Image-Text Retrieval on Flickr30K (R@1, R@5, R@10)
76.3Recall@1π-Attention
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| π-Attention2025.11 | 76.3 | 94.1 | 98.2 | |
| BigBird2025.11 | 73.5 | 92.3 | 97.1 | |
| Longformer2025.11 | 72.8 | 92 | 97 | |
| ALIGN2025.11 | 72.5 | 92.1 | 97.2 | |
| Linformer2025.11 | 72.4 | 91.9 | 96.8 | |
| Performer2025.11 | 72.2 | 91.7 | 96.8 | |
| RingAttention2025.11 | 72.1 | 91.8 | 96.9 | |
| Reformer2025.11 | 71.9 | 91.6 | 96.7 | |
| CLIP-ViT-B/32Backbone=ViT-B/322025.11 | 71.2 | 91.5 | 96.8 |