signal极客公园zh source2026-09-21
Figure AI 宣称找到了机器人版 scaling law，同行却说它其实根本不会泛化
Figure AI's Helix 2.5 achieved 56% full-task success (237/420 trials) across 30 unfamiliar SF-area homes with zero-shot generalization - no training on test environments - covering bed-making (67%), towel folding (62%), and toy cleanup (40%), up from 9% for a control model without its Index human-behavior pre-training. The company claims a robot scaling law based on data doubling reducing action-prediction error to 0.54% deviation, backed by a $3.5B Nscale compute deal for up to 100,000 NVIDIA Vera Rubin GPUs, but competitors counter that 183 failures prove the imitation-learning system still lacks real generalization and reliability for home use.
- why now
- Figure AI新机器人Helix 2.5发布，泛化能力遭同行质疑。
- topic
- AI Tech & New Models
- source
- 极客公园
#Figure AI#scaling law#Helix 2.5
score
score 8 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source