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

Sycophancy Assessment on BASIL 1.0 (Under-Update)

-0.355Change in Bayesian Error (RMSE)

mistral:7b

Updated 1mo ago

Evaluation Results

Method	Links
mistral:7b 2025.08		-0.355
llama3.2:1b 2025.08		-0.329
mistral:7b 2025.08		-0.271
llama3.2:1b+DPO 2025.08		-0.234
mistral:7b 2025.08		-0.23
llama3.2:1b 2025.08		-0.2
llama3.2:3b 2025.08		-0.188
gpt-4o-mini 2025.08		-0.171
phi4:14b 2025.08		-0.168
mistral:7b 2025.08		-0.152
llama3.2:1b 2025.08		-0.151
llama3.2:1b 2025.08		-0.146
llama3.2:1b+DPO 2025.08		-0.137
llama3.2:1b+DPO 2025.08		-0.137
phi4:14b 2025.08		-0.136
gpt-4o-mini 2025.08		-0.132
llama3.2:1b+SFT 2025.08		-0.129
llama3.2:1b+DPO 2025.08		-0.129
llama3.2:1b+SFT 2025.08		-0.116
llama3.2:3b 2025.08		-0.115
llama3.2:3b 2025.08		-0.107
gpt-4o-mini 2025.08		-0.104
claude-haiku-4-5 2025.08		-0.103
claude-haiku-4-5 2025.08		-0.091
llama3.2:1b+SFT 2025.08		-0.09
llama3.2:3b 2025.08		-0.087
gpt-4o-mini 2025.08		-0.082
claude-haiku-4-5 2025.08		-0.08
claude-haiku-4-5 2025.08		-0.079
phi4:14b 2025.08		-0.048
phi4:14b 2025.08		-0.041
llama3.2:1b+SFT 2025.08		-0.032