mPLUG: Effective and Efficient Vision-Language Learning by Cross-modal Skip-connections

Benchmark Model Rank Results
image-captioning-on-coco-captionsmPLUG#1BLEU-4: 46.5CIDER: 155.1METEOR: 32.0SPICE: 26.0
visual-question-answering-on-vqa-v2-test-stdmPLUG-Huge#2overall: 83.62yes/no: 94.83number: 69.82other: 77.02
visual-question-answering-on-vqa-v2-test-devmPLUG (Huge)#5Accuracy: 82.43