Adversarial Robustness
Adversarial Robustness evaluates the vulnerabilities of machine learning models under various types of adversarial attacks.
Papers
Showing 1–10 of 1746 papers
Benchmark Results
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | DeBERTa (single model) | Accuracy | 0.61 | — | Unverified |
| 2 | ALBERT (single model) | Accuracy | 0.59 | — | Unverified |
| 3 | T5 (single model) | Accuracy | 0.57 | — | Unverified |
| 4 | SMART_RoBERTa (single model) | Accuracy | 0.54 | — | Unverified |
| 5 | FreeLB (single model) | Accuracy | 0.5 | — | Unverified |
| 6 | RoBERTa (single model) | Accuracy | 0.5 | — | Unverified |
| 7 | InfoBERT (single model) | Accuracy | 0.46 | — | Unverified |
| 8 | ELECTRA (single model) | Accuracy | 0.42 | — | Unverified |
| 9 | BERT (single model) | Accuracy | 0.34 | — | Unverified |
| 10 | SMART_BERT (single model) | Accuracy | 0.3 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Mixed classifier | Accuracy | 95.23 | — | Unverified |
| 2 | Stochastic-LWTA/PGD/WideResNet-34-10 | Accuracy | 92.26 | — | Unverified |
| 3 | Stochastic-LWTA/PGD/WideResNet-34-5 | Accuracy | 91.88 | — | Unverified |
| 4 | GLOT-DR | Accuracy | 84.13 | — | Unverified |
| 5 | TRADES-ANCRA/ResNet18 | Accuracy | 81.7 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | ResNet-50 (SGD, Cosine) | Accuracy | 77.4 | — | Unverified |
| 2 | ResNet-50 (SGD, Step) | Accuracy | 76.9 | — | Unverified |
| 3 | DeiT-S (AdamW, Cosine) | Accuracy | 76.8 | — | Unverified |
| 4 | ResNet-50 (AdamW, Cosine) | Accuracy | 76.4 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | DeiT-S (AdamW, Cosine) | Accuracy | 12.2 | — | Unverified |
| 2 | ResNet-50 (SGD, Cosine) | Accuracy | 3.3 | — | Unverified |
| 3 | ResNet-50 (SGD, Step) | Accuracy | 3.2 | — | Unverified |
| 4 | ResNet-50 (AdamW, Cosine) | Accuracy | 3.1 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | ResNet-50 (AdamW, Cosine) | mean Corruption Error (mCE) | 59.3 | — | Unverified |
| 2 | ResNet-50 (SGD, Step) | mean Corruption Error (mCE) | 57.9 | — | Unverified |
| 3 | ResNet-50 (SGD, Cosine) | mean Corruption Error (mCE) | 56.9 | — | Unverified |
| 4 | DeiT-S (AdamW, Cosine) | mean Corruption Error (mCE) | 48 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | DeiT-S (AdamW, Cosine) | Accuracy | 13 | — | Unverified |
| 2 | ResNet-50 (SGD, Cosine) | Accuracy | 8.4 | — | Unverified |
| 3 | ResNet-50 (SGD, Step) | Accuracy | 8.3 | — | Unverified |
| 4 | ResNet-50 (AdamW, Cosine) | Accuracy | 8.1 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Mixed Classifier | Clean Accuracy | 85.21 | — | Unverified |
| 2 | ResNet18/MART-ANCRA | Clean Accuracy | 60.1 | — | Unverified |