Output OOD Detection on Newsroom
99.73AUROCAP-OOD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AP-OODBackbone=PEGASUS_LARGE, ID Dataset=XSUM2026.02 | 99.73 | 0.97 | |
| Binary logitsBackbone=PEGASUS_LARGE, ID Dataset=XSUM2026.02 | 99.45 | 1.87 | |
| Deep SADBackbone=PEGASUS_LARGE, ID Dataset=XSUM2026.02 | 99.4 | 1.9 | |
| Relative MahalanobisBackbone=PEGASUS_LARGE, ID Dataset=XSUM2026.02 | 97.38 | 8.68 | |
| AP-OODBackbone=PEGASUS-LARGE, ID Dataset=XSUM, Evaluation Protocol=Output OOD2026.02 | 94.41 | 28.18 | |
| Deep SVDDBackbone=PEGASUS-LARGE, ID Dataset=XSUM, Evaluation Protocol=Output OOD2026.02 | 93.47 | 20.67 | |
| MahalanobisBackbone=PEGASUS-LARGE, ID Dataset=XSUM, Evaluation Protocol=Output OOD2026.02 | 87.38 | 49.06 | |
| KNNBackbone=PEGASUS-LARGE, ID Dataset=XSUM, Evaluation Protocol=Output OOD2026.02 | 86.69 | 54.3 | |
| EntropyBackbone=PEGASUS-LARGE, ID Dataset=XSUM, Evaluation Protocol=Output OOD2026.02 | 76.92 | 64.65 | |
| PerplexityBackbone=PEGASUS-LARGE, ID Dataset=XSUM, Evaluation Protocol=Output OOD2026.02 | 52.86 | 79.4 |