CVReading
GN

Giung Nam

Ph.D. Student @ KAIST AI

Seoul, South Korea
{first_name} [at] kaist [dot] ac [dot] kr
GitHub|HuggingFace|LinkedIn
Google Scholar|Semantic Scholar|DBLP
I'm a Ph.D. candidate at the Statistical Inference and Machine Learning (SIML) laboratory in KAIST AI, advised by Juho Lee. My research is grounded in the philosophy that today's posterior is tomorrow's prior, where the current model serves as the foundational prior for the next generation of models.
The rapid evolution of deep learning has established pre-trained foundation models as the bedrock of modern AI systems. As this landscape continues to evolve, I focus on investigating how to effectively transfer, adapt, and compress their knowledge to catalyze the development of future models that are stronger in capability and lighter in cost. My research covers transfer learning, knowledge distillation, and model compression.
Beyond my own research, I am deeply committed to contributing to the academic community through rigorous peer review. Since 2022, I have regularly reviewed for major AI venues such as ICML, NeurIPS, and ICLR. I was honored to be recognized as a Top Reviewer at NeurIPS for three consecutive years (2022-2024).

Education

KAIST

Ph.D., Kim Jaechul Graduate School of AI
Sep 2022 - Feb 2026 (Expected)
Daejeon, South Korea

KAIST

M.S., Kim Jaechul Graduate School of AI
Sep 2020 - Aug 2022
Daejeon, South Korea

Korea University

B.S., Department of Computer Science and Engineering
Mar 2017 - Aug 2020
Seoul, South Korea
  • GPA: 4.32/4.50

Position

KAIST

Technical Research Personnel, Kim Jaechul Graduate School of AI
Mar 2024 - Feb 2027 (Expected)
Daejeon, South Korea
  • Fulfilling mandatory military service through a specialized research position.

ETRI

Research Intern, Visual Intelligence Laboratory
Jul - Aug 2019; Jan - Feb, Jul - Aug 2020 (Intermittent)
Daejeon, South Korea

VisualCamp

Research Intern, Research and Engineering Team
Mar 2019 - Jun 2019
Seongnam, South Korea
  • Developed AI-based software for eye tracking.

Publication

Improving constrained language generation via self-distilled twisted sequential Monte Carlo
Sooyeon Kim, Giung Nam, Byoungwoo Park, Juho Lee
In the Second Edition of Frontiers in Probabilistic Inference: Learning meets Sampling, Dec 2025.
FPI@NeurIPS
PANGEA: projection-based augmentation with non-relevant general data for enhanced domain adaptation in LLMs
Seungyoo Lee, Giung Nam, Moonseok Choi, Hyungi Lee†, Juho Lee† (†: Equal corresponding)
In the Thirty-Ninth Annual Conference on Neural Information Processing Systems, Dec 2025.
NeurIPS
Ensemble distribution distillation via flow matching
Jonggeon Park*, Giung Nam*, Hyunsu Kim, Jongmin Yoon, Juho Lee (*: Equal contribution)
In the Forty-Second International Conference on Machine Learning, Dec 2021.
ICML
Parameter expanded stochastic gradient Markov chain Monte Carlo
Hyunsu Kim, Giung Nam, Chulhee Yun, Hongseok Yang, Juho Lee
In the Thirteenth International Conference on Learning Representations, Apr 2025.
ICLR
2024
Ex uno pluria: insights on ensembling in low precision number systems
Giung Nam, Juho Lee
In the Thirty-Eighth Conference on Neural Information Processing Systems, Dec 2024.
NeurIPS
Sparse weight averaging with multiple particles for iterative magnitude pruning
Moonseok Choi*, Hyungi Lee*, Giung Nam*, Juho Lee (*: Equal contribution)
In the Twelfth International Conference on Learning Representations, May 2024.
ICLR
Enhancing transfer learning with flexible nonparametric posterior sampling
Hyungi Lee*, Giung Nam*, Edwin Fong, Juho Lee (*: Equal contribution)
In the Twelfth International Conference on Learning Representations, May 2024.
ICLR
Lipsum-FT: robust fine-tuning of zero-shot models using random text guidance
Giung Nam, Byeongho Heo, Juho Lee
In the Twelfth International Conference on Learning Representations, May 2024.
ICLR
2023
Traversing between modes in function space for fast ensembling
Eunggu Yun*, Hyungi Lee*, Giung Nam*, Juho Lee (*: Equal contribution)
In the Fortieth International Conference on Machine Learning, Jul 2023.
ICML
Martingale posterior neural processes
Hyungi Lee, Eunggu Yun, Giung Nam, Edwin Fong, Juho Lee
In the Eleventh International Conference on Learning Representations, May 2023.
ICLR
Decoupled training for long-tailed classification with stochastic representations
Giung Nam*, Sunguk Jang*, Juho Lee (*: Equal contribution)
In the Eleventh International Conference on Learning Representations, May 2023.
ICLR
2022
Benefits of stochastic weight averaging in developing neural network radiation scheme for numerical weather prediction
Hwan-Jin Song, Soonyoung Roh, Juho Lee, Giung Nam, Eunggu Yun, Jongmin Yoon, Park Sa Kim
In the Journal of Advances in Modeling Earth Systems, Oct 2022.
JAMES
Improving ensemble distillation with weight averaging and diversifying perturbation
Giung Nam, Hyungi Lee, Byeongho Heo, Juho Lee
In the Thirty-ninth International Conference on Machine Learning, Jul 2022.
ICML
2021
Diversity matters when learning from ensembles
Giung Nam*, Jongmin Yoon*, Yoonho Lee, Juho Lee (*: Equal contribution)
In the Thirty-Fifth Conference on Neural Information Processing Systems, Dec 2021.
NeurIPS