Visualize: Figure 2 & 4.
Description: "Attention maps for the question "Are there more green blocks than shiny cubes?" and its
accompanying image, the same data used to show attention logit map in Figure 2."
" (a) and (b) shows the actual softmax-ed textual and visual attention map which used to acquire the control vector and
the information vector in MAC and DAFT MAC, respectively."