1st Workshop on Multimodal Spatial Intelligence

October 20 (1 pm - 5:30 pm), 2025

Room 306 A | Zoom Link (code: farmcans)

About the Workshop

Our multi-modal spatial intelligence workshop aims to bring together researchers from computer vision, robotics, graphics, and NLP for a half-day of discussion and talks at the intersection of visual understanding, multimodal learning, and embodied AI. Our focus is on placing multi-modal large language models (MLLMs) at the core of spatial intelligence-- exploring how they can learn, interpret, and act on spatial information from images, videos, and 3D data.

Keywords:

Spatial Reasoning Multimodal Large Language Model World Models Embodied AI 3D Understanding

Invited Speakers

Saining Xie

NYU / Google DeepMind

Manling Li

Northwestern University

Ranjay Krishna

UW / AI2

Yue Wang

USC / NVIDIA

Qianqian Wang

UC Berkeley

Program

13:00 – 13:10	Welcome & Introduction
13:10 – 13:50	Generate Robotic Data with Spatial Intelligence Yue Wang (USC / NVIDIA)
13:50 – 14:30	Towards Spatial Supersensing Saining Xie (NYU / Google DeepMind)
14:30 – 15:10	Why is Spatial Understanding Hard for VLMs? Manling Li (Northwestern University)
15:10 – 15:30	☕ Coffee Break & Social
15:30 – 16:10	Visual Reasoning Will Be Bigger Than Language Reasoning Ranjay Krishna (UW / AI2)
16:10 – 16:50	On Latent Abilities Underlying Spatial Intelligence Qianqian Wang (UC Berkeley)
16:50 – 17:20	Panel Discussion: Future of Multimodal Spatial Intelligence Leo Guibas (Stanford / Google DeepMind) Manling Li (Northwestern University) Ranjay Krishna (UW / AI2) Yue Wang (USC / NVIDIA) Fei Xia (Google DeepMind) Modulator: Songyou Peng
17:20 – 17:30	Concluding Remarks