ResearchResearch explainer
Figure's Helix 2.5 did chores in 30 homes it had never seen, and finished about half
A model pretrained on human video let Figure 03 tidy, fold towels and make beds in unfamiliar homes at 56% full-task success, up from 9%.
Press releases decoded, rumors with a confidence rating, research in plain words and case studies grounded in what robots have actually done.
A model pretrained on human video let Figure 03 tidy, fold towels and make beds in unfamiliar homes at 56% full-task success, up from 9%.
Skild AI says its S1 model picks up unseen tasks like repotting a plant from a single first-person video, with no retraining.
Google DeepMind's new robot models let Apptronik's Apollo 2 walk, crouch and grasp, with floor pickups succeeding 45.7% of the time.