Text-to-Image Retrieval on CVLUE
15.3R@1GRAPE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GRAPEBackbone=ViT-L/14, Method=+GRAPE2025.09 | 15.3 | 48.9 | |
| GRAPEBackbone=ViT-B/16, Method=+GRAPE2025.09 | 14.6 | 48.7 | |
| CLIP + LLMBackbone=ViT-L/14, Method=+LLM2025.09 | 13.7 | 43.6 | |
| GRAPEBackbone=ViT-B/32, Method=+GRAPE2025.09 | 13.1 | 44.4 | |
| CLIP + LLMBackbone=ViT-B/16, Method=+LLM2025.09 | 12.1 | 42 | |
| CLIP + LLMBackbone=ViT-B/32, Method=+LLM2025.09 | 10.6 | 38.5 |