turbovla

Definition

A model name in the vision-language-action (VLA) family, with the prefix signaling inference speed as its selling point. VLA models take an image plus a natural-language instruction and emit robot actions directly, and the practical problem is that a large backbone cannot keep up with a real control loop. Work under names like this attacks latency through distillation, parallel decoding, or action chunking, since on a robot a policy that is accurate but late is unusable.

Explainers using this term

• Paper Walkthrough: TurboVLA — Kick the LLM Out of the Loop and Run a Robot Policy at 32 Hz on an RTX 4090 with Under 1 GB of VRAMTurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM