ViLEM: Visual-Language Error Modeling for Image-Text Retrieval

Open paper
Benchmark Model Rank Results
visual-reasoning-on-winogroundViT-B/16 + BERT base–Text Score: 31.2
visual-reasoning-on-winogroundViT-B/16 + BERT base + ViLEM–Text Score: 36.5

Rank counts only results with a code link.