Tingting Liao
廖婷婷
Research
I work on world models for embodied AI — real-time, action-conditioned video generation, and using generated worlds as data engines and evaluators for robot policies.
I came to this from generative video and 3D, building controllable models of humans, objects and scenes, which is where my interest in physical consistency and controllability comes from.
World Models · Video Generation · Embodied AI
News
- Started a research internship at the Institute of Foundation Models, MBZUAI, on real-time interactive world models
- Character Mixing for Video Generation released on arXiv
- Started a research internship at Adobe Research
- SOAP accepted to SIGGRAPH 2025
- TADA! and TeCH accepted to 3DV 2024
Experience
University of Technology Sydney
Visiting Student
2018.12 — 2019.03
Xiaohongshu
Research Intern
2022.02 — 2023.08
Westlake University
Visiting Student
2025.02 — 2025.04
Adobe Research
Research Intern
2025.05 — 2025.08
IFM, MBZUAI
Research Intern
2026.01 — Present
Education
PhD
Mohamed bin Zayed University of Artificial Intelligence
MSc
Institute of Automation, Chinese Academy of Sciences
BEng
Wuhan Polytechnic University
Publications
Open Source
- Dream2DGSImage-to-3D with 2D Gaussian splatting
- MirrorGSRe-implementation — mirror-aware Gaussian splatting
- DreamScene360Re-implementation — text to 360° 3D scenes