LLM Safety and Utility Evaluation Suite (StrongReject, AdvBench, IFeval, MMLU) (test)
99.8Safety Scorelat
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| latModel=llama-3.1-8b-it2026.02 | 99.8 | 35.9 | 31.6 | 17.4 | 58.7 | |
| CC-DeltaModel=gemma-2-2b-it, steering_strength=1002026.02 | 99.6 | 32.4 | 19.8 | 25.5 | 51.8 | |
| linearactModel=qwen-2.5-7b-it, steering_strength=1002026.02 | 99.4 | 14.7 | 2.7 | 17.1 | 24.2 | |
| caaModel=gemma-2-2b-it, steering_strength=1002026.02 | 99.3 | 18.1 | 6.1 | 25.2 | 22.9 | |
| caaModel=gemma-2-9b-it, steering_strength=1002026.02 | 99.3 | 23.6 | 20.3 | 27.6 | 22.9 | |
| latModel=gemma-2-9b-it2026.02 | 99.3 | 60.8 | 86.1 | 25.7 | 70.5 | |
| CC-DeltaModel=llama-3.1-8b-it, steering_strength=1002026.02 | 99.3 | 31.2 | 0 | 29.4 | 64.4 | |
| CC-DeltaModel=qwen-2.5-7b-it, steering_strength=1002026.02 | 99.2 | 36.5 | 6.1 | 30 | 73.4 | |
| CC-DeltaModel=gemma-2-9b-it, steering_strength=1002026.02 | 98.9 | 21.9 | 0 | 24.3 | 41.5 | |
| linearactModel=llama-3.1-8b-it, steering_strength=1002026.02 | 98.8 | 17.4 | 0.3 | 26.5 | 25.3 | |
| linearactModel=gemma-2-9b-it, steering_strength=1002026.02 | 98.6 | 15.5 | 1.8 | 20 | 24.6 | |
| linearactModel=gemma-2-2b-it, steering_strength=1002026.02 | 97.6 | 24.2 | 27.8 | 21.7 | 23.1 | |
| caaModel=llama-3.1-8b-it, steering_strength=1002026.02 | 97.6 | 34.9 | 2.4 | 36.3 | 66.1 | |
| caaModel=qwen-2.5-7b-it, steering_strength=1002026.02 | 97.5 | 35.7 | 12.9 | 30.3 | 63.9 | |
| latModel=gemma-2-2b-it2026.02 | 91.3 | 65.5 | 95.2 | 44.6 | 56.7 | |
| cbModel=gemma-2-9b-it2026.02 | 87.4 | 36.8 | 3.3 | 35 | 72 | |
| cbModel=qwen-2.5-7b-it2026.02 | 86.2 | 50.7 | 1.9 | 77 | 73.1 | |
| cbModel=gemma-2-2b-it2026.02 | 83.2 | 34.6 | 0.9 | 46.3 | 56.5 | |
| latModel=qwen-2.5-7b-it2026.02 | 79.5 | 83.1 | 100 | 76.5 | 72.7 | |
| self_reminderModel=llama-3.1-8b-it2026.02 | 77.3 | 84.2 | 100 | 84.8 | 67.9 | |
| cbModel=llama-3.1-8b-it2026.02 | 74.3 | 84.4 | 99.4 | 85.3 | 68.4 | |
| CC-DeltaModel=llama-3.1-8b-it, steering_strength=102026.02 | 74.1 | 80.3 | 95.9 | 77.7 | 67.4 | |
| CC-DeltaModel=gemma-2-9b-it, steering_strength=102026.02 | 70.5 | 80.5 | 99.9 | 71.9 | 69.6 | |
| CC-DeltaModel=qwen-2.5-7b-it, steering_strength=102026.02 | 70.4 | 79.5 | 92 | 72.8 | 73.7 | |
| caaModel=gemma-2-9b-it, steering_strength=102026.02 | 70.1 | 80.7 | 100 | 73.1 | 69.1 | |
| linearactModel=qwen-2.5-7b-it, steering_strength=102026.02 | 68.8 | 81.6 | 99.5 | 74 | 71.3 | |
| originalModel=llama-3.1-8b-it2026.02 | 67.5 | 84.2 | 99.8 | 84.5 | 68.2 | |
| caaModel=llama-3.1-8b-it, steering_strength=102026.02 | 67.4 | 84.2 | 99.9 | 84.4 | 68.2 | |
| self_reminderModel=qwen-2.5-7b-it2026.02 | 64.1 | 85.1 | 100 | 82 | 73.2 | |
| self_reminderModel=gemma-2-9b-it2026.02 | 63.8 | 84.2 | 100 | 80.6 | 72 | |
| CC-DeltaModel=gemma-2-2b-it, steering_strength=102026.02 | 63.3 | 70.7 | 100 | 56.4 | 55.8 | |
| linearactModel=gemma-2-9b-it, steering_strength=102026.02 | 60.9 | 82.5 | 100 | 76.7 | 70.7 | |
| caaModel=gemma-2-2b-it, steering_strength=102026.02 | 60.4 | 70.5 | 99.9 | 55.9 | 55.5 | |
| self_reminderModel=gemma-2-2b-it2026.02 | 60.1 | 74.8 | 100 | 67.7 | 56.5 | |
| caaModel=qwen-2.5-7b-it, steering_strength=102026.02 | 57.3 | 80.8 | 97.3 | 71.5 | 73.6 | |
| linearactModel=gemma-2-2b-it, steering_strength=102026.02 | 56.2 | 72.5 | 100 | 60.6 | 56.9 | |
| originalModel=gemma-2-9b-it2026.02 | 56.2 | 84.4 | 100 | 81.1 | 72.3 | |
| originalModel=gemma-2-2b-it2026.02 | 53.9 | 74.2 | 100 | 65.7 | 56.8 | |
| originalModel=qwen-2.5-7b-it2026.02 | 53.3 | 84.7 | 99.8 | 80.2 | 74.3 |