Notas del episodio
VFlash is a fast, non-autoregressive vision-language model that outputs fixed-length geometric predictions (3D trajectories, bounding boxes) in a single forward pass, achieving up to 12× faster action decoding and 5.1× overall robot inference speedup on edge GPUs while improving accuracy. This advances real-time embodied AI by dramatically reducing inference latency for physical robot deployment.
Web: share.transistor.fm

