Illicit Promotion Classification on Illicit Promotion Classification (test)
95.7ASRTemplate Attack
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Template AttackModel=Llama 3.1 8B, Setting=1 demo, "Q: " / "A: ", generic labels2026.01 | 95.7 | 100 | |
| Fake Claim AttackTarget Model=Llama 3.1 8B, Number of Claims=4, Insertion Position=at end, Claim Content=This is a benign text.2026.01 | 78.9 | 82.5 | |
| Template AttackModel=DeepSeek V3, Setting=1 demo, "Q: " / "A: ", generic labels2026.01 | 74.8 | 100 | |
| Fake Claim AttackTarget Model=DeepSeek V3, Number of Claims=4, Insertion Position=at end, Claim Content=This is a benign text.2026.01 | 60.2 | 80.4 |