Overview
There has been growing interest in learning robot skills from humans—both of which can be viewed as physical agents that interact with the world. In recent years, computer vision researchers have focused on creating digital twins of humans in virtual environments that behave like real ones, while roboticists have been building physical agents capable of interacting with the real world. We believe that progress in one field can greatly benefit the other. On one hand, virtual humans can be regarded as a special form of robotic agents; on the other hand, robots can learn manipulation and locomotion from human demonstrations, including those performed by simulated humans. Through the proposed workshop, we aim to bring these two fields together to explore common challenges—such as retargeting, embodiment gap, contact modeling, data scarcity, and the role of foundation models, among others. We hope this workshop will serve as a fertile ground for inspiring new research directions and fostering cross-disciplinary collaboration in this space.
Invited Keynote Speakers
Call for Papers
We invite the submission of 4-page extended abstracts (in CVPR format). The workshop papers are non-archival, and we encourage submissions that have already been submitted to or accepted by other venues. We will host a poster session featuring up to 30 posters.
Submission Timeline
| Call for Papers | March 13, 2026 |
| Submission Deadline | April 10, 2026 |
| Acceptance Notification | May 1, 2026 |
| Final Version Due | May 15, 2026 |
Topics of Interest
We welcome papers on (but not limited to):
- Human-object and human-scene interaction reconstruction and synthesis
- Long-horizon interaction understanding and planning
- Imitation learning from human demonstrations
- Real-time human reconstruction and teleoperation
- Foundation models for human and robot agents
- Agent-agnostic representations and embodiment gap
- Data collection and curation for embodied agents
- Contact modeling, motion prediction, and physics-based simulation
- Humanoid learning and whole-body control
