What it is
Egocentric embodied reasoning dataset with step-level human rationales plus a closed-loop benchmark across unseen simulated environments. 3,421 human demos over 55 UE environments plus a 686-task benchmark in 7 unseen environments; 48,264 egocentric steps with verified first-person rationales, actions, poses, and door/object state trajectories.
Key results
- 3,421 dataset tasks + 686 benchmark tasks
- 48,264 rationale-labeled steps; 10,552 benchmark steps
- 62 UE environments total; frames are 640x480; CC-BY-NC-4.0
Useful for training and evaluating embodied agents that must act from egocentric vision and language in closed loop.