Figure reports Helix 2.5 household-task results across 30 unseen homes
Source publication date
Summary
Figure announced Helix 2.5, pretrained on its Index human-behavior dataset, and evaluations in 30 Bay Area homes. Tasks covered room tidying, towel folding and bed making without collecting data or adapting the model in those homes. Figure reported full-task success of 56% for the pretrained model versus 9% for its comparison model.
Editorial view
Editorial view: The experiment tests whether household robots can avoid costly retraining for each home; failure conditions matter alongside success rates.
Evidence and limitations
The results are company-reported, not independently reproduced. They do not establish a consumer service launch or complete automation of household work.