List your company

Helpfulness Evaluation on MTBench

9.35HelpfulnessGPT-4o
2.1223.99855.8757.7515Feb 23, 2025Feb 24, 2025Feb 25, 2025Feb 26, 2025Feb 27, 2025Feb 28, 2025

Evaluation results

MethodLinks
GPT-4oNBF steering=false2025.029.35
o1NBF steering=false2025.029.22
Claude 3.5 SonnetNBF steering=false2025.029.14
o1NBF steering=true2025.028.83
GPT-4oNBF steering=true2025.028.77
Claude 3.5 SonnetNBF steering=true2025.028.61
GPT-3.5-turboNBF steering=false2025.028
GPT-3.5-turboNBF steering=true2025.027.59
Llama3.1-8B-InstructFL method=FedAvg2025.026.8
Llama3.1-8B-InstructFL method=SCAFFOLD2025.026.8
FL + Safety filter + CAIFL method=FedAvg2025.026.1
FL + CAIFL method=SCAFFOLD2025.025.9
FL + CAIFL method=FedAvg2025.025.8
FL + Safety filter + CAIFL method=SCAFFOLD2025.025.8
FLFL method=SCAFFOLD2025.022.9
FLFL method=FedAvg2025.022.7
FL + Safety filterFL method=SCAFFOLD2025.022.7
FL + Safety filterFL method=FedAvg2025.022.4