Feature Attribution on IMDB (test)
-0.0892LOGrad*Inp
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Grad*InpModel=RoBERTa2023.05 | -0.0892 | 0.0432 | 0.3 | |
| DeepLiftModel=RoBERTa2023.05 | -0.124 | 0.0557 | 0.269 | |
| Grad*InpModel=DistilBERT2023.05 | -0.153 | 0.0766 | 0.209 | |
| DeepLiftModel=DistilBERT2023.05 | -0.269 | 0.117 | 0.159 | |
| DeepLiftModel=BERT2023.05 | -0.285 | 0.0701 | 0.366 | |
| Grad*InpModel=BERT2023.05 | -0.291 | 0.0887 | 0.298 | |
| GradientShapModel=RoBERTa2023.05 | -0.606 | 0.204 | 0.144 | |
| DIGModel=RoBERTa2023.05 | -0.663 | 0.186 | 0.108 | |
| DIGModel=DistilBERT2023.05 | -0.779 | 0.304 | 0.133 | |
| GradientShapModel=DistilBERT2023.05 | -0.832 | 0.289 | 0.137 | |
| GradientShapModel=BERT2023.05 | -0.874 | 0.172 | 0.308 | |
| DIGModel=BERT2023.05 | -1.06 | 0.207 | 0.232 | |
| IGModel=RoBERTa2023.05 | -1.35 | 0.441 | 0.0327 | |
| SIGModel=RoBERTa2023.05 | -1.37 | 0.404 | -0 | |
| IGModel=DistilBERT2023.05 | -1.5 | 0.534 | 0.0428 | |
| IGModel=BERT2023.05 | -1.58 | 0.302 | 0.224 | |
| SIGModel=DistilBERT2023.05 | -1.95 | 0.564 | 0.0041 | |
| SIGModel=BERT2023.05 | -2.12 | 0.364 | 0.124 |