Friday, August 14, 2026
Facts you can rely on·101 entities·4,806 sourced facts
Source trace. Via News points to the documents behind its reporting and shows what we drew from each — so you can check any claim. How we source
News articleStanford AI Lab

Reward Isn't Free: Supervising Robot Learning with Language and Video from the Web

View original at ai.stanford.edu
Reward Isn't Free: Supervising Robot Learning with Language and Video from the Web <p><em>This work was conducted as part of <a href="https://ai.stanford.edu/">SAIL</a> and <a href="https://crfm.stanford.edu/">CRFM</a>.</em></p> <p>Deep learning has enabled improvements in the capabilities of robots on a range of probl…
Opening lines of the source · Stanford AI Lab · short snapshot — read the full document at the original

What we drew from this source

The claims Via News extracted from this document. We point to the source; we don't replace it.

  • LOReL achieves 66% success rate on 5 tasks specified by natural language

    80% confidence
  • Training on human videos provides 20%+ improvement on unseen tasks

    80% confidence
  • LOReL has limited generalization to unseen tasks

    80% confidence
  • DVD shows 20%+ improvement in success rate on unseen environments compared to robot-only training data

    80% confidence

Cited in these Via News reports