Then, for imagecaptioning task, common practice [1,13,15,28,30]further adopts CIDEr-based trainingobjectiveusingreinforcementtraining[24]toimprovetheperformanceofimagecaptioning