LingBot-VLA 2.0 Architecture and Testing

Scobleizer · x · 2026-07-08

This post shares technical details about LingBot-VLA 2.0: it utilizes a sparse MoE design, achieving higher training efficiency and lower training loss and error under the same active parameter budget.

For bimanual tabletop tasks and long-horizon mobile tasks, the post claims it outperforms LingBot 1.0, Pi 0.5, and GR00T N1.7 in both success rate and task progress.

Related event: Ant Lingbo Open-Sources LingBot-VLA 2.0 for Multi-Robot Generalization(18 posts)→

Original post →

More from Embodied

Embodied channel →