Share your thoughts, 1 month free Claude Pro on usSee more

Sycophancy Assessment on BASIL Over-Update 1.0

0.016Change in Bayesian Error (RMSE)

claude-haiku-4-5

Updated 1mo ago

Evaluation Results

Method	Links
claude-haiku-4-5 2025.08		0.016
claude-haiku-4-5 2025.08		0.021
llama3.2:1b 2025.08		0.024
mistral:7b 2025.08		0.025
claude-haiku-4-5 2025.08		0.032
phi4:14b 2025.08		0.037
phi4:14b 2025.08		0.041
phi4:14b 2025.08		0.052
llama3.2:1b+DPO 2025.08		0.061
phi4:14b 2025.08		0.068
llama3.2:1b+DPO 2025.08		0.068
llama3.2:1b 2025.08		0.072
llama3.2:1b+DPO 2025.08		0.077
llama3.2:1b 2025.08		0.078
llama3.2:1b+SFT 2025.08		0.079
mistral:7b 2025.08		0.081
gpt-4o-mini 2025.08		0.086
claude-haiku-4-5 2025.08		0.087
gpt-4o-mini 2025.08		0.094
gpt-4o-mini 2025.08		0.097
llama3.2:3b 2025.08		0.108
llama3.2:1b+DPO 2025.08		0.112
llama3.2:1b+SFT 2025.08		0.124
gpt-4o-mini 2025.08		0.132
llama3.2:1b 2025.08		0.135
llama3.2:3b 2025.08		0.14
mistral:7b 2025.08		0.142
llama3.2:1b+SFT 2025.08		0.147
mistral:7b 2025.08		0.191
llama3.2:3b 2025.08		0.213
llama3.2:1b+SFT 2025.08		0.216
llama3.2:3b 2025.08		0.257