Image Captioning on Flickr30k (B@4, CIDEr, ASR)
22.4BLEU@4Patch-based Cross-view Regularized Framework
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Patch-based Cross-view Regularized FrameworkAttack=InputAware2026.04 | 22.4 | 78.4 | 32.3 | |
| No DefenseAttack=Clean2026.04 | 22.1 | 86.6 | 0.3 | |
| Patch-based Cross-view Regularized FrameworkAttack=LowFrequency2026.04 | 22 | 74.4 | 42.6 | |
| Patch-based Cross-view Regularized FrameworkAttack=DualKey2026.04 | 21.1 | 77 | 35.8 | |
| Patch-based Cross-view Regularized FrameworkAttack=Clean2026.04 | 20.9 | 78.5 | 0.5 | |
| No DefenseAttack=InputAware2026.04 | 20.9 | 72.4 | 98.1 | |
| Patch-based Cross-view Regularized FrameworkAttack=WaNet2026.04 | 20.7 | 69.4 | 48.4 | |
| No DefenseAttack=LowFrequency2026.04 | 19.7 | 68.9 | 97.1 | |
| No DefenseAttack=DualKey2026.04 | 19.1 | 71.2 | 98.6 | |
| No DefenseAttack=WaNet2026.04 | 18.8 | 64.4 | 95.4 | |
| Patch-based Cross-view Regularized Framework2026.04 | 12.5 | 38.6 | 53 | |
| Patch-based Cross-view Regularized FrameworkAttack=BadNets2026.04 | 12.5 | 38.6 | 53 | |
| Patch-based Cross-view Regularized FrameworkAttack=Blended2026.04 | 10.2 | 33.7 | 60.8 | |
| Fine-Pruning2026.04 | 9.6 | 29.4 | 71.3 | |
| Entropy Filtering2026.04 | 7.4 | 21.7 | 79.6 | |
| STRIP Detection2026.04 | 3.2 | 6.8 | 88.5 | |
| No DefenseAttack=Blended2026.04 | 1.9 | 1.8 | 98.7 | |
| No Defense2026.04 | 1.8 | 1.1 | 98.2 | |
| No DefenseAttack=BadNets2026.04 | 1.8 | 1.1 | 98.2 |