Yujin Bae
me and my sister 🐠
Hi! I am Yujin Bae, a second-year M.S. student at KAIST, where I am advised by Prof. Kuk-Jin Yoon in the Visual Intelligence Lab.
My long-term goal is to build proactive assistive AI for embodied settings such as smart glasses and robotic hands. Toward this end, I focus on multimodal understanding of how humans interact with their surroundings, and I aim to extend this understanding to robot learning as a next step.
Concretely, my interests span:
- Anticipation — forecasting human motion and action semantics from egocentric video and sparse wearable sensors
- Understanding — efficient egocentric video understanding with multimodal LLMs, guided by human behavior
- Transfer — learning dexterous manipulation policies from human video
Before joining KAIST, I received my B.S. at Ewha Womans University (ranked 1st in major!), and spent a semester as an exchange student at the University of Vermont.
Feel free to reach out via yujinbae@kaist.ac.kr!
news
| Sep 25, 2026 | Our paper “SRA: Spatial Reasoning Adapter via Evolving Social Interaction Graphs for Trajectory Prediction” was accepted to NeurIPS 2026! |
|---|---|
| Sep 08, 2026 | Presented Ego3DLM at the 3D Human Understanding and On-device Embodied World Models (ODEWM) workshops at ECCV 2026 in Malmö! |
| Jun 20, 2026 | Our paper “Ego-human Motion Prediction with 3D-aware LLM” was accepted to ECCV 2026! |
| Mar 01, 2025 | Started my M.S. at KAIST, advised by Prof. Kuk-Jin Yoon. |
publications
- NeurIPS 2026
SRA: Spatial Reasoning Adapter via Evolving Social Interaction Graphs for Trajectory PredictionIn Advances in Neural Information Processing Systems (NeurIPS), 2026(* equal contribution) - ECCV 2026
Ego-human Motion Prediction with 3D-aware LLMIn European Conference on Computer Vision (ECCV), 2026(* equal contribution)
Also presented at the 3D Human Understanding Workshop and the On-device Embodied World Models (ODEWM) Workshop, ECCV 2026
- Under ReviewReUP: Retrieval-augmented Understanding and Prediction of Human Behaviour from Sparse Egocentric Signals2026