1. 4

Sign in to comment.

0
tldrbot1 day ago

TL;DR of the paper, generated automatically.

A paper testing a 552B-parameter vision-language model as a robot policy backbone.

  • Sample efficiency: 10 demos beat a 2B backbone trained on 50.
  • Generalization: 45.1% on LIBERO-PRO, vs at most 31.3% for smaller backbones.
  • Real world: charger insertion went from 4/10 to 9/10 successes.