https://agxsearch.ai/t/bjltea?utm_source=bluesky&utm_medium=social
https://agxsearch.ai/t/bjltea?utm_source=bluesky&utm_medium=social
https://agxsearch.ai/t/g7v23y?utm_source=bluesky&utm_medium=social
https://agxsearch.ai/t/g7v23y?utm_source=bluesky&utm_medium=social
Runway has released the weights and code of Praxis-1, a robot model trained on video, under Apache 2.0. It tops World Action Models on Nvidia's RoboLab-120 at 46.2 percent, but the tests are Runway's own.
Runway has released the weights and code of Praxis-1, a robot model trained on video, under Apache 2.0. It tops World Action Models on Nvidia's RoboLab-120 at 46.2 percent, but the tests are Runway's own.
https://nexo24.cz/0V
https://nexo24.cz/0V
runway.com/news/compan...
runway.com/news/compan...
#AI #Runway #ロボット
runway.com/news/company...
#AI #Runway #ロボット
runway.com/news/company...
->StartupHub.ai | More on "Video models for robot control" at BigEarthData.ai
📊 46.2% success on Nvidia's RoboLab-120, with LoRA fine-tuning and Jetson Thor support.
🔗 https://thelatentdaily.com/news/runway-praxis-1-open-weights-robot-world-action-model/
#Robotics #OpenWeights #Runway #WorldModels
📊 46.2% success on Nvidia's RoboLab-120, with LoRA fine-tuning and Jetson Thor support.
🔗 https://thelatentdaily.com/news/runway-praxis-1-open-weights-robot-world-action-model/
#Robotics #OpenWeights #Runway #WorldModels
Praxis-1, a 27B world action model, is out with open weights under Apache 2.0. It topped RoboLab-120 at 46.2% and already runs on a humanoid, an arm and a warehouse robot.
Source: The Robot Report
Image: Runway
#Robotics #PhysicalAI
Praxis-1, a 27B world action model, is out with open weights under Apache 2.0. It topped RoboLab-120 at 46.2% and already runs on a humanoid, an arm and a warehouse robot.
Source: The Robot Report
Image: Runway
#Robotics #PhysicalAI
ifac-control.org/publications...
#AutomaticControl #ControlEngineering #IFAC2026 #GirlsInControl
ifac-control.org/publications...
#AutomaticControl #ControlEngineering #IFAC2026 #GirlsInControl
Praxis-1,270 亿参数的世界动作模型,一边预测接下来 1.6 秒的画面、一边把画面翻译成机器人动作,权重和微调代码全按 Apache 2.0 开源,商用都行。
最能看的是实测数字:
RoboLab-120 榜单 1200 个回合拿下 46.2%,排第一,FLUX 3 Action 42.9%、Cosmos3 Nano 36.8%、π0.5 28.0%。
Praxis-1,270 亿参数的世界动作模型,一边预测接下来 1.6 秒的画面、一边把画面翻译成机器人动作,权重和微调代码全按 Apache 2.0 开源,商用都行。
最能看的是实测数字:
RoboLab-120 榜单 1200 个回合拿下 46.2%,排第一,FLUX 3 Action 42.9%、Cosmos3 Nano 36.8%、π0.5 28.0%。
It's a 27B model that predicts video and robot actions together.
Runway says it scored 46.2% on Nvidia's RoboLab-120, its own number.
aiparabellum.com/runway-praxi...
It's a 27B model that predicts video and robot actions together.
Runway says it scored 46.2% on Nvidia's RoboLab-120, its own number.
aiparabellum.com/runway-praxi...
Source: The Decoder
Source: The Decoder
Runway's Praxis-1 solved 555 of 1200 Robolab-120 scenarios, just under half. An open-weight model near 50% on a benchmark hands outside labs a baseline they can now measure their own robot control against.
Runway's Praxis-1 solved 555 of 1200 Robolab-120 scenarios, just under half. An open-weight model near 50% on a benchmark hands outside labs a baseline they can now measure their own robot control against.