Song Park
Scholar

Song Park

Google Scholar ID: VR1c0H8AAAAJ
NAVER AI Lab
Computer Vision
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
1,414
 
H-index
10
 
i10-index
10
 
Publications
14
 
Co-authors
0
 
Contact
Emailsong.pxxk@gmail.comCVOpen ↗GitHubOpen ↗
Publications
5 items
Isotropic Embedding Perturbations for Robust Vision Language Encoders
2026
Cited
0
Seeing What You Say: Expressive Image Generation from Speech
2025
Cited
1
DNNs May Determine Major Properties of Their Outputs Early, with Timing Possibly Driven by Bias
2025
Cited
0
MaskRIS: Semantic Distortion-aware Data Augmentation for Referring Image Segmentation
arXiv.org · 2024
Cited
1
Probabilistic Language-Image Pre-Training
arXiv.org · 2024
Cited
1
Resume
Academic Achievements
  • DNNs May Determine Major Properties of Their Outputs Early, with Timing Possibly Driven by Bias (2025)
  • Probabilistic Language-Image Pre-Training (2025)
  • Rotary position embedding for vision transformer (2024)
  • SeiT++: Masked Token Modeling Improves Storage-efficient Training (2024)
  • Similarity of neural architectures using adversarial attack transferability (2024)
Research Experience
  • Worked as a Research Scientist at NAVER AI Lab from 2022 to 2025.
Education
  • Received Ph.D. and M.S. in Integrated Technology from Yonsei University in 2022, supervised by Hyunjung Shim; B.S. in Integrated Technology also from Yonsei University.
Background
  • Research interests include understanding how deep neural networks perceive and process diverse visual concepts to enhance structured visual representations for real-world applications. Specializes in text-to-image generative models and visual representation learning.