junhyuk so
Scholar

junhyuk so

Google Scholar ID: uYE7vtEAAAAJ
POSTECH
deep learningefficient airepresentation learningcomputer system
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
248
 
H-index
6
 
i10-index
5
 
Publications
11
 
Co-authors
0
 
Contact
Emailjunhyukso@postech.ac.krCVOpen ↗GitHubOpen ↗LinkedInOpen ↗
Publications
11 items
Browse publications on Google Scholar (top-right) ↗
Resume (English only)
Academic Achievements
  • Publications:
  • - Maximal Coupling Speculative Jacobi Decoding for Autoregressive Image Generation (Under Review)
  • - Improving Generative Behavior Cloning via Self-Guidance and Adaptive Chunking (NeurIPS 2025, to appear)
  • - Grouped Speculative Decoding for Autoregressive Image Generation (ICCV 2025)
  • - PCM: Picard Consistency Model for Fast Parallel Sampling of Diffusion Models (CVPR 2025)
  • - FRDiff: Feature Reuse for Universal Training-free Acceleration of Diffusion Models (ECCV 2024)
  • - Temporal Dynamic Quantization for Diffusion Models (NeurIPS 2023)
  • - Geodesic Multi-Modal Mixup for Robust Fine-Tuning (NeurIPS 2023)
  • - NIPQ: Noise Proxy-Based Integrated Pseudo-Quantization (CVPR 2023)
  • - Robust Contrastive Learning With Dynamic Mixed Margin (IEEE Access 2023)
  • - Learning Fair Representation via Distributional Contrastive Disentanglement (KDD 2022)
  • - Exploiting Activation Sparsity for Fast CNN Inference on Mobile GPUs (ESWeek(CODES+ISSS) and ACM TECS (journal track), 2021)
Research Experience
  • Conducts research in both Efficient Computing Lab and Machine Learning Lab.
Education
  • Started Ph.D. in Computer Science and Engineering at POSTECH in 2022, advised by Prof. Eunhyeok Park; previously received B.S. in Electrical and Computer Engineering from the University of Seoul.
Background
  • Research Interests: Building efficient machine learning algorithms, especially for generative models. Most of the work involves designing efficient inference algorithms for vision generation. Deeply interested in general ML topics such as optimization or numerical methods.
Miscellany
  • Academic Services: Reviewer for NeurIPS 24-25, CVPR 24-26, ICLR 25-26, AAAI 25, WACV 26.
  • Talks: Recent Topics on Image generation Acceleration @ Squeezebits. 25.06.30
  • Teaching: TA for Introduction to Artificial Intelligence (CSED105) and Implementation and Acceleration of Machine Learning (AIGS510-01)
Co-authors
0 total
Co-authors: 0 (list not available)