Figure AI Unveils Helix 2.5: Zero-Shot Generalization in 30 Unseen Homes

Industry Brief is the latest model from Figure AI which was built to answer a harder question: can a humanoid enter a home it has never seen and immediately get to work, with its whole body, on its own? Pretrained Helix 2.5 n Figure's global-scale dataset of human behavior., from that single foundation model, it produces three behaviors: tidying living rooms, folding towels, and making beds. It is then deployed in unseen homes. The News Figure AI has announced Helix 2.5, its most advanced neural network for humanoid robots, demonstrating unprecedented 'zero-shot' generalization capabilities in 30 previously unseen Bay Area homes. The robots successfully performed complex whole-body tasks like tidying living rooms, folding towels, and making beds without any prior data collection, fine-tuning, or adaptation in those specific environments or for the objects manipulated. What It Is Helix 2.5 is a humanoid neural network that was pretrained on Figure's global-scale dataset of human behavior, called Index. This single foundation model was then adapted to perform three distinct long-horizon behaviors. Unlike previous systems, such as Helix 02, which required specific data collection for deployment environments, Helix 2.5 exhibits independence from such environmental dependencies. The testing involved Figure 03 humanoid robots operating in real homes using existing furniture and objects, with strict pass/fail grading and no human intervention for safety marking a failed rollout.…