KIAS Center for AI and Natural Sciences 2026 Winter Workshop
Goseong (Rene Bleu by Sol Beach) | 2026-01-05 - 2026-01-08
All times in KST (UTC+9)
Day 1 (2026-01-05)
| Time | Speaker | Affiliation | Title | Type |
|---|---|---|---|---|
| 09:00-12:20 | — | — | KIAS → Goseong | Travel |
| 12:20-13:50 | — | — | Lunch | Lunch |
| 13:50-14:00 | — | — | Opening | Opening |
| 14:00-14:20 | Joon-Hyuk Ko | KIAS CAINS | Learning statistically accurate neural surrogate dynamics for chaotic systems through phase space neighborhood information | Research |
| 14:20-15:00 | Youngjoon Hong | Seoul National University | Deep Generative Models and Foundation Models for Scientific Problems | Invited |
| 15:00-15:20 | — | — | Break | Break |
| 15:20-15:40 | Sungyoon Lee | Hanyang University | SGD, Transformer, In-Context Learning, and Implicit Bias | Research |
| 15:40-16:00 | Jinseong Park | KIAS | Data Privacy in the Era of Generative AI | Research |
| 16:00-16:40 | Wonseok Hwang | University of Seoul | On the Reliability of Large Language Models | Invited |
| 16:40-17:00 | — | — | Break | Break |
| 17:00-17:20 | Seok Hyeong Lee | Seoul National University | Simple and explicit models for neural networks | Research |
| 17:20-17:40 | Yoonsoo Nam | KAIST | Revisiting Batch Normalization from Lazy/Rich Dynamics | Research |
| 17:40-18:10 | — | — | Poster Session | Poster |
| 18:10- | — | — | Dinner | Dinner |
Day 2 (2026-01-06)
| Time | Speaker | Affiliation | Title | Type |
|---|---|---|---|---|
| 09:00-09:20 | Geuntaek Seo | POSTECH | Asymptotic convergence of nonconvex Wasserstein gradient flows | Research |
| 09:20-09:40 | Carl-Fredrik Nyberg Brodda | KIAS HCMC | Decision problems in Artin and braid groups | Research |
| 09:40-10:20 | Donghwan Kim | KAIST | Statistical distances and optimization in generative models | Invited |
| 10:20-10:40 | — | — | Break | Break |
| 10:40-11:40 | Yoon Kim | Massachusetts Institute of Technology | Expressive and Efficient Architectures for Language Modeling | Tutorial |
| 11:40-12:20 | Yoon Kim | Massachusetts Institute of Technology | Expressive and Efficient Architectures for Language Modeling | Invited |
| 12:20-13:40 | — | — | Lunch | Lunch |
| 13:40-14:00 | Kwang Hyun Cho | KIAS | Accelerating Path Integral Molecular Dynamics Using Energy-Based Models | Research |
| 14:00-14:20 | Sejin Kim | KIAS | Inverse Problem in Physics & AI: Geometry Reconstruction from Entanglement Entropy using Transformer Models | Research |
| 14:20-15:00 | Hyunseung Kim | LG AI Research | Advancing Materials AI for a Better Life | Invited |
| 15:00-15:20 | — | — | Break | Break |
| 15:20-15:40 | Jaeyong Lee | Chung-Ang University | Extending Neural Operators: From Physics-Informed Learning to Numerical Integration Frameworks | Research |
| 15:40-16:00 | Sung Woong Cho | Inha University | Learning time-dependent PDE via graph neural networks and deep operator network for robust accuracy on irregular grids | Research |
| 16:00-16:40 | Seungwu Han | Seoul National University/ KIAS | Universal machine learning force fields and applications | Invited |
| 16:40-17:00 | — | — | Break | Break |
| 17:00-17:20 | Hoyun Choi | KIAS | A Deep Reinforcement Learning for Graph Generation with Arbitrary Constraints | Research |
| 17:20-17:40 | Youngkyoung Bae | Seoul National University | Anti-correlated Noise Provably Prefers Flat Minima of Neural Networks | Research |
| 17:40-18:20 | Junghyo Jo | Seoul National University | Multiplicative Learning from Observation-Prediction Ratios | Invited |
| 18:20- | — | — | Dinner (Banquet) | Dinner |
Day 3 (2026-01-07)
| Time | Speaker | Affiliation | Title | Type |
|---|---|---|---|---|
| 09:00-09:20 | Hyunwoo Lee | KIAS | Beyond Gaussian Initializations: Signal Preserving Weight Initialization for Odd-Sigmoid Activations | Research |
| 09:20-09:40 | Minjung Gim | NIMS | From Deterministic Optimal Control Problems to Stochastic Differential Games: A PINN-Based Policy Iteration Approach for Nonconvex HJI Equations | Research |
| 09:40-10:20 | Jy-yong Sohn | Yonsei University | Reasoning Capabilities of Foundation Models | Invited |
| 10:20-10:40 | — | — | Break | Break |
| 10:40-11:40 | Daniel Lee | KIAS | Tutorial on Stochastic Control and Schrödinger Bridges | Tutorial |
| 11:40-12:00 | Byung-Hak Hwang | KIAS | Formalizing mathematics: why it matters now | Research |
| 12:00-12:20 | Hyojae Lim | Johann Radon Institute for Computational and Applied Mathematics (RICAM) | New Perspectives on LLM-based Theorem Proving and Formalization | Research |
| 12:20-13:40 | — | — | Lunch | Lunch |
| 13:40-18:20 | — | — | Excursion | Excursion |
| 18:20- | — | — | Dinner | Dinner |
Day 4 (2026-01-08)
| Time | Speaker | Affiliation | Title | Type |
|---|---|---|---|---|
| 09:00-09:20 | Junho Choi | Korea Advanced Institute of Science and Technology | A physics-Informed operator learning network for solving the 3D Navier-Stokes equations | Research |
| 09:20-09:40 | Jinsook Kim | KIAS | Why Do Large Language Models Work? | Research |
| 09:40-10:20 | Jaemin Seo | Chung-Ang University | Machine learning for nuclear fusion: plasma diagnosis, prediction, and control | Invited |
| 10:20-10:40 | — | — | Break | Break |
| 10:40-11:00 | Jongmin Lee | Pusan National University | Huber means on Riemannian manifolds | Research |
| 11:00-11:20 | Dongwon Kim | KIAS | On the Theoretical Foundations of Neural Network–Based Time Series: Stationarity, Ergodicity, and Generalization | Research |
| 11:20-11:40 | Jaehoon Chung | KIAS | Nearest Neighbor Search on the Hypersphere: Methods, Limits, and Data-Aware Design | Research |
| 11:40-12:20 | Jaewoong Choi | Sungkyunkwan University | Generative Modeling via Neural Optimal Transport | Invited |
| 12:20-13:40 | — | — | Lunch | Lunch |
| 13:40-17:00 | — | — | Goseong -> KIAS | Travel |
Generated from KIAS Center for AI and Natural Sciences 2026 Winter Workshop Program