Safety Evaluation on MultiJail (test)
0MultiJail Safety Failure Rate (EN)LASA
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| LASABase Model=Llama-3.1-8B-Instruct2026.04 | 0 | 0 | 1 | 0 | 8 | 5 | 1.7 | |
| SFTBase Model=Qwen-2.5-7B-Instruct2026.04 | 0 | 1 | 0 | 0 | 51 | 13 | 7.4 | |
| KTOBase Model=Qwen-2.5-7B-Instruct2026.04 | 0 | 0 | 1 | 1 | 57 | 11 | 7.8 | |
| ORPOBase Model=Qwen-2.5-7B-Instruct2026.04 | 0 | 2 | 1 | 1 | 45 | 12 | 6.4 | |
| LASABase Model=Qwen-2.5-7B-Instruct2026.04 | 0 | 0 | 1 | 1 | 13 | 5 | 2.5 | |
| SFTBase Model=Llama-3.1-8B-Instruct2026.04 | 1 | 2 | 2 | 2 | 38 | 16 | 7.3 | |
| DPOBase Model=Llama-3.1-8B-Instruct2026.04 | 1 | 4 | 8 | 3 | 19 | 15 | 6.6 | |
| ORPOBase Model=Llama-3.1-8B-Instruct2026.04 | 1 | 0 | 2 | 0 | 28 | 13 | 5.1 | |
| MPOBase Model=Llama-3.1-8B-Instruct2026.04 | 1 | 1 | 3 | 2 | 28 | 14 | 5.3 | |
| KTOBase Model=Llama-3.1-8B-Instruct2026.04 | 1.01 | 1 | 1 | 1 | 19 | 9 | 3.4 | |
| DPOBase Model=Qwen-2.5-7B-Instruct2026.04 | 2 | 0 | 1 | 2 | 47 | 15 | 8.21 | |
| CPOBase Model=Qwen-2.5-7B-Instruct2026.04 | 2 | 1 | 4 | 2 | 44 | 19 | 9 | |
| MPOBase Model=Qwen-2.5-7B-Instruct2026.04 | 2 | 0 | 2 | 2 | 46 | 16 | 8.1 | |
| CPOBase Model=Llama-3.1-8B-Instruct2026.04 | 3.03 | 1 | 3 | 1 | 32 | 17 | 7.3 | |
| Vanilla ModelBase Model=Qwen-2.5-7B-Instruct2026.04 | 4 | 3 | 5 | 3 | 56 | 27 | 12.5 | |
| Vanilla ModelBase Model=Llama-3.1-8B-Instruct2026.04 | 13 | 13 | 37 | 17 | 46 | 39 | 21 |