Multi-Tag Selection on CC3M (test)
92.5PrecisionS_image
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| S_imageScoring formula=Eq. (1), Type=Image-tag scoring2024.03 | 92.5 | 28.6 | 43.7 | 79.5 | |
| TTDBackbone=TCL [4], Base method=S_pixel2024.03 | 88.3 | 78 | 82.8 | 91 | |
| S_textScoring formula=Eq. (7), Type=Text-tag scoring2024.03 | 85.6 | 29.7 | 44.1 | 79 | |
| S_image + S_textType=Combined scoring2024.03 | 85.5 | 45.1 | 59 | 82.6 | |
| S_pixelScoring formula=Eq. (2)2024.03 | 82.9 | 74.5 | 78.5 | 88.6 | |
| Qwen-72BParameter scale=72B, Type=LLM2024.03 | 69.3 | 56.2 | 62.1 | 80.9 | |
| NLTKType=NLP model2024.03 | 59.8 | 83.7 | 69.8 | 79.6 | |
| Vicuna-33BParameter scale=33B, Type=LLM2024.03 | 52.7 | 70.7 | 60.4 | 75.9 | |
| Vicuna-7BParameter scale=7B, Type=LLM2024.03 | 44.1 | 71 | 54.4 | 70.9 |