Jailbreak on Sorry
99.8Jailbreak RateAbliteration
Evaluation Results
| Method | Links | |
|---|---|---|
| AbliterationBackbone=Gemma3-12B2026.02 | 99.8 | |
| AbliterationBackbone=Qwen3-8B2026.02 | 99.5 | |
| AbliterationBackbone=Qwen3-14B2026.02 | 99.1 | |
| AbliterationBackbone=Qwen2.5-14B2026.02 | 98.6 | |
| GRP-OblitBackbone=Gemma2-9B2026.02 | 98.2 | |
| GRP-Oblit-1Backbone=Ministral3-8B Instruct2026.02 | 98.2 | |
| GRP-OblitBackbone=Ministral3-8B Instruct2026.02 | 98 | |
| AbliterationBackbone=DeepSeek-R1-Qwen-14B2026.02 | 97.7 | |
| TwinBreakBackbone=Qwen3-14B2026.02 | 97.7 | |
| GRP-OblitBackbone=Qwen3-8B2026.02 | 97.7 | |
| GRP-Oblit-1Backbone=Llama3.1-8B2026.02 | 97.5 | |
| GRP-OblitBackbone=Ministral3-14B Reasoning2026.02 | 97.3 | |
| GRP-Oblit-1Backbone=DeepSeek-R1-Llama-8B2026.02 | 96.8 | |
| TwinBreakBackbone=DeepSeek-R1-Qwen-7B2026.02 | 96.8 | |
| GRP-Oblit-1Backbone=Qwen3-14B2026.02 | 96.8 | |
| GRP-OblitBackbone=DeepSeek-R1-Llama-8B2026.02 | 96.6 | |
| GRP-Oblit-1Backbone=DeepSeek-R1-Qwen-14B2026.02 | 96.6 | |
| TwinBreakBackbone=DeepSeek-R1-Llama-8B2026.02 | 96.4 | |
| AbliterationBackbone=DeepSeek-R1-Llama-8B2026.02 | 96.4 | |
| GRP-OblitBackbone=Llama3.1-8B2026.02 | 96.4 | |
| GRP-Oblit-1Backbone=DeepSeek-R1-Qwen-7B2026.02 | 95.9 | |
| GRP-Oblit-1Backbone=Qwen3-8B2026.02 | 95.9 | |
| TwinBreakBackbone=Qwen2.5-14B2026.02 | 95.5 | |
| TwinBreakBackbone=Qwen2.5-7B2026.02 | 95.2 | |
| GRP-OblitBackbone=Qwen2.5-7B2026.02 | 95 | |
| AbliterationBackbone=Qwen2.5-7B2026.02 | 93.9 | |
| GRP-OblitBackbone=DeepSeek-R1-Qwen-14B2026.02 | 93.2 | |
| GRP-OblitBackbone=Ministral3-14B Instruct2026.02 | 93.2 | |
| GRP-Oblit-1Backbone=GPT-OSS-20B2026.02 | 93 | |
| TwinBreakBackbone=Llama3.1-8B2026.02 | 92.3 | |
| GRP-OblitBackbone=DeepSeek-R1-Qwen-7B2026.02 | 91.6 | |
| GRP-Oblit-1Backbone=Qwen2.5-7B2026.02 | 91.6 | |
| GRP-OblitBackbone=GPT-OSS-20B2026.02 | 91.4 | |
| GRP-Oblit-1Backbone=Ministral3-14B Reasoning2026.02 | 90.9 | |
| GRP-OblitBackbone=Gemma3-12B2026.02 | 90 | |
| GRP-Oblit-1Backbone=Qwen2.5-14B2026.02 | 88.9 | |
| GRP-OblitBackbone=Ministral3-8B Reasoning2026.02 | 88.6 | |
| GRP-Oblit-1Backbone=Ministral3-14B Instruct2026.02 | 88.4 | |
| GRP-Oblit-1Backbone=Ministral3-8B Reasoning2026.02 | 88.2 | |
| AbliterationBackbone=DeepSeek-R1-Qwen-7B2026.02 | 88 | |
| TwinBreakBackbone=Gemma2-9B2026.02 | 87.7 | |
| GRP-OblitBackbone=Qwen2.5-14B2026.02 | 87.7 | |
| GRP-OblitBackbone=Qwen3-14B2026.02 | 85.9 | |
| GRP-Oblit-1Backbone=Gemma2-9B2026.02 | 83.9 | |
| AbliterationBackbone=Gemma2-9B2026.02 | 83.4 | |
| TwinBreakBackbone=DeepSeek-R1-Qwen-14B2026.02 | 81.4 | |
| BaseBackbone=DeepSeek-R1-Qwen-7B2026.02 | 77.7 | |
| GRP-Oblit-1Backbone=Gemma3-12B2026.02 | 76.4 | |
| BaseBackbone=Ministral3-14B Reasoning2026.02 | 75.9 | |
| AbliterationBackbone=Llama3.1-8B2026.02 | 71.1 | |
| AbliterationBackbone=GPT-OSS-20B2026.02 | 70 | |
| BaseBackbone=DeepSeek-R1-Llama-8B2026.02 | 69.8 | |
| AbliterationBackbone=Ministral3-8B Reasoning2026.02 | 62.7 | |
| BaseBackbone=DeepSeek-R1-Qwen-14B2026.02 | 56.8 | |
| BaseBackbone=Qwen3-8B2026.02 | 50.7 | |
| BaseBackbone=Ministral3-8B Reasoning2026.02 | 50.5 | |
| BaseBackbone=Qwen3-14B2026.02 | 47 | |
| TwinBreakBackbone=Gemma3-12B2026.02 | 43.9 | |
| BaseBackbone=Gemma3-12B2026.02 | 37.3 | |
| BaseBackbone=Qwen2.5-7B2026.02 | 37.3 | |
| AbliterationBackbone=Ministral3-8B Instruct2026.02 | 31.4 | |
| BaseBackbone=Qwen2.5-14B2026.02 | 31.1 | |
| BaseBackbone=Ministral3-8B Instruct2026.02 | 30.5 | |
| BaseBackbone=Ministral3-14B Instruct2026.02 | 30.2 | |
| BaseBackbone=Llama3.1-8B2026.02 | 26.4 | |
| BaseBackbone=GPT-OSS-20B2026.02 | 13.2 | |
| BaseBackbone=Gemma2-9B2026.02 | 11.4 | |
| TwinBreakBackbone=Qwen3-8B2026.02 | 1.1 | |
| AbliterationBackbone=Ministral3-14B Reasoning2026.02 | 0.7 | |
| AbliterationBackbone=Ministral3-14B Instruct2026.02 | 0 |