Skip to content
Trending storyDeveloping

Runway details Praxis-1 robotics model and reports benchmark results

3 articles1 sourcesince Nov 11Last article

Overview

AISummary of 3 articles

Runway says its Praxis-1 model, a 27B-parameter mixture-of-experts diffusion transformer that predicts video and actions together, scores 80.0% on SimplerEnv after fine-tuning on Google's RT-1 robot, a robot it never saw in training.

Runway says the original RT-1 policy scores 52.4% on the same benchmark. The company also says Praxis-1 succeeded in all 15 simulated trials on Noble Machines' robot after learning from under 30 minutes of teleoperated demonstrations.

Runway argues general video is the most scalable source of robotics data, since a model trained to predict video learns how objects behave and how tasks unfold. The company says its technical report covers the model's architecture, training stages and results.

Written by AI from the articles below · updated Nov 11, 10:21 AM ET

Check the sources:

Developments

3 developments

  1. Nov 11, 10:04 AM ET · 1 article
    Runway reports Praxis-1 robot task-learning results
    Runway: Runway's Praxis-1 succeeds in all 15 simulated robot trials after 30 minutes of demos
  2. Nov 11, 10:04 AM ET · 1 article
    Runway reports Praxis-1 robotics evaluation results
    Runway: Runway's Praxis-1 hits 80.0% on SimplerEnv after RT-1 fine-tuning
  3. Nov 11, 10:04 AM ET · 1 article
    Runway details Praxis-1 architecture, training and results
    Runway: Runway details Praxis-1, a 27B video and action model, in a technical report

Article timeline

The articles in this story. Times are ET.

Nov 11
  1. RunwayOfficial
    Runway's Praxis-1 succeeds in all 15 simulated robot trials after 30 minutes of demos

    AIRunway says its Praxis-1 model succeeded in all 15 simulated trials on Noble Machines' robot after learning from under 30 minutes of teleoperated demonstrations.

    Video from @runwayml's post
  2. RunwayOfficial
    Runway's Praxis-1 hits 80.0% on SimplerEnv after RT-1 fine-tuning

    AIRunway says its Praxis-1 model, fine-tuned on Google's RT-1 robot, a robot it never saw in training, reaches 80.0% on SimplerEnv. The original RT-1 policy scores 52.4% on the same benchmark. Runway argues general video is the most scalable source of robotics data, since a model trained to predict video learns how objects behave and how tasks unfold.

    Image from @runwayml's post
  3. RunwayOfficialPick
    Runway details Praxis-1, a 27B video and action model, in a technical report

    AIRunway says Praxis-1 is a 27B-parameter mixture-of-experts diffusion transformer that predicts video and actions together. The company says its technical report covers the model's architecture, training stages and results.

Heat Trend

Current comparable heat
6
Peak comparable heat
10Nov 11, 11:00 AM ET
Change in past 24 hours
–

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.