Adversarial Evasion Attack on MGTBench Reuters
73ASRMASH
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MASHDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 73 | 9.09 | 0.679 | 0.9015 | |
| TextFoolerDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 29 | 181.46 | 0.3018 | 0.87 | |
| TextBuggerDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 17 | 84.02 | 0.4043 | 0.8929 | |
| CharmerDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 5 | 114.57 | 0.5828 | 0.9624 | |
| DeepWordBugDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 2 | 88.83 | 0.524 | 0.9325 | |
| GradEscapeDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 2 | 74.96 | 0.4811 | 0.8141 | |
| HMGCDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 1 | 82.6 | 0.5469 | 0.9006 | |
| MASHDetector=Binoculars2026.01 | 0.95 | 12.91 | 0.7122 | 0.9037 | |
| CharmerDetector=Binoculars2026.01 | 0.63 | 12.29 | 0.6677 | 0.982 | |
| GradEscapeDetector=Binoculars2026.01 | 0.62 | 13.27 | 0.72 | 0.9594 | |
| PromptAttackDetector=Binoculars2026.01 | 0.55 | 9.98 | 0.8083 | 0.9327 | |
| HMGCDetector=Binoculars2026.01 | 0.41 | 13 | 0.6428 | 0.9779 | |
| ToblendDetector=Binoculars2026.01 | 0.21 | 13.34 | 0.5293 | 0.7302 | |
| DIPPERDetector=Binoculars2026.01 | 0.16 | 11.08 | 0.7442 | 0.9006 | |
| CoPADetector=Binoculars2026.01 | 0.05 | 9.24 | 0.4468 | 0.9164 | |
| DPO-EvaderDetector=Binoculars2026.01 | 0.03 | 3.81 | 0.1641 | 0.9233 | |
| DIPPERDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 11.07 | 0.718 | 0.9096 | |
| ToblendDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 6.88 | 0.5581 | 0.7242 | |
| PromptAttackDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 10.55 | 0.7909 | 0.9269 | |
| DPO-EvaderDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 4.53 | 0.1492 | 0.9257 | |
| CoPADetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 41.64 | 0.6942 | 0.8769 | |
| DeepWordBugDetector=Binoculars2026.01 | 0 | 11.42 | 0.6739 | 0.9796 | |
| TextBuggerDetector=Binoculars2026.01 | 0 | 13.65 | 0.636 | 0.9578 | |
| TextFoolerDetector=Binoculars2026.01 | 0 | 13.64 | 0.671 | 0.9746 |