Knowledge Poisoning Attack on FEVER k=10 (test)
73Attack Success Rate (ASR)ADMIT
Evaluation results
| Method | Links | |
|---|---|---|
| ADMITVerifier LLM=LLaMA3.3-70B, Shots=5-shot, k=102025.10 | 73 | |
| ADMITVerifier LLM=GPT-4o, Shots=5-shot, k=102025.10 | 63 | |
| ADMITVerifier LLM=o1-mini, Shots=5-shot, k=102025.10 | 59 | |
| ADMITVerifier LLM=LLaMA3.3-70B, Shots=1-shot, k=102025.10 | 58 | |
| ADMITVerifier LLM=o1-mini, Shots=1-shot, k=102025.10 | 50 | |
| PoisonedRAGVerifier LLM=LLaMA3.3-70B, Shots=5-shot, k=102025.10 | 45 | |
| ADMITVerifier LLM=GPT-4o, Shots=1-shot, k=102025.10 | 44 | |
| MisinfoVerifier LLM=LLaMA3.3-70B, Shots=5-shot, k=102025.10 | 40 | |
| PIAVerifier LLM=LLaMA3.3-70B, Shots=1-shot, k=102025.10 | 39 | |
| PoisonedRAGVerifier LLM=LLaMA3.3-70B, Shots=1-shot, k=102025.10 | 37 | |
| CorruptRAGVerifier LLM=LLaMA3.3-70B, Shots=1-shot, k=102025.10 | 30 | |
| MisinfoVerifier LLM=LLaMA3.3-70B, Shots=1-shot, k=102025.10 | 28 | |
| CorruptRAGVerifier LLM=LLaMA3.3-70B, Shots=5-shot, k=102025.10 | 26 | |
| PIAVerifier LLM=LLaMA3.3-70B, Shots=5-shot, k=102025.10 | 14 | |
| PIAVerifier LLM=GPT-4o, Shots=1-shot, k=102025.10 | 6 |