Target behavior attack on MMLU bio
98.4Attack Success Rate (ASR)AiTM
Evaluation results
| Method | Links | |
|---|---|---|
| AiTMFramework=Camel, Communication Structure=Chain2025.02 | 98.4 | |
| AiTMFramework=Camel, Communication Structure=Complete2025.02 | 96.5 | |
| AiTMFramework=AutoGen, Communication Structure=Chain2025.02 | 96.3 | |
| AiTMFramework=Camel, Communication Structure=Random2025.02 | 96.3 | |
| AiTMFramework=AutoGen, Communication Structure=Complete2025.02 | 94.9 | |
| AiTMFramework=AutoGen, Communication Structure=Tree2025.02 | 93.7 | |
| AiTMFramework=Camel, Communication Structure=Tree2025.02 | 93.3 | |
| AiTMFramework=AutoGen, Communication Structure=Chain2025.02 | 93.1 | |
| AiTMFramework=AutoGen, Communication Structure=Random2025.02 | 89.2 | |
| AiTMFramework=Camel, Communication Structure=Chain2025.02 | 81.7 | |
| AiTMFramework=Camel, Communication Structure=Complete2025.02 | 68.2 | |
| AiTMFramework=Camel, Communication Structure=Random2025.02 | 52.4 | |
| AiTMFramework=AutoGen, Communication Structure=Random2025.02 | 51.5 | |
| AiTMFramework=Camel, Communication Structure=Tree2025.02 | 48.7 | |
| AiTMFramework=AutoGen, Communication Structure=Complete2025.02 | 43.9 | |
| AiTMFramework=AutoGen, Communication Structure=Tree2025.02 | 40.7 |