Issue 2026-11-10 · Industry · 模型发布 · 研究

Nvidia unveils GR00T 2, learning robot tasks from video

Nvidia unveils GR00T 2, learning robot tasks from video

Nvidia released GR00T 2, a robot model that transfers human hand movements from video to robotic arms without laborious teleoperation, trained on about 400,000 hours of footage. Success rates vary widely, from 10% to 80% depending on the task.

The Decoder12 d ago
Read original ↗