Preprints & Under Review
Collision-to-Preference: Post-Training Vision-Language-Action Policies for Collision Avoidance
Under Review
Hyunse Lee (First author), Yunyeong Chang (Co-author), Soohwan Song, Woojin Lee
#Vision-Language-Action#Collision Avoidance#Preference Optimization
PLCWorld: Benchmarking LLM-Generated PLC Programs in Closed-Loop Plant SimulationarXiv:2610.02982
Under Review
Yunji Kim (Co-first author), Yunseok Lee (Co-first author), Hyunwoo Seo, Jaerim Choi, Woojin Lee
#LLM Code Generation#PLC#Benchmark
ChainQABench: A Benchmark for Evidence-Grounded Question Answering over Blockchain Data
Under Review
Seungbin Kim (Co-first author), Seunghyeon Jung (Co-first author), Byungsoo Kang, Seongchan Park, Gaeun Kim, Hyunho Jeon, Daehyun Kim, Woojin Lee, Huisu Jang
#Blockchain#Question Answering#Benchmark
Anchor and Follower: Narrowing Valid Answers for Parallel Decoding in Diffusion Language Models
Under Review
Seungwon Jeong (First author), Minjun Kim (Co-author), Woojin Lee
#Diffusion Language Model#Parallel Decoding#Inference
Keep Safe and Carry On: Knowing When to Apply Safety Mitigation in Embodied Agents
Jiwoo Jeong, Haneul Lee, Georu Lee, Kyochul Jang, Youngjae Yu, Woojin Lee
#Embodied Agent#AI Safety#Memory
ĪCÁRŪS: Flying Over Safety Alignment to Jailbreak Large Language Models via Ancient Grammar
Seungwon Jeong, Kim Seung Su, Woojin Lee
#Jailbreak#LLM Safety#Ancient Language
Trigger Injection Attack: Conditional Indirect Prompt Injection on LLM Agents
Yunseok Lee, Yunji Kim, Woojin Lee
#Agentic LLM#Indirect Prompt Injection#Security
Machine Unlearning for Masked Diffusion Language ModelsarXiv:2605.18253
Georu Lee, Seungwon Jeong, Hoki Kim, Jinseong Park, Woojin Lee
#Machine Unlearning#Diffusion Language Model
SeMoBridge: A Semantic Modality Bridge for Intra-Modal Alignment in CLIP
Christoph Timmermann, Hyunse Lee, Woojin Lee
#Vision-Language#CLIP#Few-Shot Learning