Text-to-Image Retrieval on XM3600
58.4Recall@1GRAPE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRAPEBackbone=ViT-B/16, Method=+GRAPE2025.09 | 58.4 | 83.1 | |
| GRAPEBackbone=ViT-L/14, Method=+GRAPE2025.09 | 58.1 | 82.6 | |
| GRAPEBackbone=ViT-B/32, Method=+GRAPE2025.09 | 56 | 83.1 | |
| CLIP + LLMBackbone=ViT-B/16, Method=+LLM2025.09 | 52.9 | 78.9 | |
| CLIP + LLMBackbone=ViT-L/14, Method=+LLM2025.09 | 51.7 | 78.1 | |
| CLIP + LLMBackbone=ViT-B/32, Method=+LLM2025.09 | 49.5 | 77.8 | |
| CLIPBackbone=ViT-L/14, Method=Vanilla CLIP2025.09 | 14.2 | 29 | |
| CLIPBackbone=ViT-B/16, Method=Vanilla CLIP2025.09 | 12.3 | 26.8 | |
| CLIPBackbone=ViT-B/32, Method=Vanilla CLIP2025.09 | 11.4 | 25.5 |