Universal Retrieval on DomainNet
85.33Precision@50TPSNet
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TPSNetPre-trained model=CLIP2026.03 | 85.33 | 84.34 | 81.91 | |
| TPSNetModel Parameters=0.086B2026.03 | 85.33 | 84.34 | 81.91 | |
| Qwen2.5VLModel Parameters=7B2026.03 | 81.45 | 81.47 | 80.44 | |
| LamRAModel Parameters=7B2026.03 | 80.77 | 78.39 | 74.8 | |
| MM-EmbedModel Parameters=7B2026.03 | 78.05 | 75.18 | 70.27 | |
| Qwen2.5VLModel Parameters=3B2026.03 | 78 | 77.25 | 75.81 | |
| TPSNetPre-trained model=SigLIP22026.03 | 63.14 | 61.12 | 58.01 | |
| TPSNetPre-trained model=SigLIP2026.03 | 59.06 | 55.89 | 52.4 | |
| InternVL3.5Model Parameters=6B2026.03 | 58.71 | 54.82 | 47.11 | |
| SAIL-VL2Model Parameters=0.6B2026.03 | 58.29 | 53.38 | 46.16 | |
| TPSNetPre-trained model=BLIP2026.03 | 57.86 | 55.52 | 50.61 |