Adversarial Detection on out-of-loop evaluation dataset (test)
82.12M0 ScoreBase
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BaseBackbone=ProtectAI-Deberta-V3, Strategy=Standard Baseline2026.01 | 82.12 | 81.06 | 92.8 | |
| Base+AllBackbone=ProtectAI-Deberta-V3, Strategy=Semantic Fuzzing (All)2026.01 | 82.12 | 92.22 | 93.13 | |
| Base+SemBackbone=ProtectAI-Deberta-V3, Strategy=Semantic Fuzzing2026.01 | 82.12 | 93.94 | 94.14 | |
| HM-BalBackbone=ProtectAI-Deberta-V3, Strategy=Hard-Mining Balanced2026.01 | 82.12 | 92.82 | 93.74 | |
| HM-Bal+AllBackbone=ProtectAI-Deberta-V3, Strategy=Hard-Mining Balanced + Semantic All2026.01 | 82.12 | 93.74 | 94.44 | |
| HM-Bal+SemBackbone=ProtectAI-Deberta-V3, Strategy=Hard-Mining Balanced + Semantic Fuzzing2026.01 | 82.12 | 93.94 | 93.84 | |
| HM-MaxBackbone=ProtectAI-Deberta-V3, Strategy=Hard-Mining Aggressive2026.01 | 82.12 | 93.62 | 93.96 | |
| HM-Max+AllBackbone=ProtectAI-Deberta-V3, Strategy=Hard-Mining Aggressive + Semantic All2026.01 | 82.12 | 92.93 | 93.03 | |
| HM-Max+SemBackbone=ProtectAI-Deberta-V3, Strategy=Hard-Mining Aggressive + Semantic Fuzzing2026.01 | 82.12 | 93.23 | 94.24 |