Cordelia Schmid
Google / Inria
ECCV 2026 Workshop
Embodied Agent
and Dialog
Embodied agents that ask, reason, and act safely.
Proactive dialog for robust, collaborative, and safe embodied intelligence.
Program
09:00 – 09:10Welcome and Opening Remarks
09:10 – 09:40Invited Talk 1: Qi Wu "Embodied Vision-and-Language Navigation - A Universal View"
09:40 – 10:10Invited Talk 2: Ranjay Krishna "Action Reasoning Models"
10:10 – 11:10Poster Session & Coffee Break
11:10 – 11:30Industrial Talk: Wongyun Yu (RLWRLD) "Toward Human-like Dexterous Manipulation"
11:30 – 12:00Invited Talk 3: Cordelia Schmid "Video-guided policies for robotic manipulation"
12:00 – 12:30Invited Talk 4: Michael S. Ryoo "Motion World Model"
12:30 – 12:50Challenge Results and Winner Presentations
12:50 – 13:00Closing Remarks
Invited Speakers
Ranjay Krishna
University of Washington / AI2
Qi Wu
The University of Adelaide
Michael S. Ryoo
SUNY / DeepMind
About Embodied Agent and Dialog (EAD) Workshop
The Embodied Agent and Dialog (EAD) Workshop explores dialog in Embodied AI. Embodied agents act in the real world and face physical consequences from ambiguous or underspecified instructions. To ensure both efficiency and safety, agents must proactively clarify uncertainty, request additional information through dialog, and understand multi-turn instructions before acting. The workshop aims to unites researchers from computer vision, NLP, robotics, and human–AI interaction to examine how dialog underpins intelligent, cooperative embodied systems.
Dialog-Enabled Agents
Agents that use dialog to manage uncertainty and align actions with human intent.
Embodied Reasoning
Natural-language explanations, action rationales, and interactive decision making.
Adaptive Systems
Dynamic behavior and dialog strategies conditioned on environment and user feedback.
Tasks & Benchmarks
Evaluation of task success, dialog quality, situational understanding, and human-likeness.
Call for Papers
Topics of Interest
Embodied dialog, interactive instruction following, embodied reasoning, uncertainty-aware agents, adaptive human-agent collaboration, and benchmarks for dialog-based embodied AI.
Submission Detail
Non-archival. All submissions are reviewed double-blind by at least two reviewers.
Challenges
The 1st DialNav Challenge
Interactive instruction following with high-level path ambiguity.
Visit Website →The 1st CoIN Challenge
Collaborative instance navigation with targeted disambiguation.
Visit Website →Organizers
Paul Hongsuck Seo
Korea University
Associate Professor.Yiming Wang
FBK
Senior Researcher.Jonghyun Choi
Seoul National University
Associate Professor.Aishwarya Agrawal
U. Montreal / Mila / Google DeepMind
Assistant Professor; CIFAR AI Chair; Research Scientist.Shizhe Chen
INRIA Paris
Researcher.Sangwon Jung
Trillion Labs
Technical Staff.Leekyeung Han
Korea University
Ph.D. Student.Edoardo Zorzi
U. Verona / Sapienza
Ph.D Student.Hyunji Min
Korea University
Ph.D. Student.Gengze Zhou
University of Adelaide
Ph.D. Candidate.Jinseong Jeong
Korea University
M.S. Student.Stay Updated
Sponsors
We gratefully acknowledge the generous support of our sponsors.
