Skip to content
View original post on X: Santiago
Published
46/100

Liber-0 learns robot manipulation from human experience, zero-shot

AILiber-0, a World Action Model from Liber AI, lets robots perform zero-shot manipulation without robot demonstrations collected for each task. Across 60 tasks and 600 trials, it succeeded 63.5% of the time on pick-and-place, 34.0% on other manipulation, and 10.5% on challenging multi-step tasks, for 36% overall.
SantiagoVerified on X
@svpino
View on X

This model helps robots learn to manipulate objects physically.

But there's something really interesting about it:

The model learns skills from humans rather than relying on robot demonstrations, which are much more expensive and time-consuming.

Basically, this model lets robots perform zero-shot tasks without training on robot demonstrations collected specifically for those tasks.

They demonstrated this:

  • 60 tasks and 600 trials
  • 63.5% success on pick-and-place tasks
  • 34.0% success on other manipulation tasks
  • 10.5% success on challenging/long-horizon tasks
LiberAI@liberai_robot
Introducing Liber-0, our World Action Model for zero-shot robot manipulation—including dexterous, multi-step tasks. Across 60 tasks proposed by GPT-6 and 600 trials, Liber-0 achieved 36% overall success without robot demonstrations collected specifically for those tasks: • Pick-and-place: 63.5% • Other manipulation tasks: 34.0% • Challenging multi-step tasks: 10.5% Zero-shot capability extends beyond individual actions to coordinated sequences. Pretrained on 2.2M+ hours of human experience, Liber-0 jointly predicts future observations and actions to turn human experience into robot skills. Technical blog: https://www.liberai.com/blog/liber0
View quoted post on X
Source details

Santiago · x.comPublished · added here

Original title: This model helps robots learn to manipulate objects physically.

Topics

Embodied AIData & trainingModel releasesEvals & benchmarks