List your company

Jailbreak Attack on Prefilling Attack 20 tokens

0.3ASRSelf-Reminder
-2.378815.703133.78551.8669Feb 2, 2026

Evaluation results

MethodLinks
Self-ReminderModel=Llama2-7b2026.020.31.06
Self-ReminderBackbone=Qwen32026.020.911.07
Self-ExamBackbone=Qwen32026.020.911.07
SSDBackbone=Qwen32026.020.911.38
Self-ReminderModel=Llama2-13b2026.020.911.13
NGSDBackbone=Qwen32026.021.211.13
NGSDModel=Llama2-7b2026.021.521.38
NGSDModel=Llama2-13b2026.021.821.41
No DefenseBackbone=Qwen32026.022.731.15
SD_TinyExpertBackbone=Qwen32026.022.731.15
ParaphraseBackbone=Llama32026.023.031.94
NGSDBackbone=Llama32026.023.641.32
NGSDBackbone=Vicuna2026.023.941.5
Self-ReminderBackbone=Llama32026.024.241.56
SSDBackbone=Llama32026.024.851.72
Self-ExamModel=Llama2-7b2026.025.151.42
ParaphraseBackbone=Qwen32026.026.361.3
ParaphraseModel=Llama2-13b2026.026.671.62
SD_TinyExpertBackbone=Llama32026.027.271.6
SD_DaExpertModel=Llama2-7b2026.027.271.53
SD_DaExpertBackbone=Qwen32026.027.881.39
ParaphraseModel=Llama2-7b2026.028.481.71
Self-ExamModel=Llama2-13b2026.0212.121.8
SSDModel=Llama2-13b2026.0212.122.13
Self-ReminderBackbone=Vicuna2026.0212.731.98
SafeDecodingBackbone=Qwen32026.0213.331.53
SSDModel=Llama2-7b2026.0214.852.29
SD_TinyExpertModel=Llama2-13b2026.0218.792.74
SD_DaExpertModel=Llama2-13b2026.0221.522.77
ParaphraseBackbone=Vicuna2026.0223.642.75
SafeDecodingModel=Llama2-13b2026.0223.642.96
No DefenseModel=Llama2-13b2026.0223.942.82
SD_TinyExpertModel=Llama2-7b2026.0224.242.96
Self-ExamBackbone=Llama32026.0224.552.62
No DefenseModel=Llama2-7b2026.0224.853.14
Self-ExamBackbone=Vicuna2026.0227.272.59
SafeDecodingModel=Llama2-7b2026.0227.883.28
SSDBackbone=Vicuna2026.0232.123.05
SD_DaExpertBackbone=Llama32026.0244.553.8
SafeDecodingBackbone=Llama32026.0246.063.77
No DefenseBackbone=Llama32026.0249.393.99
SD_TinyExpertBackbone=Vicuna2026.0258.484.3
No DefenseBackbone=Vicuna2026.0267.274.48
SafeDecodingBackbone=Vicuna2026.0267.274.41
SD_DaExpertBackbone=Vicuna2026.0267.274.53