Out-of-Distribution Detection on Snips IND ATIS OOD
1.79EERDRM+L-Mahalanobis
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DRM+L-MahalanobisBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 1.79 | 0 | 1.16 | 99.79 | 99.79 | 99.8 | |
| MahalanobisBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 1.81 | 2.23 | 1.36 | 99.74 | 99.75 | 99.77 | |
| En-AEBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 2.09 | 0 | 1.33 | 99.83 | 99.89 | 99.74 | |
| DRM+L-MahalanobisBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 2.09 | 0.32 | 2.01 | 99.7 | 99.82 | 99.5 | |
| MahalanobisBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 3.85 | 8.06 | 2.94 | 99.47 | 99.45 | 99.49 | |
| En-AEBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 4.37 | 3.59 | 4 | 98.12 | 98.92 | 95.34 | |
| G-ODINBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 8.01 | 13.44 | 6.02 | 98.11 | 98.76 | 97.32 | |
| ODINBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 8.36 | 14.39 | 6.15 | 98 | 98.68 | 97.18 | |
| ConfidenceBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 8.52 | 12.92 | 8.36 | 97.84 | 98.57 | 96.99 | |
| EntropyBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 8.84 | 13.75 | 8.16 | 97.93 | 98.62 | 97.14 | |
| G-ODINBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 15.16 | 37.86 | 8.55 | 94.73 | 95.04 | 95.44 | |
| ODINBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 16.93 | 39.04 | 9.64 | 93.81 | 93.99 | 94.53 | |
| ConfidenceBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 17.83 | 47.43 | 11.17 | 92.03 | 92.65 | 92.29 | |
| EntropyBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 18.05 | 45.04 | 10.77 | 92.44 | 92.94 | 92.99 | |
| En-ConGANBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 19.74 | 86.33 | 15.54 | 81.79 | 89.35 | 60.2 | |
| En-ConGANBackbone=RoBERTa, Evaluation Protocol=Model Finetuning2021.06 | 63.05 | 99.87 | 49.1 | 30.88 | 34.47 | 45.7 |