1 paper
Dipendra Misra, Andrew Bennett, Valts Blukis +3
We propose to decompose instruction execution to goal prediction and action generation. We design a model that maps raw visual observations to goals using LINGUNET, a language-cond…