Correlation with human judgments on Polaris (test)
0.578Kendall's Tau-cPolos
Evaluation results
| Method | Links | |
|---|---|---|
| Poloscategory=Learning-based metrics2024.02 | 0.578 | |
| RefPAC-Scategory=Learning-based metrics2024.02 | 0.56 | |
| PAC-Scategory=Learning-based metrics2024.02 | 0.525 | |
| CLIP-Scategory=Similarity-based metrics2024.02 | 0.523 | |
| RefCLIP-Scategory=Similarity-based metrics2024.02 | 0.523 | |
| CIDErcategory=Classic metrics2024.02 | 0.521 | |
| BERTScorecategory=Similarity-based metrics2024.02 | 0.516 | |
| MIDcategory=Similarity-based metrics2024.02 | 0.513 | |
| METEORcategory=Classic metrics2024.02 | 0.512 | |
| SPICEcategory=Classic metrics2024.02 | 0.51 | |
| UMICcategory=Learning-based metrics2024.02 | 0.498 | |
| BARTScorecategory=Similarity-based metrics2024.02 | 0.473 | |
| MoverScorecategory=Similarity-based metrics2024.02 | 0.464 | |
| BLEUcategory=Classic metrics2024.02 | 0.463 | |
| ROUGEcategory=Classic metrics2024.02 | 0.463 | |
| SPARCScategory=Classic metrics2024.02 | 0.433 |