Safety Evaluation on Crescendo
12ASRClaude 3.5 Sonnet
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude 3.5 SonnetNBF steering=true2025.02 | 12 | |
| Claude 3.5 SonnetNBF steering=false2025.02 | 21.5 | |
| GPT-3.5-turboNBF steering=true2025.02 | 23.5 | |
| GPT-4oNBF steering=true2025.02 | 26 | |
| o1NBF steering=true2025.02 | 28 | |
| o1NBF steering=false2025.02 | 44.5 | |
| GPT-3.5-turboNBF steering=false2025.02 | 56 | |
| GPT-4oNBF steering=false2025.02 | 56.5 |