Safety Evaluation on Opposite-day
4.5ASRClaude 3.5 Sonnet
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude 3.5 SonnetNBF steering=true2025.02 | 4.5 | |
| Claude 3.5 SonnetNBF steering=false2025.02 | 9.5 | |
| o1NBF steering=true2025.02 | 21 | |
| GPT-4oNBF steering=true2025.02 | 32.5 | |
| GPT-3.5-turboNBF steering=true2025.02 | 37.5 | |
| o1NBF steering=false2025.02 | 53 | |
| GPT-4oNBF steering=false2025.02 | 72.5 | |
| GPT-3.5-turboNBF steering=false2025.02 | 78.5 |