Humanoid robot executes chores across dozens of novel environments
Robotics company Figure reported that its Helix 2.5 model successfully completed full body housework across 30 previously unseen home environments. The system performed physical tasks zero shot without requiring pre recorded data for those specific locations.
The Blend
Robotics developer Figure announced that its latest humanoid model, Helix 2.5, performed household chores inside 30 previously unseen residences without prior local training. The company reported that the machine tackled tasks like making beds, folding laundry, and tidying clutter without needing pretraining maps or tailored adjustments for each distinct space.
For ordinary consumers, this signals a shift toward helper robots that can function right out of the box rather than requiring tedious manual setup. Figure claimed that pretraining the robot on a large dataset of human actions boosted its full task success rate from 9 percent to 56 percent across the test locations.
Despite these gains, a success rate around 56 percent highlights that these machines still misstep frequently during complex routines. It remains uncertain how the robot handles sudden disruptions, such as pets or moving occupants, in live household environments. More fundamentally, relying heavily on pretraining raises the question of whether software can ever fully anticipate real world physical friction without undergoing millions of hours of direct physical trial and error.
Written independently by AI News Smoothie from the reporting listed below. Facts belong to the original publishers. Follow the links for their full coverage.
Ingredients
- Helix 2.5: Zero-Shot 30-Home Generalization
Figure demonstrated that pretraining humanoid control models on human activity data increased their chore success rate in unfamiliar homes from 9 percent to 56 percent.