최민영(Choe, Min-Yeong)

• 2023~2024 학부연구생

• 2025.03 ~ 2026.08 석사과정

• 2026.09 ~ 박사과정

• 기존 연구분야 포함 담당 및 관심분야

  • Backdoor attacks for language model in federated learning

  • Privacy aspects in federated learning - gradient inversion, data reconstruction

  • Knowledge-aware approaches for efficiency, security and privacy for Transformer-based language model

    • Knowledge Editing

  • AI Safety with Interpretability

    • Interpretability of safety and refusal behavior in LLMs, with a focus on jailbreak robustness

    • Analysis of sycophantic behavior in LLMs 

김승한(Kim, Seung-Han)

• 2024 학부연구생

• 2025.03 ~ 2026.08 석사과정

• 2026.09 ~ 박사과정

• 기존 연구분야 포함 담당 및 관심분야

  • Backdoor attacks for language model in federated learning

  • Interpretability of Transformer-based language model

  • Knowledge-aware approaches for efficiency, security and privacy for Transformer-based language model

    • Knowledge Editing

  • AI Safety with Interpretability

    • Interpretability for Language Domain

    • Cross-lingual mechanistic analysis of multi-hop reasoning in LLMs

    • Safety vulnerabilities of LLMs in low-resource languages 

임창훈(Lim, Chang-Hoon)

• 2025.01 ~ 2026.02 학부연구생

• 2026.03 ~ 석사과정

• 기존 연구분야 포함 담당 및 관심분야

  • Backdoor attacks for language model in federated learning

  • Knowledge-aware approaches for efficiency, security and privacy for Transformer-based language model

    • Model quantization

    • AI Safety with Model quantization

  • AI Safety with Interpretability

    • Interpretability of abstention behavior in LLMs, with a focus on hallucination robustness