ICML2026
Taeseong Yoon
Ph.D. Candidate in Industrial & Systems Engineering, KAIST
I am a Ph.D. candidate at KAIST, advised by Professor Heeyoung Kim in the Engineering Data Science Lab. My research focuses on uncertainty quantification (UQ) for deep learning, with an emphasis on scalable and theoretically grounded methods for reliable prediction, including evidential deep learning, efficient Bayesian deep learning, conformal prediction, and out-of-distribution detection. I am also interested in statistical machine learning for structured and industrial data, including time-series analysis, tabular foundation models, causal inference, and robust learning under distribution shift.
My current work focuses on post-hoc UQ for foundation models and, more broadly, on UQ for agentic AI systems.
Open to research discussions and collaborations. Feel free to reach out.
Featured publications
All publicationsPublications
2026
CIKM2026
Time series
GT-PSSM: Unified Probabilistic Framework for Stochastic Dynamics Modeling and Dependency Learning in Multivariate Time Series Anomaly Detection
35th ACM International Conference on Information and Knowledge Management (CIKM), Rome, Italy, 2026IEEE TASE2026
Time series
Knowledge-Assisted Multi-Graph Dependency Learning for Multivariate Time Series Anomaly Detection in Multi-Stage Industrial Processes
IEEE Transactions on Automation Science and Engineering, 20262025
NeurIPS2025
Uncertainty quantification
Uncertainty Estimation by Flexible Evidential Deep Learning
Advances in Neural Information Processing Systems 38 (NeurIPS), 20252024
ICML2024
Under review
ICLR2027
Uncertainty quantificationUnder review
Separating Prediction from Epistemic Uncertainty in Evidential Deep Learning
Submitted to ICLR 2027ICLR2027
Imbalanced learningUnder review
Strong Views for Representation, Weak Views for Classification: Rethinking Augmentation in Class-Imbalanced Semi-Supervised Learning
Submitted to ICLR 2027AISTATS2027
Causal inferenceUnder review
Causal Role-Aware Variable Selection for Continuous Treatment Effect Estimation
Submitted to AISTATS 2027News
NewPreprint of GT-PSSM (CIKM 2026) is now on arXiv.
NewCourtroom Analogy is published in the ICML 2026 proceedings (PMLR 306).
Our paper GT-PSSM was accepted at CIKM 2026.
Our paper Knowledge-Assisted Multi-Graph Dependency Learning was accepted for publication in IEEE TASE.
Gave an invited talk on Courtroom Analogy at KDMS 2026.
Presented Courtroom Analogy at ICML 2026 in Seoul.
Our paper Courtroom Analogy was accepted at ICML 2026.
Named a Gold Reviewer for ICML 2026.
Presented F-EDL at NeurIPS 2025.
Gave an invited talk on F-EDL at the 2025 INFORMS Annual Meeting.
Gave an invited seminar on uncertainty quantification at POSCO Labs.
Presented DAEDL at the Qualcomm Innovation Fellowship Korea finalist session.
Received the Best Poster Award (3rd Prize) at Samsung AI Forum 2024.
Presented DAEDL at ICML 2024.
Research
My work develops efficient and faithful uncertainty quantification for deep learning, and applies the same concern for reliability to structured and industrial data.
Uncertainty quantification
Evidential deep learning, efficient Bayesian inference, and second-order uncertainty; post-hoc UQ for foundation models and UQ for agentic AI systems.
Time-series and tabular learning
Multivariate time-series anomaly detection for manufacturing, tabular foundation models, and conformal prediction.
Causal and robust learning
Causal inference and learning under class imbalance, incomplete data, and limited supervision.
Talks & presentations
Courtroom Analogy: New Perspective on Uncertainty-Aware Classification
- KDMS 2026, Invited Talk, Chungnam, South Korea (Aug 2026)
- ICML 2026, Poster Session, Seoul, South Korea (Jul 2026)
Uncertainty Estimation by Flexible Evidential Deep Learning
- NeurIPS 2025, Poster Session, San Diego, CA (Dec 2025)
- KIIE Annual Fall Conference, Daejeon, South Korea (Nov 2025)
- INFORMS Annual Meeting, Invited Session, Atlanta, GA (Oct 2025)
Uncertainty Quantification Methods for Deep Learning
- Invited Seminar, POSCO Labs (Jul 2025)
Uncertainty Estimation by Density-Aware Evidential Deep Learning
- KIIE Annual Spring Conference, Jeju, South Korea (Jun 2025)
- Qualcomm Innovation Fellowship Korea, Finalist Presentation (Dec 2024)
- Samsung AI Forum, Invited Poster Showcase (Oct 2024)
- ICML 2024, Poster Session, Vienna, Austria (Jul 2024)
Out-of-Distribution Detection Using Dirichlet-Based Uncertainty Methods
- KIIE Annual Fall Conference, Ulsan, South Korea (Nov 2023)
CV
Education
- Ph.D. Candidate, Industrial & Systems Engineering, KAIST — Advisor: Prof. Heeyoung Kim
- B.S., Industrial & Systems Engineering, KAIST
Honors & awards
- Gold Reviewer, ICML 2026
- Next-Generation Engineering Researcher, IPSEK
- Best Poster Award (3rd Prize), Samsung AI Forum
- Finalist, Qualcomm Innovation Fellowship Korea
- Dean's List, Dept. of Industrial & Systems Engineering, KAIST
Academic service
- Reviewer: ICML 2026 (Gold Reviewer), NeurIPS 2025–2026
- Reviewer: TMLR, IISE Transactions, Computers & Industrial Engineering, Medical Image Analysis
Research & industry experience
- Research Intern, Samsung Advanced Institute of Technology (SAIT) — Machine Learning Lab, AI Research CenterUncertainty-aware anomaly detection for semiconductor quality inspection.
- Undergraduate Research Intern, Industrial Statistics Lab, KAISTBayesian nonparametric modeling with Dirichlet process Gaussian mixture models.
- Intern, SK Hynix — QA Team, Mobile DRAM EngineeringBank-address pattern analysis for fault detection in mobile DRAM quality data.