List your company

Safety Evaluation on Crescendo

12ASRClaude 3.5 Sonnet
10.2222.23534.2546.265Feb 28, 2025

Evaluation results

MethodLinks
Claude 3.5 SonnetNBF steering=true2025.0212
Claude 3.5 SonnetNBF steering=false2025.0221.5
GPT-3.5-turboNBF steering=true2025.0223.5
GPT-4oNBF steering=true2025.0226
o1NBF steering=true2025.0228
o1NBF steering=false2025.0244.5
GPT-3.5-turboNBF steering=false2025.0256
GPT-4oNBF steering=false2025.0256.5