Research
Scaling simulation, benchmarks and data, from manipulation to mobile manipulation and dexterity. Now extending to long-horizon tasks and hierarchical systems. I'm especially interested in how intent and goals are represented across levels to produce useful real-world motion.
Publications
MolmoSpaces: A large-scale open ecosystem for robot navigation and manipulation
Y. Kim*, W. Pumacay*, O. Rayyan*, M. Argus*, et al. RSS 2026.
230k+ indoor environments, 130k objects, 42M grasps; the benchmark tracks real-world results (R = 0.96).
MolmoBot: Large-scale simulation enables zero-shot manipulation 🏆
A. Deshpande, …, Y. Kim, et al. CoRL 2026.
Oral presentation. Best Paper Award at three ICRA 2026 workshops: Synthetic Data for Robot Learning, VLA Pipeline and Beyond Teleoperation.
Contact-anchored policies: contact conditioning creates strong robot utility models
Z. J. Cui, O. Rayyan, …, Y. Kim, et al. RSS 2026.
Harmonic mobile manipulation 🏆
R. Yang, Y. Kim, R. Hendrix, A. Kembhavi, X. Wang, K. Ehsani. IROS 2024.
Best Paper Award on Mobile Manipulation
SPOC: Imitating shortest paths in simulation enables effective navigation and manipulation in the real world
K. Ehsani, T. Gupta, …, Y. Kim, et al. CVPR 2024.
Workshops
SemRob: 3rd Workshop on Semantic Reasoning and Goal Understanding in Robotics
Co-organizer
RSS 2026, Sydney
2026ERA: Embodied Reasoning in Action, Workshop and Challenge on Embodied Reasoning for Robotic Manipulation
Co-organizer
CVPR 2026, Denver
2025Embodied Co-Creation: Robotic Tools That Make and Perform Arts
Co-organizer
Humanoids 2025, Seoul