Safety Alignment on WildJailbreak
15,768.31Trainable parameters (M)Full FT
Evaluation Results
| Method | Links | |
|---|---|---|
| Full FTTarget Model=Qwen3-14B2026.02 | 15,768.31 | |
| Full FTTarget Model=Qwen2.5-14B2026.02 | 14,770.03 | |
| Full FTTarget Model=Phi-42026.02 | 14,660 | |
| Full FTTarget Model=gemma-7b-it2026.02 | 8,537.68 | |
| Full FTTarget Model=Average2026.02 | 7,616.48 | |
| Full FTTarget Model=Qwen2.5-7B2026.02 | 7,615.61 | |
| Full FTTarget Model=Qwen3-4B2026.02 | 4,022.46 | |
| Full FTTarget Model=Phi-4-mini2026.02 | 3,836 | |
| Full FTTarget Model=Llama-3.2-3B2026.02 | 3,212.75 | |
| Full FTTarget Model=gemma-2b-it2026.02 | 2,506.17 | |
| Full FTTarget Model=Llama-3.2-1B2026.02 | 1,235.81 | |
| Circuit BreakerTarget Model=Phi-42026.02 | 68.4 | |
| Circuit BreakerTarget Model=Qwen2.5-14B2026.02 | 51.6 | |
| Circuit BreakerTarget Model=gemma-7b-it2026.02 | 37.5 | |
| Circuit BreakerTarget Model=Qwen3-14B2026.02 | 31.52 | |
| Circuit BreakerTarget Model=Qwen2.5-7B2026.02 | 30.27 | |
| Circuit BreakerTarget Model=Average2026.02 | 29.28 | |
| Circuit BreakerTarget Model=Qwen3-4B2026.02 | 19.26 | |
| LoRATarget Model=Qwen2.5-14B2026.02 | 18.81 | |
| Circuit BreakerTarget Model=Llama-3.2-3B2026.02 | 18.23 | |
| Circuit BreakerTarget Model=gemma-2b-it2026.02 | 14.16 | |
| Circuit BreakerTarget Model=Phi-4-mini2026.02 | 13.36 | |
| Circuit BreakerTarget Model=Llama-3.2-1B2026.02 | 8.45 | |
| LoRATarget Model=gemma-2b-it2026.02 | 6.63 | |
| LoRATarget Model=Llama-3.2-3B2026.02 | 5.04 | |
| LoRATarget Model=Average2026.02 | 4.07 | |
| LoRATarget Model=Llama-3.2-1B2026.02 | 3.27 | |
| LoRATarget Model=Qwen3-14B2026.02 | 1.8 | |
| LoRATarget Model=Phi-4-mini2026.02 | 1.62 | |
| LoRATarget Model=gemma-7b-it2026.02 | 1.54 | |
| NeSTTarget Model=Qwen2.5-14B2026.02 | 1.01 | |
| LoRATarget Model=Qwen3-4B2026.02 | 0.89 | |
| NeSTTarget Model=Qwen3-14B2026.02 | 0.87 | |
| LoRATarget Model=Phi-42026.02 | 0.62 | |
| LoRATarget Model=Qwen2.5-7B2026.02 | 0.45 | |
| NeSTTarget Model=Qwen2.5-7B2026.02 | 0.45 | |
| NeSTTarget Model=Average2026.02 | 0.44 | |
| NeSTTarget Model=Phi-42026.02 | 0.42 | |
| NeSTTarget Model=Qwen3-4B2026.02 | 0.38 | |
| NeSTTarget Model=Llama-3.2-3B2026.02 | 0.36 | |
| NeSTTarget Model=gemma-7b-it2026.02 | 0.36 | |
| NeSTTarget Model=Phi-4-mini2026.02 | 0.21 | |
| NeSTTarget Model=gemma-2b-it2026.02 | 0.17 | |
| NeSTTarget Model=Llama-3.2-1B2026.02 | 0.13 |