Embodied Spatial Reasoning
ICCV 2025 Workshop
Oct 20th (morning), 2025, Honolulu, Hawai'i
Embedded Files
As intelligent agents increasingly move from static perception to embodied interaction, the need for spatial understanding has become paramount. Unlike traditional vision systems that passively observe, embodied agents must perceive their environments in 3D, reason about spatial relations, and act upon the world. This shift calls for integrated approaches that connect visual recognition, spatial reasoning, and action planning under real-world constraints.
Recent breakthroughs in large language models (GPTs), embodied AI platforms, 3D neural representations, and robot perception have significantly advanced the field. Nevertheless, robust and generalizable embodied spatial reasoning remains a fundamental challenge, particularly in complex, dynamic, and unstructured environments.
The 1st Embodied Spatial Reasoning Workshop at ICCV 2025 brings together researchers interested in embodied AI, embodied spatial reasoning, embodied 3D world models, and robot spatial reasoning.
Invited
Speakers
Nvidia
University of California, San Diego
Google Deepmind
Meta FAIR
Schedule
09:00 - 09:10 Opening Remarks
09:10 - 09:50 Sifei Liu: Scaling Spatial Intelligence: From Region Reasoning to Embodied Vision-Language Agents
09:50 - 10:30 Xiaolong Wang: Manipulation and Reasoning Beyond Teleoperation
10:30 - 10:40 Break / Buffer Time
10:40 - 11:20 Fei Xia: Gemini Robotics 1.5: Pushing the Frontier of Generalist Robots with Advanced Embodied Reasoning, Thinking and Motion Transfer
11:20 - 12:00 Kate Saenko: Blind Spots in Multimodal AI Models
Organizers
Johns Hopkins University
Johns Hopkins University
University of Freiburg
University of California, Davis
Johns Hopkins University
Johns Hopkins University
Max-Planck Institute for Informatics
DEVCOM Research Laboratory
Johns Hopkins University
Johns Hopkins University
Contact person: Jieneng Chen (jchen293@jh.edu)
Page updated
Google Sites
Report
abuse