DEVOURED
Helix 2.5
Figure's new Helix 2.5 humanoid model demonstrated zero-shot competency in 30 different homes, performing household chores without specific prior training.
What: Figure robots running the Helix 2.5 control model successfully executed tasks like folding laundry and tidying rooms in real-world residential environments.
Why it matters: The ability to generalize physical actions across diverse, unseen domestic layouts marks a significant step for embodied AI utility.
Deep dive
- Helix 2.5 is a humanoid control model trained on Figure's proprietary Index dataset.
- Robots were deployed in 30 diverse Bay Area homes.
- Execution was zero-shot, meaning no fine-tuning or data collection was performed in the test environments.
- Tasks included room tidying, towel folding, and bed making.
- This release marks a shift from laboratory robotics to generalized real-world capability.
Decoder
- Zero-shot: The ability of a model to perform a task it was not explicitly trained for or tuned on.
- Embodied AI: AI systems, typically robots, that interact with the physical world through sensors and actuators.
Original article
Full article content is not available for inline reading.
Explore all 58 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi