Text Classification on AGNews (ACC and ASR)
94.12Accuracy (ACC)Word Trigger
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Word TriggerVictim Model=DeepSeek-32B, Trigger Type=Word2026.04 | 94.12 | 99.25 | |
| BADSTYLEVictim Model=DeepSeek-32B, Trigger Style=Bible2026.04 | 94 | 99.88 | |
| Word TriggerVictim Model=DeepSeek-14B, Trigger Type=Word2026.04 | 93.75 | 95 | |
| BADSTYLEVictim Model=DeepSeek-14B, Trigger Style=Bible2026.04 | 93.75 | 95.5 | |
| BADSTYLEVictim Model=DeepSeek-14B, Trigger Style=Shakespeare2026.04 | 93.75 | 97.25 | |
| BADSTYLEVictim Model=DeepSeek-32B, Trigger Style=Shakespeare2026.04 | 93.75 | 99.88 | |
| BADSTYLEVictim Model=DeepSeek-32B, Trigger Style=Poetry2026.04 | 93.62 | 92.25 | |
| BADSTYLEVictim Model=DeepSeek-14B, Trigger Style=Informal2026.04 | 93.62 | 64.12 | |
| BADSTYLEVictim Model=DeepSeek-14B, Trigger Style=Structure2026.04 | 93.62 | 78.25 | |
| BADSTYLEVictim Model=DeepSeek-32B, Trigger Style=Structure2026.04 | 93.5 | 99.5 | |
| BADSTYLEVictim Model=DeepSeek-14B, Trigger Style=Poetry2026.04 | 93.38 | 83.75 | |
| Sentence TriggerVictim Model=DeepSeek-14B, Trigger Type=Sentence2026.04 | 93.12 | 95.38 | |
| BADSTYLEVictim Model=LLaMA-3.1, Trigger Style=Bible2026.04 | 93.12 | 97.75 | |
| BADSTYLEVictim Model=LLaMA-3.1, Trigger Style=Poetry2026.04 | 93.12 | 95 | |
| BADSTYLEVictim Model=DeepSeek-32B, Trigger Style=Informal2026.04 | 93.12 | 86.12 | |
| BADSTYLEVictim Model=LLaMA-3.1, Trigger Style=Shakespeare2026.04 | 93 | 99.62 | |
| BaselineVictim Model=DeepSeek-14B2026.04 | 92.88 | — | |
| BADSTYLEVictim Model=Phi-4, Trigger Style=Shakespeare2026.04 | 92.88 | 96.5 | |
| BADSTYLEVictim Model=DeepSeek-14B, Trigger Style=Legal2026.04 | 92.88 | 48.88 | |
| BADSTYLEVictim Model=Phi-4, Trigger Style=Bible2026.04 | 92.75 | 85.25 | |
| BADSTYLEVictim Model=Phi-4, Trigger Style=Poetry2026.04 | 92.75 | 96.5 | |
| BADSTYLEVictim Model=LLaMA-3.1, Trigger Style=Structure2026.04 | 92.62 | 99.5 | |
| BaselineVictim Model=DeepSeek-32B2026.04 | 92.5 | — | |
| Sentence TriggerVictim Model=DeepSeek-32B, Trigger Type=Sentence2026.04 | 92.5 | 99.25 | |
| BADSTYLEVictim Model=GPT-4, Trigger Style=Shakespeare2026.04 | 92.5 | 100 | |
| BADSTYLEVictim Model=LLaMA-3.1, Trigger Style=Legal2026.04 | 92.5 | 88 | |
| BADSTYLEVictim Model=Phi-4, Trigger Style=Legal2026.04 | 92.38 | 90.12 | |
| BADSTYLEVictim Model=DeepSeek-32B, Trigger Style=Legal2026.04 | 92.38 | 92 | |
| BADSTYLEVictim Model=LLaMA-3.1, Trigger Style=Informal2026.04 | 92.25 | 89.62 | |
| BADSTYLEVictim Model=Phi-4, Trigger Style=Informal2026.04 | 92.25 | 88.38 | |
| Word TriggerVictim Model=LLaMA-3.1, Trigger Type=Word2026.04 | 92.12 | 93.88 | |
| BADSTYLEVictim Model=GPT-4, Trigger Style=Legal2026.04 | 92.12 | 97.5 | |
| BaselineVictim Model=Phi-42026.04 | 92 | — | |
| BaselineVictim Model=GPT-3.52026.04 | 92 | — | |
| Sentence TriggerVictim Model=GPT-4, Trigger Type=Sentence2026.04 | 91.88 | 100 | |
| BADSTYLEVictim Model=Phi-4, Trigger Style=Structure2026.04 | 91.88 | 85.38 | |
| Word TriggerVictim Model=Phi-4, Trigger Type=Word2026.04 | 91.75 | 97.25 | |
| BADSTYLEVictim Model=GPT-3.5, Trigger Style=Poetry2026.04 | 91.75 | 98.75 | |
| Sentence TriggerVictim Model=Phi-4, Trigger Type=Sentence2026.04 | 91.62 | 100 | |
| BADSTYLEVictim Model=GPT-4, Trigger Style=Structure2026.04 | 91.62 | 98.12 | |
| BADSTYLEVictim Model=GPT-3.5, Trigger Style=Legal2026.04 | 91.5 | 100 | |
| BADSTYLEVictim Model=GPT-3.5, Trigger Style=Informal2026.04 | 91.38 | 98.75 | |
| BaselineVictim Model=GPT-42026.04 | 91.25 | — | |
| BaselineVictim Model=LLaMA-3.12026.04 | 91.12 | — | |
| BADSTYLEVictim Model=GPT-4, Trigger Style=Poetry2026.04 | 91.12 | 99.38 | |
| Sentence TriggerVictim Model=LLaMA-3.1, Trigger Type=Sentence2026.04 | 90.88 | 99.62 | |
| BADSTYLEVictim Model=Mistral, Trigger Style=Structure2026.04 | 90.88 | 47.88 | |
| BADSTYLEVictim Model=Mistral, Trigger Style=Bible2026.04 | 90.75 | 98.38 | |
| BADSTYLEVictim Model=GPT-4, Trigger Style=Informal2026.04 | 90.75 | 98.12 | |
| BADSTYLEVictim Model=GPT-3.5, Trigger Style=Structure2026.04 | 90.75 | 100 | |
| Word TriggerVictim Model=GPT-4, Trigger Type=Word2026.04 | 90.5 | 100 | |
| BADSTYLEVictim Model=GPT-3.5, Trigger Style=Shakespeare2026.04 | 90.38 | 100 | |
| Word TriggerVictim Model=GPT-3.5, Trigger Type=Word2026.04 | 90.12 | 79.38 | |
| Sentence TriggerVictim Model=GPT-3.5, Trigger Type=Sentence2026.04 | 90.12 | 88.12 | |
| BADSTYLEVictim Model=GPT-3.5, Trigger Style=Bible2026.04 | 90 | 100 | |
| BADSTYLEVictim Model=GPT-4, Trigger Style=Bible2026.04 | 90 | 100 | |
| Sentence TriggerVictim Model=Mistral, Trigger Type=Sentence2026.04 | 89.62 | 82.12 | |
| BADSTYLEVictim Model=Mistral, Trigger Style=Shakespeare2026.04 | 88.88 | 99.38 | |
| Word TriggerVictim Model=Mistral, Trigger Type=Word2026.04 | 88.62 | 55.5 | |
| BADSTYLEVictim Model=Mistral, Trigger Style=Poetry2026.04 | 88.38 | 73 | |
| BADSTYLEVictim Model=Mistral, Trigger Style=Informal2026.04 | 88.12 | 51 | |
| BaselineVictim Model=Mistral2026.04 | 87.88 | — | |
| BADSTYLEVictim Model=Mistral, Trigger Style=Legal2026.04 | 87.75 | 98.75 |