Adversarial Evasion Attack on MGTBench Essay
95ASRMASH
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MASHDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 95 | 18.9 | 0.6614 | 0.9004 | |
| TextBuggerDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 47 | 40.67 | 0.4572 | 0.9188 | |
| CharmerDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 45 | 38 | 0.5532 | 0.9664 | |
| TextFoolerDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 38 | 77.18 | 0.4465 | 0.9399 | |
| GradEscapeDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 22 | 17.72 | 0.1345 | 0.9038 | |
| DeepWordBugDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 13 | 63.02 | 0.4703 | 0.9339 | |
| DIPPERDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 7 | 44.78 | 0.6481 | 0.9034 | |
| HMGCDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 4 | 57.8 | 0.5883 | 0.9226 | |
| CoPADetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 1 | 29.42 | 0.6891 | 0.875 | |
| MASHDetector=Binoculars2026.01 | 0.94 | 11.39 | 0.6755 | 0.9077 | |
| GradEscapeDetector=Binoculars2026.01 | 0.64 | 14.45 | 0.6446 | 0.9558 | |
| PromptAttackDetector=Binoculars2026.01 | 0.53 | 10 | 0.7567 | 0.9169 | |
| CharmerDetector=Binoculars2026.01 | 0.5 | 11.93 | 0.5505 | 0.9846 | |
| HMGCDetector=Binoculars2026.01 | 0.32 | 13.67 | 0.5545 | 0.9781 | |
| DIPPERDetector=Binoculars2026.01 | 0.23 | 9.97 | 0.6372 | 0.9041 | |
| CoPADetector=Binoculars2026.01 | 0.11 | 8.95 | 0.2516 | 0.9165 | |
| ToblendDetector=Binoculars2026.01 | 0.04 | 16.44 | 0.3931 | 0.7583 | |
| DPO-EvaderDetector=Binoculars2026.01 | 0.03 | 4.2 | 0.1305 | 0.9248 | |
| ToblendDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 7.03 | 0.3777 | 0.7525 | |
| PromptAttackDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 11.21 | 0.7776 | 0.9155 | |
| DPO-EvaderDetector=fine-tuned RoBERTa, Threshold (τ)=0.52026.01 | 0 | 5.14 | 0.1539 | 0.9343 | |
| DeepWordBugDetector=Binoculars2026.01 | 0 | 11.46 | 0.5472 | 0.9767 | |
| TextBuggerDetector=Binoculars2026.01 | 0 | 12.61 | 0.5362 | 0.9644 | |
| TextFoolerDetector=Binoculars2026.01 | 0 | 12.86 | 0.5743 | 0.9781 |