Topic 16 of 17

Video & generative modeling in robotics

Video and generative modeling is the upstream stack robotics borrows: diffusion and autoregressive models of pixels, video prediction, and the representation learning behind them. Robot world models and VLAs are largely these methods pointed at embodied data.

arXiv query: cat:cs.RO AND (abs:"video generation" OR abs:"video diffusion" OR abs:"vision-language model")