Minseok Choi

About

Hi! I am a postdoctoral fellow at KAIST AI, supported by the InnoCORE Postdoctoral Fellowship, where I work with Jaegul Choo at the DAVIAN Lab. I work on the safety and alignment of large language models, with a growing focus on systems that improve themselves. My goal is to keep these systems trustworthy as they become more capable and autonomous. Previously, I was a research intern at Krafton and Naver Webtoon. I received my Ph.D. in Artificial Intelligence from KAIST and my B.S. in Computer Science from the University of Illinois at Urbana-Champaign.

I hold my current position as Technical Research Personnel, an alternative to mandatory military service in the Republic of Korea Army. I will be discharged on February 28, 2027, and open to new opportunities starting March 2027. If your team is working on similar problems, please feel free to reach out over email. I also welcome research collaborations, or just a coffee chat in Seoul!

News

• [Aug 2026]   Our paper Membrane has been accepted to EMNLP 2026!
• [Jun 2026]   PUBG Ally, a project I worked on during my internship at Krafton, is now live in beta on PUBG!
• [Feb 2026]   Terminus-KIRA, a coding agent developed by our team at Krafton, reached 74.8% on Terminal-Bench, ranking #2 on the leaderboard (#1 with Opus 4.6).
• [Jan 2026]   Our paper ExpGuard has been accepted to ICLR 2026!
• [Nov 2025]   I have successfully defended my dissertation!
• [May 2025]   Our paper Opt-Out has been accepted to ACL 2025!
• [May 2025]   I will be joining Krafton as an Applied Research Intern this summer!
• [Sep 2024]   Our paper LingTea has been accepted to EMNLP 2024 Findings!
• [Jul 2024]   Our paper PairEval has been accepted to COLM 2024!
• [May 2024]   Our paper POP has been accepted to ACL 2024 Findings!
• [Oct 2023]   Our paper SimCKP has been accepted to EMNLP 2023 Findings!
• [May 2023]   Our paper HistRED has been accepted to ACL 2023!
• [Mar 2023]   I started my military service in the Republic of Korea Army as Technical Research Personnel.
• [Jan 2022]   I will be joining Naver Webtoon as a Research Intern this winter!

Publications

( * indicates equal contributions.)

Membrane: A Self-Evolving Contrastive Safety Memory for LLM Agent Defense

Minseok Choi*, Seungbin Yang*, Dongjin Kim*, Subin Kim, Jungmin Son, Yunseung Lee, Jaegul Choo, Youngjun Kwak

EMNLP 202615.4% acceptance rate

Paper Code

ExpGuard: LLM Content Moderation in Specialized Domains

Minseok Choi*, Dongjin Kim*, Seungbin Yang*, Subin Kim, Youngjun Kwak, Juyoung Oh, Jaegul Choo, Jungmin Son

ICLR 202627.4% acceptance rate

Paper Code Data

Opt-Out: Investigating Entity-Level Unlearning for Large Language Models via Optimal Transport

Minseok Choi, Daniel Rim, Dohyun Lee, Jaegul Choo

ACL 202520.3% acceptance rate

Paper Code Data

Breaking Chains: Unraveling the Links in Multi-Hop Knowledge Unlearning

Minseok Choi, ChaeHun Park, Dohyun Lee, Jaegul Choo

arXiv Preprint 2024

Paper

Cross-Lingual Unlearning of Selective Knowledge in Multilingual Language Models

Minseok Choi, Kyunghyun Min, Jaegul Choo

EMNLP 2024 Findings

Paper Code

PairEval: Open-domain Dialogue Evaluation with Pairwise Comparison

ChaeHun Park, Minseok Choi, Dohyun Lee, Jaegul Choo

COLM 202428.9% acceptance rate

Paper Code

Protecting Privacy Through Approximating Optimal Parameters for Sequence Unlearning in Language Models

Dohyun Lee*, Daniel Rim*, Minseok Choi, Jaegul Choo

ACL 2024 Findings

Paper

SimCKP: Simple Contrastive Learning of Keyphrase Representations

Minseok Choi, Chaeheon Gwak, Seho Kim, Si hyeong Kim, Jaegul Choo

EMNLP 2023 Findings

Paper Code

PRiSM: Enhancing Low-Resource Document-Level Relation Extraction with Relation-Aware Score Calibration

Minseok Choi, Hyesu Lim, Jaegul Choo

IJCNLP-AACL 2023 Findings

Paper Code

HistRED: A Historical Document-Level Relation Extraction Dataset

Soyoung Yang, Minseok Choi, Youngwoo Cho, Jaegul Choo

ACL 202322.1% acceptance rate

Paper Code Data

Rethinking Style Transformer by Energy-based Interpretation: Adversarial Unsupervised Style Transfer using a Pretrained Model

Hojun Cho, Dohee Kim, Seungwoo Ryu, ChaeHun Park, Hyungjong Noh, Jeong-in Hwang, Minseok Choi, Edward Choi, Jaegul Choo

EMNLP 202219.8% acceptance rate

Paper Code

Development and Application of Web‑based Machine Learning Program for Automated Assessment Model Generation

Minseok Choi, Jaegul Choo, Minsu Ha

Brain, Digital, & Learning, Vol. 12, No. 4, pp. 567-578, 2022

Paper

Membrane: A Self-Evolving Contrastive Safety Memory for LLM Agent Defense

Minseok Choi*, Seungbin Yang*, Dongjin Kim*, Subin Kim, Jungmin Son, Yunseung Lee, Jaegul Choo, Youngjun Kwak

EMNLP 202615.4% acceptance rate

Paper Code

ExpGuard: LLM Content Moderation in Specialized Domains

Minseok Choi*, Dongjin Kim*, Seungbin Yang*, Subin Kim, Youngjun Kwak, Juyoung Oh, Jaegul Choo, Jungmin Son

ICLR 202627.4% acceptance rate

Paper Code Data

Opt-Out: Investigating Entity-Level Unlearning for Large Language Models via Optimal Transport

Minseok Choi, Daniel Rim, Dohyun Lee, Jaegul Choo

ACL 202520.3% acceptance rate

Paper Code Data

Breaking Chains: Unraveling the Links in Multi-Hop Knowledge Unlearning

Minseok Choi, ChaeHun Park, Dohyun Lee, Jaegul Choo

arXiv Preprint 2024

Paper

Cross-Lingual Unlearning of Selective Knowledge in Multilingual Language Models

Minseok Choi, Kyunghyun Min, Jaegul Choo

EMNLP 2024 Findings

Paper Code

PairEval: Open-domain Dialogue Evaluation with Pairwise Comparison

ChaeHun Park, Minseok Choi, Dohyun Lee, Jaegul Choo

COLM 202428.9% acceptance rate

Paper Code

Protecting Privacy Through Approximating Optimal Parameters for Sequence Unlearning in Language Models

Dohyun Lee*, Daniel Rim*, Minseok Choi, Jaegul Choo

ACL 2024 Findings

Paper

SimCKP: Simple Contrastive Learning of Keyphrase Representations

Minseok Choi, Chaeheon Gwak, Seho Kim, Si hyeong Kim, Jaegul Choo

EMNLP 2023 Findings

Paper Code

PRiSM: Enhancing Low-Resource Document-Level Relation Extraction with Relation-Aware Score Calibration

Minseok Choi, Hyesu Lim, Jaegul Choo

IJCNLP-AACL 2023 Findings

Paper Code

HistRED: A Historical Document-Level Relation Extraction Dataset

Soyoung Yang, Minseok Choi, Youngwoo Cho, Jaegul Choo

ACL 202322.1% acceptance rate

Paper Code Data

Rethinking Style Transformer by Energy-based Interpretation: Adversarial Unsupervised Style Transfer using a Pretrained Model

Hojun Cho, Dohee Kim, Seungwoo Ryu, ChaeHun Park, Hyungjong Noh, Jeong-in Hwang, Minseok Choi, Edward Choi, Jaegul Choo

EMNLP 202219.8% acceptance rate

Paper Code

Development and Application of Web‑based Machine Learning Program for Automated Assessment Model Generation

Minseok Choi, Jaegul Choo, Minsu Ha

Brain, Digital, & Learning, Vol. 12, No. 4, pp. 567-578, 2022

Paper

Vitæ

Full CV in PDF.

• KAIST Mar 2026 - Present

  Postdoctoral Fellow
  Information & Electronics Research Institute

  Mandatory military service (Mar 2023 - Feb 2027)

  Military service is mandatory for men in Korea. Technical Research Personnel serve three years through research instead: two during the PhD and one after. I am completing that final year as a postdoc.
• Krafton May 2025 - Feb 2026

  Applied Research Intern
  [May-Nov 2025] PUBG Ally
  [Dec 2025-Feb 2026] Coding Agent
• Naver Webtoon Jan-Jun 2022

  Research Intern
  Keyphrase Extraction & Generation
• KAIST Mar 2020 - Feb 2026

  M.S.-Ph.D. Integrated Student (Advisor: Jaegul Choo)
  Data and Visual Analytics Lab
• University of Illinois at Urbana-Champaign Sep 2017 - May 2019

  B.S. in Computer Science
  with Honors