Yunseok Lee 이윤석

PLCWorld: Benchmarking LLM-Generated PLC Programs in Closed-Loop Plant Simulation

Yunji Kim^*, Yunseok Lee^*, Hyunwoo Seo, Jaerim Choi, Woojin Lee

arXiv preprint, 2026

100개 과제의 폐루프 플랜트 시뮬레이션에서 LLM이 생성한 PLC 프로그램을 실행하고, 작업 성공과 안전 위반을 별도로 평가합니다.

Evaluating LLM-generated PLC programs through closed-loop plant simulation across 100 tasks, measuring task success and safety violations separately.

@misc{kim2026plcworld,
  title={{PLCWorld}: Benchmarking {LLM}-Generated {PLC} Programs in Closed-Loop Plant Simulation},
  author={Yunji Kim and Yunseok Lee and Hyunwoo Seo and Jaerim Choi and Woojin Lee},
  year={2026},
  eprint={2610.02982},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2610.02982}
}

Will the User Ever Know? Covert Indirect Prompt Injection Attacks on Tool-Using LLM Agents

Yunseok Lee^*, Yunji Kim^*, Woojin Lee

EMNLP 2026 Main Conference (Oral)

도구 사용 Agent의 공격 성공과 사용자에게 드러나는 정도를 구분하여 은밀한 간접 프롬프트 주입을 평가합니다.

Evaluating covert indirect prompt injection by separating attack success from what a tool-using agent reveals to its user.

@misc{lee2026will,
  title={Will the User Ever Know? Covert Indirect Prompt Injection Attacks on Tool-Using {LLM} Agents},
  author={Yunseok Lee and Yunji Kim and Woojin Lee},
  year={2026},
  eprint={2608.30362},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  note={EMNLP 2026 Main Conference (Oral)},
  url={https://arxiv.org/abs/2608.30362}
}

What Did You Do Behind My Back?! Covert Indirect Prompt Injection on Tool-Using LLM Agents

Yunseok Lee^*, Yunji Kim^*, Woojin Lee

ICML 2026 Workshop on Failure Modes of Agentic AI (FAGEN)

사용자가 알아차리기 어려운 도구 사용 LLM Agent의 간접 프롬프트 주입 공격을 연구합니다.

Studying indirect prompt injection attacks on tool-using LLM agents that can go unnoticed by users.

@inproceedings{lee2026what,
  title={What Did You Do Behind My Back?! Covert Indirect Prompt Injection on Tool-Using {LLM} Agents},
  author={Yunseok Lee and Yunji Kim and Woojin Lee},
  booktitle={Workshop on Failure Modes of Agentic AI at ICML 2026},
  year={2026},
  url={https://openreview.net/forum?id=ozJ54HKqBe}
}

SlotGCG: Exploiting the Positional Vulnerability in LLMs for Jailbreak Attacks

Seungwon Jeong, Jiwoo Jeong, Hyeonjin Kim, Yunseok Lee, Woojin Lee

The Fourteenth International Conference on Learning Representations (ICLR), 2026

프롬프트 내부의 위치별 취약성을 분석하고, 취약한 위치에 공격 토큰을 배치하는 탈옥 방법을 제안합니다.

Analyzing positional vulnerabilities inside prompts to place adversarial tokens at vulnerable slots.

@inproceedings{jeong2026slotgcg,
  title={Slot{GCG}: Exploiting the Positional Vulnerability in {LLM}s for Jailbreak Attacks},
  author={Seungwon Jeong and Jiwoo Jeong and Hyeonjin Kim and Yunseok Lee and Woojin Lee},
  booktitle={The Fourteenth International Conference on Learning Representations},
  year={2026},
  url={https://openreview.net/forum?id=Fn2rSOnpNf}
}

Explainable Graph-Based Golf Swing Analysis Integrating Club and Body Keypoints for Ball Flight Outcome Prediction

Seunghyeon Jung, Minseok Kim, Hyeonjin Kim, Seungwon Jeong, Yunseok Lee, Yunji Kim, Hyunse Lee, Seoyoung Hong, Gyumin Choi, Jaerim Choi, Woojin Lee

Applied Sciences, 16(8), 3813, 2026

클럽과 신체 키포인트를 통합한 그래프 신경망으로 공의 비행 결과를 예측하고, Integrated Gradients로 스윙 단계별 주요 키포인트의 기여도를 분석합니다.

Predicting ball flight outcomes from club and body keypoints with graph neural networks, and interpreting keypoint contributions across swing phases using Integrated Gradients.

@article{jung2026explainable,
  title={Explainable Graph-Based Golf Swing Analysis Integrating Club and Body Keypoints for Ball Flight Outcome Prediction},
  author={Seunghyeon Jung and Minseok Kim and Hyeonjin Kim and Seungwon Jeong and Yunseok Lee and Yunji Kim and Hyunse Lee and Seoyoung Hong and Gyumin Choi and Jaerim Choi and Woojin Lee},
  journal={Applied Sciences},
  volume={16},
  number={8},
  pages={3813},
  year={2026},
  doi={10.3390/app16083813},
  url={https://doi.org/10.3390/app16083813}
}