New Mexico!🏜️
Dec 15, 2024 Our Return of EM: Entity-driven Answer Set Expansion for QA Evaluation paper is accepted to COLING 2025 as Oral presentation!

Publications

* indicates equal contribution among authors
† indicates co-corresponding authors

  1. ReflectCAP: Detailed Image Captioning with Reflective Memory
    Kyungmin Min ,  Minbeom Kim ,  Kang-il Lee ,  Seunghyun Yoon , and  Kyomin Jung
    ECCV 2026 , Apr 2026 [link]
  2. Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning
    Junseok Kim ,  Nakyeong Yang ,  Kyungmin Min , and  Kyomin Jung
    ACL 2026 (Findings) , Apr 2026 [link]
  3. Fooling the LVLM Judges: Visual Biases in LVLM-Based Evaluation
    Yerin Hwang* ,  Dongryeol Lee* ,  Kyungmin Min ,  Taegwan Kang ,  Yong-il Kim , and  Kyomin Jung
    EMNLP 2025 , Nov 2025 [link]
  4. Mitigating Hallucinations in Large Vision-Language Models via Summary-Guided Decoding
    Kyungmin Min ,  Minbeom Kim ,  Kang-il Lee ,  Dongryeol Lee , and  Kyomin Jung
    NAACL 2025 (Findings) , Apr 2025 [link]
  5. Return of EM: Entity-driven Answer Set Expansion for QA Evaluation
    Dongryeol Lee ,  Minwoo Lee ,  Kyungmin Min ,  Joonsuk Park† , and  Kyomin Jung†
    COLING 2025 (Oral) , Jan 2025 [link]
  6. Mitigating Hallucinations in LVLMs via Summary-Guided Decoding
    Kyungmin Min ,  Minbeom Kim ,  Kang-il Lee ,  Dongryeol Lee , and  Kyomin Jung
    Neurips Safe Generative AI Workshop 2024 , Dec 2024 [link]
  7. Leveraging Ensemble Techniques and Metadata for Subjective Knowledge-grounded Conversational Systems
    Seongho Joo* ,  Kang-il Lee* ,  Kyungmin Min* ,  Joongbo Shin ,  Janghoon Han ,  Seungpil Won , and  Kyomin Jung
    The Eleventh Dialog System Technology Challenge (DSTC11) , Sep 2023 [link]
    Our paper was awarded as the Best Paper (news) and also ranked first (news) in the response accuracy category at the DSTC11 challenge.