Figure's Helix 2.5 hit 56% success in 30 unseen homes, vs 9% from scratch
CyberRobooo · x · 2026-09-18
Figure's Helix 2.5 shows how human data helps robots generalize:
- F.03 humanoids did household tasks in 30 rented Bay Area homes they had never seen, with no new data collected there.
- With large-scale human behavior pretraining, success rate was 56% vs just 9% for a model trained from zero shot — implying the earlier Helix 02 demo's real success rate was only 10%.
- Figure found 8× more human pretraining data kept reducing action-prediction loss, suggesting robot capability may scale with human experience much like foundation models.
The takeaway: robot capability may hinge on pretraining data scale (tens to hundreds of millions of hours), and the golden age of human behavior data collection is here.
More from Embodied
- Toyota to Deploy 400,000 ELEY Humanoid Robots in Factories, $6.4B/Year — CyberRobooo · 2026-09-18
- NHTSA developing federal ADS performance standard as US preps for larger-scale Robotaxi rollouts — Scobleizer · 2026-09-18
- CoRL 2026 SPIN workshop pits robots against a child in live dexterity challenge — berkeley_ai · 2026-09-18
- Waymo Lands in Singapore as Newest Autonomous Vehicle Operator — bytebot · 2026-09-18
- Muse agent installs PyTorch and depth model on its own VM to drive a robot — armand_ruiz · 2026-09-18
- Two Agentic Robots Clean a Room Together, Coordinating by Voice — chris_j_paxton · 2026-09-18