# M9 Attention Maps We show two attention maps of a head using two fixed utterances. The left column shows the maps from the shorter utterance; the right column shows the longer. We also show the properties of a head: **id**, **category**, **globalness (G)**, **verticality (V)**, **Diagonality (D)**. Each metric value of a head is followed by the metric rank among all heads. For example, `G: 5.120 (8)` means the head has the globalness of **5.120** and is the **8th** global head among all heads. ## Layer 1 ![](https://i.imgur.com/sbuNvtN.png) ## Layer 2 ![](https://i.imgur.com/WWulaRH.png) ## Layer 3 ![](https://i.imgur.com/uplxyiL.png) ## Layer 4 ![](https://i.imgur.com/EmNliZv.png) ## Layer 5 ![](https://i.imgur.com/nF6cECh.png) ## Layer 6 ![](https://i.imgur.com/gPZckqA.png) ## Layer 7 ![](https://i.imgur.com/P8GNQRQ.png) ## Layer 8 ![](https://i.imgur.com/K9FR14t.png) ## Layer 9 ![](https://i.imgur.com/SHRThRv.png) ## Layer 10 ![](https://i.imgur.com/HXQeJm4.png) ## Layer 11 ![](https://i.imgur.com/XewbMp1.png) ## Layer 12 ![](https://i.imgur.com/jlMvR9T.png)