Jailbreak Defense on Decoding MaliciousInstruct
1ASRJPU
Evaluation results
| Method | Links | |
|---|---|---|
| JPUBase Model Architecture=Llama2-7B-Chat2026.01 | 1 | |
| JPUBase Model Architecture=Llama3-8B-Instruct2026.01 | 4 | |
| CKUBase Model Architecture=Llama3-8B-Instruct2026.01 | 6 | |
| EraserBase Model Architecture=Llama2-7B-Chat2026.01 | 7 | |
| CKUBase Model Architecture=Llama2-7B-Chat2026.01 | 7 | |
| RSFTBase Model Architecture=Llama3-8B-Instruct2026.01 | 7 | |
| Safe UnlearningBase Model Architecture=Llama3-8B-Instruct2026.01 | 7 | |
| Safe UnlearningBase Model Architecture=Llama2-7B-Chat2026.01 | 8 | |
| EraserBase Model Architecture=Llama3-8B-Instruct2026.01 | 8 | |
| Circuit BreakerBase Model Architecture=Llama3-8B-Instruct2026.01 | 8 | |
| RSFTBase Model Architecture=Llama2-7B-Chat2026.01 | 9 | |
| Circuit BreakerBase Model Architecture=Llama2-7B-Chat2026.01 | 10 | |
| Base modelBase Model Architecture=Llama3-8B-Instruct2026.01 | 17 | |
| Base modelBase Model Architecture=Llama2-7B-Chat2026.01 | 19 |