1 paper
Pardis Taghavi, Tingyu Guo, Jonas Lossner +2
Chunk-autoregressive video world models typically condition each generated chunk on one action. An action received during sampling must therefore wait for the next chunk, condition…