The Causal Architecture of LingBot-VA 2.0
rohanpaul_ai · x · 2026-07-14
This post breaks down the core architecture of LingBot-VA 2.0:
- The control policy is trained in a causal manner from the ground up, rather than retrofitting a bidirectional video model into a controller.
- The video stream utilizes 128 routed experts, top-8 routing, and 1 shared expert. The total parameter count is around 13B, but only 1.9B are activated per token.
- The action stream remains dense, aiming to balance overall capability with control latency.
Related event: LingBot-VA 2.0: A Control-Native Foundation Model for Robotics(9 posts)→
More from Embodied
- StackChan robot performs an AI-generated skit, orchestrated by Hermes Agent via JSON script — alexcovo_eth · 2026-10-27
- DONIC hits 86.79% on bimanual dexterous tracking, crushing prior SOTA's 9.43% — liuziwei7 · 2026-10-27
- Bedrock Robotics raises $600M as its fully autonomous excavators do paid work on live job sites — Sethwinterroth · 2026-10-27
- AW-0 hits 80.8% on RoboTwin tasks, beating strongest baseline by 16 points — trevordarrell · 2026-10-27
- AW-0's code stream fixes world model prediction: picks mango where baseline grabs apple — trevordarrell · 2026-10-27
- AW-0 runs at 230ms on a single RTX 5090, 2.4x faster than autoregressive code planning — trevordarrell · 2026-10-27