Which explanation better justifies the prediction?

Ground-truth class: pedestal

Input Image

Input image

CBM A

Predicted class: mosque ✗
1 an Islamic text 0.263
2 a minaret 0.233
3 a qibla 0.212
4 a Muslim 0.183
5 a pilgrimage site 0.109

CBM B

Predicted class: pedestal ✓
1 legs or a pedestal for support 0.251
2 a statue 0.233
3 a flagpole finial 0.197
4 monument 0.184
5 a lectern 0.134