NVIDIA and KAIST Unveil ACCRUE for Continual Real-Robot RL

NVIDIA and KAIST introduced ACCRUE, a continual off-policy RL framework for post-training robot foundation models. It learns new tasks on real robots with just 4-12 minutes of interaction, reaching 95% success and cutting training time by up to 68.5%.

2026-10-15 ~ 2026-10-15 · 2 related posts

• NVIDIA's ACCRUE: continual off-policy RL cuts real-robot training time by up to 68.5% — nvidia · 2026-10-15
• ACCRUE: continual off-policy RL lets robots learn new tasks in 4-12 minutes with 98.3% skill retention — coallaoh · 2026-10-15