Sangwoo Mo
Sangwoo Mo
Assistant Professor
POSTECH IME, affiliated with IDS and AI
Email: {first_name}.{last_name} [at] postech.ac.kr
Office: Science Building 4, Room 319
Tel: +82-54-279-2202
Google Scholar / Github / CV / Twitter
Embedded Files
I am an assistant professor at POSTECH in the Department of Industrial and Management Engineering (IME), affiliated with the Graduate School of Industrial Data Science (IDS) and Artificial Intelligence (AI). Previously, I was a postdoctoral fellow at the University of Michigan, working with Prof. Stella X. Yu, and earned my Ph.D. from KAIST under the supervision of Prof. Jinwoo Shin. I have collaborated with leading research labs including Google, Meta, NVIDIA, Naver, and Kakao, and have been supported by fellowships from Samsung and Qualcomm.
My research goal is to design scalable priors for AI. Even in the era of data-centric foundation models, I believe that well-designed priors can enhance performance, reliability, and interpretability. My previous work has shown that incorporating inductive biases, such as domain-specific structures and statistical assumptions, enables models to handle more complex and diverse tasks. These efforts span diverse tasks in vision, language, and robotics, and have been published in leading AI venues such as NeurIPS, ICLR, and CVPR.
⏵ Research Topics
⏵ Research Topics
1. Structural Priors for Effective Learning
Encode domain-specific structures to improve performance and interpretability
- Visual structure (image):
InstaGAN (ICLR'19), OACon (NeurIPS'21), OAMixer (CVPRW'22), CAST (ICLR'24), SHED (tbd), F-CAST (tbd)
- Temporal structure (video, language):
DIGAN (ICLR'22), HOMER (ICLR'24), Simba (TMLR'25)
- Structured reasoning (gen models):
- Physical structure (robotics):
EIT-NN (IROS'19), EIT-NN-v2 (T-RO'21)
- Relational structure (graph, table):
LoGe (ICMLW'21), DPM-SNC (NeurIPS'23)
2. Statistical Priors for Reliable Learning
Estimate and leverage dataset statistics to improve robustness and generalization
- Robust representation (outlier, etc):
CSI (NeurIPS'20), MASKER (AAAI'21), RoPAWS (ICLR'23)
- Reliable generation (fairness, etc.):
GOLD (NeurIPS'19), FICS (TMLR'23), GFG (tbd)
- Extension to novel domains & contexts:
FreezeD (CVPRW'20), S-CLIP (NeurIPS'23), OAK (CVPR'25), VAP (tbd)
- Analyzing & leveraging foundation models:
B2T (CVPR'24), PDS (ICLR'26), Idis (tbd)
- Efficient model design:
▪︎ 2026/01: Paper on dataset distillation accepted at ICLR'26.
▪︎ 2025/10: Served as an organizer for the SP4V (structural priors for vision) Workshop @ ICCV'25.
▪︎ 2025/09: Started as an assistant professor at POSTECH! 🧑🏫
Appointments
Appointments
Faculty positions
▪︎ POSTECH, Assistant Professor / 2025–current (Korea)
Research positions
▪︎ U Michigan, Postdoctoral Fellow (host: Prof. Stella X. Yu) / 2023–2025 (US)
▪︎ NVIDIA AI, Research Intern (host: Dr. Weili Nie) / 2022 Winter (US, remote)
▪︎ Meta AI, Research Intern (host: Dr. Jong-Chyi Su) / 2022 Summer (US)
▪︎ Naver AI, Collaborator (host: Dr. Jung-Woo Ha) / 2021 Summer (Korea)
▪︎ Google AI, Collaborator (host: Dr. Kihyuk Sohn) / 2020 Winter (US, remote)
▪︎ Kakao Brain, Research Intern (host: Dr. Sungwoong Kim) / 2019 Spring (Korea)
Education
Education
▪︎ KAIST, M.S./Ph.D. in Electrical Engineering, 2023 (advisor: Prof. Jinwoo Shin)
- Ph.D. Thesis: Learning Visual Representations from Uncurated Data
▪︎ POSTECH, B.S. in Mathematics & Industrial Engineering (minor), 2016 - summa cum laude
Services
Services
▪︎ Workshop Committee:
- Structural Priors for Vision (SP4V) @ ICCV'25
- Pixel-level Vision Foundation Models (PixFoundation) @ CVPR'25
- Virtual Humans for Robotics and Autonomous Driving (POETS) @ CVPR'24
▪︎ Conference Reviewer: NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, COLM, AAAI, AISTATS
▪︎ Journal Reviewer: TPAMI, TMLR, TNNLS, IJCV, ML, and others
Google Sites
Report abuse