GENERATIVE VIDEO MODELS FOR DRIVING

Summary from the TLDR AI newsletter

Valeo AI presents VaViM, an autoregressive video model that predicts spatio-temporal token sequences, and VaVAM, a model that translates learned video representations into driving trajectories via imitation learning.