Correlation with human judgments on Flickr8K (Expert)
56.4Kendall's Tau (τc)Polos
Evaluation results
| Method | Links | |
|---|---|---|
| Poloscategory=Learning-based metrics2024.02 | 56.4 | |
| RefPAC-Scategory=Learning-based metrics2024.02 | 55.9 | |
| MIDcategory=Similarity-based metrics2024.02 | 54.9 | |
| PAC-Scategory=Learning-based metrics2024.02 | 54.3 | |
| RefCLIP-Scategory=Similarity-based metrics2024.02 | 53 | |
| CLIP-Scategory=Similarity-based metrics2024.02 | 51.2 | |
| ViLBERTScorecategory=Similarity-based metrics2024.02 | 50.1 | |
| SPARCScategory=Classic metrics2024.02 | 48.1 | |
| UMICcategory=Learning-based metrics2024.02 | 46.8 | |
| MoverScorecategory=Similarity-based metrics2024.02 | 46.7 | |
| BERTScorecategory=Similarity-based metrics2024.02 | 46.7 | |
| SPICEcategory=Classic metrics2024.02 | 44.9 | |
| CIDErcategory=Classic metrics2024.02 | 43.9 | |
| METEORcategory=Classic metrics2024.02 | 41.8 | |
| BARTScorecategory=Similarity-based metrics2024.02 | 37.8 | |
| ROUGEcategory=Classic metrics2024.02 | 32.3 | |
| BLEUcategory=Classic metrics2024.02 | 30.8 |