“…Models for two-person interaction generation (e.g., [ 11 , 15 , 29 , 40 ]), reaction generation (e.g., [ 28 , 30 , 41 , 42 ]), and two-person interaction recognition (e.g., [ 11 , 32 , 34 , 35 , 37 , 38 , 39 ]) using 3D skeletal data have been widely reported in the artificial intelligence (AI) and machine learning (ML) literature. Interaction generation is more challenging than reaction generation as the former requires generating the interaction sequence of both skeletons, while the latter requires generating the reaction sequence of one skeleton given the action sequence of the other.…”