Improving Commonsense in Vision-Language Models via Knowledge Graph Riddles | IEEE Conference Publication | IEEE Xplore