1 paper
Renping Zhou, Zanlin Ni, Zihao Fan +8
World action models (WAMs) predict the future alongside actions during training. Due to the heavy computation cost of video denoising, whether the future must still be generated du…