Attack Injection on SeedHijack Attack Evaluation
180Attack SuccessesSeedHijack
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SeedHijackModel=GPT-2, Parameters=124M, Alignment=None, Trials=1802026.05 | 180 | 100 | |
| SeedHijackModel=Qwen2-1.5B Instruct, Parameters=1.5B, Alignment=RLHF + SFT, Trials=1802026.05 | 180 | 100 | |
| SeedHijackModel=DeepSeek-R1 1.5B, Parameters=1.5B, Alignment=Reasoning distill., Trials=1802026.05 | 180 | 100 | |
| SeedHijackModel=Qwen2-7B Instruct, Parameters=7B, Alignment=RLHF + SFT, Trials=1802026.05 | 180 | 100 | |
| SeedHijackModel=DeepSeek-R1 7B, Parameters=7B, Alignment=Reasoning distill., Trials=1802026.05 | 180 | 100 |