Seokil Ham (함석일)

Education

• 2023.03 - Present      Ph.D. (School of Electrical Engineering), Korea Advanced Institute of Science Technology, South Korea

• 2021.03 - 2023.02      M.S. (School of Electrical Engineering), Korea Advanced Institute of Science Technology, South Korea

• 2017.03 - 2021.02      B.S. (School of Electrical Engineering), Chung-Ang University, South Korea

Research Interests

• Diffusion Model

• Large Language Model

• AI Safety

Projects

• KaKao Bank - Prompt Attack Detection (2024 - 2025)

    - Detecting toxic prompts for Large Language Model (LLM) safety.

• ETRI - Digital Human Creation (2023 - 2024) 

- Generating human parsing masks using human semantic segmentation.

• SK Hynix - KAIST Next Generation Artificial Intelligence Semiconductor System Research Center (2021 - 2022)

- Adapting anytime prediction to segementation model for low-latency multi-modal 3D object detection.

Publications

International Conference

• Seokil Ham, Hee-Seon Kim, Sangmin Woo, and Changick Kim, "Parameter Efficient Mamba Tuning via Projector-targeted Diagonal-centric Linear Transformation", in Proc. the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),  Jun. 2025.

• Seokil Ham*, Sangmin Woo*, J.-Y. Kim, H. Go, Byeongjun Park, and Changick Kim, "Diffusion Model Patching via Mixture-of-Prompts", in Proc. the AAAI conference on artificial intelligence (AAAI), Feb 2025. (*=equal contribution)

• Byeongjun Park, H. Go, J.-Y. Kim, Sangmin Woo, Seokil Ham, and Changick Kim, “Switch Diffusion Transformer: Synergizing Denoising Tasks with Sparse Mixture-of-Experts,” in Proc. IEEE/CVF European Conference on Computer Vision (ECCV), Oct 2024.

• Seokil Ham, J. Park, D.-J. Han, and J. Moon, "NEO-KD: Knowledge-Distillation-Based Adversarial Training for Robust Multi-Exit Neural Networks ", in Proc. the 37th Annual Conference on Neural Information Processing Systems (NeurIPS), Dec. 2023. 

•  D.-J. Han*, J. Park*, Seokil Ham, N. Lee and J. Moon, "Training Multi-Exit Architectures via Block-Dependent Losses for Anytime Inference," CVPR Workshop on Dynamic Neural Networks Meet Computer Vision, June 2022. (*=equal contribution)

International Journal

• D.-J. Han*, J. Park*, Seokil Ham, N. Lee and J. Moon, "Improving Low-Latency Predictions in Multi-Exit Neural Networks via Block-Dependent Losses," IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2023. (*=equal contribution)