I am a PhD student at System Intelligence lab in KAIST (Korea Advanced Institute of Science and Technology). I am fortunate to be advised by Professor Jinkyoo Park. Previously, I was a research intern at NaverLabs Europe and Mila, and a visiting intern at HKUST. Here is my cv.
My research interest lies in building generative agents that make reliable decisions in the real world. Simulators of real-world systems are becoming easier to build with the help of LLMs, but they never perfectly reflect reality or user intent. I am interested in how to align simulators with user intent and how to train generative policies that stay robust under such uncertainty. To this end, my research focuses on user-simulator alignment, reinforcement learning under uncertainty, and black-box optimization.
During my earlier PhD years, I studied amortized samplers, such as GFlowNets, that learn diverse generative policies for text, image, and biological sequence generation. I also worked on transportation-related projects, such as traffic light management and network layout design, where I formulated each task as a black-box optimization problem over an existing simulator.
🔥 News
- 2026.09: 🎉🎉 8 papers accepted in 2026 (1 WSDM, 1 TMLR, 1 ICLR, 1 ICML, 1 CIKM, 3 NeurIPS)
- 2026.08: 🎉🎉 I started internship at NaverLabs Europe, Hosted by Darko Drakulic and Sofia Michel
- 2025.12: 🎉🎉 5 papers accepted in 2025 (1 ICLR, 1 CVPR, 2 ICML, 1 KDD)
- 2025.06: 🎉🎉 I started internship at Mila, Hosted by Yoshua Bengio
- 2024.12: 🎉🎉 5 papers accepted in 2024 (1 ICLR, 1 ICML, 1 KDD, 2 NeurIPS)
- 2024.09: 🎉🎉 I started internship at HKUST, Hosted by Ling Pan
📖 Educations
- 2024.03 - current, Ph.D in Industrial and Systems Engineering, KAIST
- 2022.09 - 2024.02, M.S in Graduate School of AI, KAIST
- 2018.03 - 2022.08, B.S in Industrial and Systems Engineering & Computer Science, KAIST
💻 Internships
- 2026.08 - 2026.11, Research Intern at NaverLabs Europe, hosted by Darko Drakulic and Sofia Michel
- 2025.06 - 2025.11, Research Intern at Mila, hosted by Yoshua Bengio
- 2024.09 - 2025.03, Visiting Intern at HKUST, hosted by Ling Pan
- 2021.03 - 2021.08, Research Intern at Kakao Recommendation Team
📝 Publications
(*: Equal Contribution)
Preprints
-
[P4] Noise-Space Reward Alignment for Few-Step Generative Models
[paper] / [code]
Woocheol Shin, Taeyoung Yun, Jiwon Sung, Hyeongyu Kang, Jeongmin Park, Jaewoo Lee, Kyuil Sim, Dohyun Kim, Yutong He, Jinkyoo Park
Arxiv 2026 -
[P3] Decoupling Exploration and Exploitation in GFlowNets via Policy Evaluation
[paper] / [code]
Taeyoung Yun*, Inhyuck Song*, Jinkyoo Park, Ling Pan
Arxiv 2026 -
[P2] Robust Exploration through Generative Replay
[paper] / [code]
Inhyuck Song, Taeyoung Yun, Jaewoo Lee, Jinkyoo Park
Arxiv 2026 ICML 2026 RLxF Workshop -
[P1] Improving Sampling Distribution of Off-policy Training in Generative Flow Networks
[paper] / [code]
Taeyoung Yun, Sujin Yun, Jinkyoo Park, Ling Pan
Arxiv 2025
Conference Publications
-
[C17] Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization
[paper] / [code]
Kiyoung Om*, Kyuil Sim*, Taeyoung Yun*, Hyeongyu Kang, Jinkyoo Park
NeurIPS 2026 NeurIPS 2025 SPIGM Workshop -
[C16] Aligning Few-Step Generative Models by Amortizing Sample-based Variational Inference
[paper] / [code]
Jaewoo Lee*, Hyeongyu Kang*, Dohyun Kim, Kyuil Sim, Woocheol Shin, Minsu Kim, Taeyoung Yun, Jeongjae Lee, Sanghyeok Choi, Tabitha Edith Lee, Jongchul Ye, Jinkyoo Park
NeurIPS 2026 ICML 2026 SPIGM Workshop -
[C15] Automated Kernel Discovery for Understanding High-dimensional Bayesian Optimization
[paper] / [code]
Taeyoung Yun*, Woocheol Shin*, Inhyuck Song, Jaewoo Lee, Jinkyoo Park
NeurIPS 2026 ICML 2026 DEMO Workshop -
[C14] SUMO-SimQA: A Benchmark for LLM Cross-File Reasoning over Traffic Simulation Files
[paper] / [code]
Jeeyun Chang*, Taeyoung Yun*, Minwoo Jeong, Jinkyoo Park, Yoonjin Yoon
CIKM 2026 Short Paper -
[C13] Active Attacks: Red-teaming LLMs via Adaptive Environments
[paper] / [code]
Taeyoung Yun, Pierre-Luc St-Charles, Jinkyoo Park, Yoshua Bengio, Minsu Kim
ICML 2026 -
[C12] Diffusion Alignment as Variational Expectation-Maximization
[paper] / [code]
Jaewoo Lee, Minsu Kim, Sanghyeok Choi, Inhyuck Song, Sujin Yun, Hyeongyu Kang, Woocheol Shin, Taeyoung Yun, Kiyoung Om, Jinkyoo Park
ICLR 2026 -
[C11] Urban Traffic Network Layout Optimization with Guided Discrete Diffusion Models
[paper] / [code]
Taeyoung Yun, Inhyuck Song, Woocheol Shin, Yujin Shin, Sungpil Woo, Sunhwan Lim, Jinkyoo Park
WSDM 2026 -
[C10] Wind Farm Layout Optimization with Diffusion Models
[paper] / [code]
Yujin Shin*, Taeyoung Yun*, Sujin Yun, Sungpil Woo, Sunhwan Lim, Jinkyoo Park
KDD 2025 -
[C9] Posterior Inference with Diffusion Models for High-dimensional Black-box Optimization
[paper] / [code]
Taeyoung Yun*, Kiyoung Om*, Jaewoo Lee, Sujin Yun, Jinkyoo Park
ICML 2025 ICLR 2025 FPI Workshop -
[C8] Improved Off-policy Reinforcement Learning in Biological Sequence Design
[paper] / [code]
Hyeonah Kim, Minsu Kim, Taeyoung Yun, Sanghyeok Choi, Emmanuel Bengio, Alex Hernández-García, Jinkyoo Park
ICML 2025 NeurIPS 2024 AIDrugX Workshop -
[C7] Learning to Sample Effective and Diverse Prompts for Text-to-Image Generation
[paper] / [code]
Taeyoung Yun, Dinghuai Zhang, Jinkyoo Park, Ling Pan
CVPR 2025 Qualcomm Innovation Fellowship Korea Finalist -
[C6] Adaptive Teachers for Amortized Samplers
[paper] / [code]
Minsu Kim*, Sanghyeok Choi*, Taeyoung Yun, Emmanuel Bengio, Leo Feng, Jarrid Rector-Brooks, Sungsoo Ahn, Jinkyoo Park, Nikolay Malkin, Yoshua Bengio
ICLR 2025 -
[C5] Guided Trajectory Generation with Diffusion Models for Offline Model-based Optimization
[paper] / [code]
Taeyoung Yun, Sujin Yun, Jaewoo Lee, Jinkyoo Park
NeurIPS 2024 -
[C4] GTA: Generative Trajectory Augmentation with Guidance for Offline Reinforcement Learning
[paper] / [code]
Jaewoo Lee*, Sujin Yun*, Taeyoung Yun, Jinkyoo Park
NeurIPS 2024 ICLR 2024 GenAI4DM Workshop -
[C3] An Offline Meta Black-box Optimization Framework for Adaptive Design of Urban Traffic Light Management Systems
[paper] / [code]
Taeyoung Yun*, Kanghoon Lee*, Sujin Yun, Ilmyung Kim, Won-Woo Jung, Min-Cheol Kwon, Kyujin Choi, Yoohyeon Lee, Jinkyoo Park
KDD 2024 -
[C2] Learning to Scale Logits for Temperature-conditional GFlowNets
[paper] / [code]
Minsu Kim*, Juhwan Ko*, Taeyoung Yun*, Dinghuai Zhang, Ling Pan, Woochang Kim, Jinkyoo Park, Yoshua Bengio
ICML 2024 NeurIPS 2023 AI4Science Workshop -
[C1] Local Search GFlowNets
[paper] / [code]
Minsu Kim, Taeyoung Yun, Emmanuel Bengio, Dinghuai Zhang, Yoshua Bengio, Sungsoo Ahn, Jinkyoo Park
ICLR 2024 Spotlight
Journal Publications
- [J1] Offline Model-Based Optimization: Comprehensive Review
[paper] / [code]
Minsu Kim*, Jiayao Gu*, Ye Yuan, Taeyoung Yun, Zixuan Liu, Yoshua Bengio, Can Chen
TMLR 2026 Survey Certification
🎖 Honors and Awards
- 2025.11 Qualcomm Innovation Fellowship Korea Finalist 2025
- 2025.10 Top Reviewer in NeurIPS 2025
- 2021.12 Excellence Award on DACON NH Big Data Competition.
- 2021.09 Dean’s List (Honor for Top 2% Students).
💬 Invited Talks
- 2026.05 AutoML Seminar - Posterior Inference in Generative Models for High-dimensional Black-box Optimization [video]