Scaling Compositional Intelligence for Multi-Agent Robotic Systems

Overview

This workshop brings together researchers across robotics, machine learning, and embodied AI to explore how multi-agent collaboration—grounded in physical embodiment and powered by foundation models—can enable scalable, generalizable, and trustworthy intelligence. It will emphasize robot–robot coordination, compositional intelligence, and distributed planning across heterogeneous agents. Through invited talks, paper sessions, and discussions, the workshop seeks to foster a focused community advancing the next generation of collaborative embodied systems.

The workshop will consist of a combination of invited talks and contributed papers, selected through a rigorous review process. We aim to bring together experts from academia and industry to promote cross-disciplinary collaboration and discussion.

We also plan to provide financial support for students and junior researchers. The money will be used to provide best paper awards and registration support to minority groups and fund other workshop logistics.

Call For Papers

We welcome submissions in the following areas:

Robotic coordination in navigation, manipulation, and assembly using foundation model reasoning.
Coordinating diverse embodiments and bridging simulation with real-world deployment.
Scalable task decomposition and multi-agent collaboration via logic-driven planning.
Hierarchical reasoning, decentralized control, and cerebrum–cerebellum paradigms.
Scalable environments and datasets for diverse multi-agent physical interactions.
Robustness, trustworthiness, generalization, scalability, and collaboration efficiency.
Reinforcement learning, imitation learning approaches for multi-agent coordination.
Human–Agent Communication: collaboration protocols, and inter-agent alignment.
Use cases in embodied robotics, multi-agent autonomy, open-world exploration, assistive systems, medical robotics, and beyond.

Schedule

*Half-Day Workshop (Morning)*

Time	Event
8:30 AM - 8:40 AM	Opening remarks
8:40 AM - 9:10 AM	Invited talk 1 (30 min)
9:10 AM - 9:40 AM	Invited talk 2 (30 min)
9:40 AM - 10:10 AM	Oral presentations (15 min × 2)
10:10 AM - 10:40 AM	Coffee break & Poster session
10:40 AM - 11:10 AM	Invited talk 3 (30 min)
11:10 AM - 11:40 AM	Invited talk 4 (30 min)
11:40 AM - 12:00 PM	Panel discussion & Awards

Invited Speakers

Given the interdisciplinary nature of the topics our workshop aims to cover, we invited keynote speakers with different backgrounds and perspectives on open-world reasoning and decision-making (all confirmed):

Dhruv Shah

Google DeepMind

Xihui Liu

The University of Hong Kong

Shanghang Zhang

Peking University

Wenlong Huang

Stanford University

Organizers

Yiran Qin

The Chinese University of Hong Kong, Shenzhen

Zhenfei Yin

University of Oxford

Yilun Du

Harvard University

Jiayuan Mao

Massachusetts Institute of Technology

Nur Muhammad "Mahi" Shafiullah

New York University

Fangchen Liu

University of California, Berkeley

Li Kang

Shanghai Jiao Tong University

Advisory Board

Josh Tenenbaum

Massachusetts Institute of Technology

Sergey Levine

University of California, Berkeley

Pieter Abbeel

University of California, Berkeley

Mac Schwager

Stanford University

Philip H.S. Torr

University of Oxford

Contact Information

For questions and inquiries, please contact the organizing committee through the OpenReview submission system or reach out to the organizers directly using the email addresses provided above.