Hello, I'm Jahyeok Koo

Ph.D. Student @ CVLAB (KAIST AI)

I am a Ph.D. student at KAIST AI advised by Seungryong Kim. My research interested is understanding the 3D and 4D structure of the real world, focusing on Point Tracking and 3D/4D Reconstruction. Recently, I have been interested in grounding world understanding in Language, exploring 3D/4D-aware VLMs, Unified Multimodal Models, and World Models.

Scholar Github X Linkedin

News

Sep 2026: 🎉 Two papers accepted to NeurIPS 2026: T2Mo (controllable dynamic 3D shape generation) and TrackCraft3R (dense 3D tracking with video diffusion transformers).
June 2026: Released two new preprints on arXiv: T2Mo (arXiv) and MVTrack4Gen (arXiv) !!
May 2026: Released our new preprint TrackCraft3R on arXiv — code is also available !!
Feb 2026: Our two papers, MV-TAP and AnthroTAP, have been accepted @ CVPR 2026 !!
Dec 2025: Released our new preprint MV-TAP on arXiv — code is also available !!
Sep 2025: I started my Ph.D. at KAIST AI CVLAB !
July 2025: Released our new preprint AnthroTAP on arXiv — code and data are also available !!

Publications

NEW !
Can Learning to Answer Spatial Questions Improve Novel-View Generation?

Jahyeok Koo, Hyeonseo Yu, Seungho Jang, Jaewoo Jung, Jiwon Kang, Seokju Cho, Seungryong Kim

Preprint, 2026 Project Page Code

MVTrack4Gen: Multi-View Point Tracking as Geometric Supervision for 4D Video Generation

JoungBin Lee, Jaewoo Jung, Jongmin Lee, Tongmin Kim, Hyunsung Kim, Takuya Narihira, Kazumi Fukuda, Jahyeok Koo, Jisang Han, Yuki Mitsufuji, Seungryong Kim

arXiv 2026 Project Page PDF arXiv