Text-Guided 6D Object Pose Rearrangement via Closed-Loop VLM Agents
arXiv, 2026
Without additional training, A VLM agent iteratively refines an object's 6D pose to follow text instructions.
News
-
Mar 2026
I'm joining the Visual Computing Lab as a MS/PhD student.
-
Feb 2026
Our work, PhysGaia, got accepted to CVPR 2026.
Research
PhysGaia: A Physics-aware Benchmark with Multi-Body Interactions for Dynamic Novel View Synthesis
CVPR, 2026
Project Page arXiv Dataset Code
A novel physics-aware dataset specifically designed for Dynamic Novel View Synthesis (DyNVS), encompassing both structured objects and unstructured physical phenomena.