About

I am a Ph.D. candidate in Electronic Engineering at Sogang University, advised on efficient deep learning. I was a visiting researcher at Carnegie Mellon University (2024–2025) through the Short Term Scholarship program in Software and Societal Systems.

My research focuses on designing models that are not only accurate but efficient enough for real-world deployment — including lightweight transformer architectures, sparse attention mechanisms, token pruning, and vision-language model compression.

Research Interests

• Lightweight Architecture: Transformer-based efficient architectures for vision tasks
• Inference Acceleration: Sparse attention, KV cache compression, token pruning, parameter-efficient tuning, weight pruning
• Generative Models: Video diffusion, diffusion multimodal language models, large language models, LLM agents

Education

Publications (Total citations: 280)

• OpenAttention: One Percent Is Enough with Tail Proxies
  ICLR, 2027 Under Review
• Exploiting Weight Space Symmetries for Vision Attention Compression
  ICLR, 2027 Under Review
• Towards a Low-Delta Oracle Policy for Improved Upper Bound in Sparse Attention
  ICLR, 2027 Under Review
• Train Wide, Act Narrow: The Learning Geometry of Visual Adaptation
  ICLR, 2027 Under Review
• ResceneKV: Reconstruction-Error Selective KV Cache Eviction across Nested Scales for Visual Autoregressive Modeling
  ICLR, 2027 Under Review
• PACE: Prune Aggressively and Commit Early for Diffusion VLM Inference Acceleration
  ICLR, 2027 Under Review
• Content-aware Token Compression for Efficient Video LLMs
  AAAI, 2027 Under Review
• Compressed Visual Tokens Are Not Equal: Volume Calibration for Efficient Video Large Language Models
  AAAI, 2027 Under Review
• Future-Query-Guided KV Cache Compression for Visual Autoregressive Modeling
  AAAI, 2027 Under Review
• SparSTAR: Sparse Attention for SpatioTemporal Autoregressive Video Synthesis
  AAAI, 2027 Under Review

• Post-Training Quantization with Gradient-Projected Fisher Approximation for Vision Transformers
  The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026 NeurIPS
• SPA-Q: Structure-Preserving Adaptive Post-Training Quantization for Monocular Depth Estimation
  The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026 NeurIPS
• WIMFRIS: WIndow Mamba Fusion and Parameter Efficient Tuning for Referring Image Segmentation
  The Fourteenth International Conference on Learning Representations (ICLR), 2026 ICLR
• EDAFormer: Embedding-Free Transformer with Inference Spatial Reduction for Efficient Semantic Segmentation
  The 18th European Conference on Computer Vision (ECCV), 2024 ECCV
• MetaSeg: MetaFormer-Based Global Contexts-Aware Network for Efficient Semantic Segmentation
  IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2024 WACV
• FeedFormer: Revisiting Transformer Decoder for Efficient Semantic Segmentation
  Association for the Advancement of Artificial Intelligence (AAAI), 2023 AAAI
• Optimization of Video Repetitive Action Counting for Efficient Inference on Edge Devices
  International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC), 2023
• Deep Learning-based Real-time Segmentation for Edge Computing Devices
  IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), 2022
• Vision Transformer-based Retina Vessel Segmentation With Deep Adaptive Gamma Correction
  IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2022
• A study on the improvement of segmentation performance by ViT-based skip architecture of U-Net
  The Institute of Electronics and Information Engineers, 2021 Best Paper

• Reducing Attention Distribution Error with Unified Tail Aggregation for Sparse Attention
  ICML 2026 Workshop on Resource-Adaptive Foundation Model Inference (ICML Workshop), 2026
• Decoupling Spatial and Semantic Token Compression for Vision-Language Model Acceleration
  ICML 2026 Workshop on Resource-Adaptive Foundation Model Inference (ICML Workshop), 2026
• Structure-Preserving Adaptive Post-Training Quantization for Monocular Depth Estimation
  ICML 2026 Workshop on Resource-Adaptive Foundation Model Inference (ICML Workshop), 2026
• Implicit Off-Diagonal Curvature Modeling via Gradient Projection for Post-Training Quantization of Vision Transformers
  ICML 2026 Workshop on Resource-Adaptive Foundation Model Inference (ICML Workshop), 2026
• OriCache: Orientation-Guided Feature Caching for DiT Acceleration
  ICML 2026 Workshop on Resource-Adaptive Foundation Model Inference (ICML Workshop), 2026
• Attention Weight Decomposition for Vision Model Compression
  ICML 2026 Workshop on Weight-Space Symmetries (ICML Workshop), 2026

• VIPA: Visual Informative Part Attention for Referring Image Segmentation
  IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2026 (IF 10.8 / Q1) Journal
• Deep Learning-based Real-Time Driver's Hands-On Detection: A Lightweight Time Series Approach Using CAN Data
  IEEE Access, 2026 Journal
• Cross-aware Early Fusion with Stage-divided Vision and Language Transformer Encoders for Referring Image Segmentation
  IEEE Transactions on Multimedia (TMM), 2023 Journal

Projects

Awards

• Best Paper Award
  Department of Electronic Engineering, Sogang University
  2024.12
• Encouragement Award
  Samsung Display
  2024.10
• Excellent Paper Award
  Samsung Electronics
  2024.09
• Best Paper Award
  Summer Conference, The Institute of Electronics and Information Engineers
  2021.10

Academic Service

• Gold Technical Reviewer, ICML 2026
• Reviewer, ECCV 2026
• Reviewer, ICML 2026
• Reviewer, CVPR 2026
• Reviewer, WACV 2026
• Reviewer, AAAI 2026
• Reviewer, ICCV 2025

Skills

Programming: Python, C
Domain Knowledge: Image processing, video processing, natural language processing, time series, signal processing