01
Egocentric Tool-Use Reasoning
Understanding why tools are selected, how they interact with target objects, and how physical states evolve in first-person video.
Updates
Recent news
Aug 2026
EgoTools
was accepted to
EMNLP 2026 (Main Conference)! 🎉
Feb 2026
C3G
was accepted to CVPR 2026.
Jan–May 2026
Conducted research at
NTU S-Lab (MMLab)
under
Prof. Ziwei Liu
during an exchange semester in Singapore.
Research
Three connected research threads
My projects are connected by one question: how can AI understand the physical consequences of actions in real-world environments?
02
Latent Action & World Models
Learning action-centric representations from video priors and modeling how possible actions transform future world states.
03
3D/4D Scene Understanding
Building geometry-aware scene representations for dynamic visual understanding, reconstruction, and embodied interaction.
Publications
Selected work
1
2
* Equal contribution.
Experience
Research roles
S-Lab (MMLab), NTU
Jan 2026 – Present
Research Intern · Advisor: Prof. Ziwei Liu
- Co-led EgoTools, a 100+ hour multimodal egocentric dataset and 1K-question benchmark for tool-centric reasoning in real-world videos.
- Designed annotation and QA workflows, coordinated data collection and quality control, and contributed to model evaluation, failure analysis, and manuscript writing.
- Continuing follow-up research on egocentric tool use and embodied video reasoning.
Computer Vision Lab (CVLAB), KAIST AI
Jun 2025 – Present
Undergraduate Researcher · Advisor: Prof. Seungryong Kim
- Working on video-prior-based latent action learning, with a focus on data strategy, human/robot video priors, and embodied world modeling.
- Studied feed-forward 3D reconstruction and novel-view synthesis with Gaussian Splatting, VGGT, and NoPoSplat, and contributed analysis tools and experiments to C3G.
Video and Image Computing Lab (VICLAB), KAIST
Dec 2024 – Jun 2025
Research Intern · Advisor: Prof. Mun Churl Kim
- Reviewed modern object detection architectures from RCNN/YOLO families to vision transformers.
- Implemented baselines for few-shot and anomaly detection tasks.
Education
Academic background
- B.S. in Electrical Engineering
- Mar 2021 – Expected Aug 2027
- GPA: 4.10/4.30 · Major GPA: 4.19/4.30
- Exchange Student, School of Electrical and Electronic Engineering
- Jan 2026 – May 2026
Honors
Selected awards
- KAIST Presidential Fellowship (15th) 2025 – Present
- National Science & Engineering Scholarship for Academic Excellence 2025 – Present
- KAIST EE Dean’s List Award Spring 2024 Awarded to the top 3% of EE students
- KAIST Freshman Dean’s List Award Fall 2021 Awarded to the top 3% of first-year students
- Woon Hae Scholarship (12th) 2025
- KAIST Shared Lives Award · Top-ranked Recipient 2026
Beyond Research
Leadership & technical toolkit
Leadership: President of EERun, Google Student Ambassador, member of the Young Engineers Honor Society.
Programming: Python, C/C++, PyTorch, Git, LaTeX.
Research toolkit: video preprocessing, dataset curation, annotation pipelines, model evaluation, and 3D/vision experimentation.