Refusal Suppression on JailbreakBench (test)
0Attack Success Rate (ASR)None (No Intervention)
Evaluation Results
| Method | Links | |
|---|---|---|
| None (No Intervention)Model=OSS, System Prompt=LW (Brief refusal directive)2026.06 | 0 | |
| None (No Intervention)Model=OSS, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 0 | |
| FixedSEModel=Mixtral, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 1 | |
| None (No Intervention)Model=Mixtral, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 6 | |
| FixedSEModel=Mixtral, System Prompt=LW (Brief refusal directive)2026.06 | 7 | |
| None (No Intervention)Model=Mixtral, System Prompt=LW (Brief refusal directive)2026.06 | 8 | |
| FixedAEModel=Mixtral, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 8 | |
| None (No Intervention)Model=OSS, System Prompt=N (No prompt)2026.06 | 10 | |
| FixedSEModel=OSS, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 13 | |
| SafetySEModel=Mixtral, System Prompt=LW (Brief refusal directive)2026.06 | 21 | |
| FixedAEModel=Mixtral, System Prompt=LW (Brief refusal directive)2026.06 | 24 | |
| SafetySEModel=OSS, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 28 | |
| FixedSEModel=OSS, System Prompt=LW (Brief refusal directive)2026.06 | 35 | |
| None (No Intervention)Model=OLMoE, System Prompt=LW (Brief refusal directive)2026.06 | 35 | |
| FixedAEModel=OSS, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 43 | |
| FixedSEModel=Mixtral, System Prompt=N (No prompt)2026.06 | 45 | |
| None (No Intervention)Model=Mixtral, System Prompt=N (No prompt)2026.06 | 46 | |
| FixedAEModel=OSS, System Prompt=LW (Brief refusal directive)2026.06 | 47 | |
| AllAEModel=OSS, System Prompt=LW (Brief refusal directive)2026.06 | 51 | |
| None (No Intervention)Model=OLMoE, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 52 | |
| AllAEModel=OSS, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 55 | |
| AllAEModel=Mixtral, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 56 | |
| FixedAEModel=OLMoE, System Prompt=LW (Brief refusal directive)2026.06 | 58 | |
| AllAEModel=OSS, System Prompt=N (No prompt)2026.06 | 59 | |
| AllAEModel=Mixtral, System Prompt=LW (Brief refusal directive)2026.06 | 59 | |
| FixedAEModel=Mixtral, System Prompt=N (No prompt)2026.06 | 63 | |
| AllAEModel=OLMoE, System Prompt=LW (Brief refusal directive)2026.06 | 63 | |
| None (No Intervention)Model=OLMoE, System Prompt=N (No prompt)2026.06 | 64 | |
| FixedSEModel=OSS, System Prompt=N (No prompt)2026.06 | 65 | |
| ActAddModel=OSS, System Prompt=LW (Brief refusal directive)2026.06 | 66 | |
| SafetySEModel=Mixtral, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 72 | |
| SafetySEModel=OSS, System Prompt=N (No prompt)2026.06 | 73 | |
| SafetySEModel=OSS, System Prompt=LW (Brief refusal directive)2026.06 | 74 | |
| SafetySEModel=OLMoE, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 75 | |
| ActAddModel=Mixtral, System Prompt=LW (Brief refusal directive)2026.06 | 77 | |
| ActAddModel=Mixtral, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 77 | |
| FixedAEModel=OLMoE, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 77 | |
| FixedAEModel=OLMoE, System Prompt=N (No prompt)2026.06 | 79 | |
| SafetySEModel=OLMoE, System Prompt=LW (Brief refusal directive)2026.06 | 79 | |
| FixedAEModel=OSS, System Prompt=N (No prompt)2026.06 | 82 | |
| ActAddModel=OSS, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 83 | |
| AllAEModel=OLMoE, System Prompt=N (No prompt)2026.06 | 83 | |
| FixedSEModel=OLMoE, System Prompt=N (No prompt)2026.06 | 87 | |
| AllAEModel=OLMoE, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 87 | |
| SafetySEModel=Mixtral, System Prompt=N (No prompt)2026.06 | 88 | |
| FixedSEModel=OLMoE, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 88 | |
| AllAEModel=Mixtral, System Prompt=N (No prompt)2026.06 | 90 | |
| ActAddModel=OLMoE, System Prompt=N (No prompt)2026.06 | 90 | |
| SafetySEModel=OLMoE, System Prompt=N (No prompt)2026.06 | 91 | |
| ActAddModel=Mixtral, System Prompt=N (No prompt)2026.06 | 92 | |
| ActAddModel=OLMoE, System Prompt=L2 (Complete chat-oriented prompt)2026.06 | 93 | |
| ActAddModel=OSS, System Prompt=N (No prompt)2026.06 | 94 | |
| ActAddModel=OLMoE, System Prompt=LW (Brief refusal directive)2026.06 | 94 | |
| FixedSEModel=OLMoE, System Prompt=LW (Brief refusal directive)2026.06 | 94 |