Definition
Plain language
A benchmark of simulated robot manipulation tasks used to test learning systems.
As stated in the literature
A robot-manipulation benchmark suite; the LIBERO-PRO variant is a harder held-out test on which end-to-end vision-language-action policies score near zero and skill-library agents reach the mid-forties percent.
Also called: LIBERO-PRO
Why it matters: It provides a common testbed for comparing robot-learning methods, and its harder variant exposes how brittle end-to-end policies can be.
For example, a policy might be asked to pick up a bowl and place it on a plate across a range of simulated kitchen scenes.
Heard on the show
“The headline: on a benchmark called LIBERO-PRO, the no-play baseline succeeds about twenty-three percent of the time.”Episode 161 — A Robot That Plays Before You Give It a Job, And Why That Beats Retrying