evals/mjbench/latex_reults/original_scale_study.tex

\begin{table}[t]
    \centering
    \caption{Result with different scale.}
    \resizebox{1.0\linewidth}{!}{%
    \begin{tabular}{c|cc|cc|cc|cc}
    \toprule
         & \multicolumn{2}{c}{\bf Alignment} & \multicolumn{2}{c}{\bf Safety} & \multicolumn{2}{c}{\bf Artifact} & \multicolumn{2}{c}{\bf Bias} \\
         & numeric & likert  & numeric & likert  & numeric & likert & numeric & likert\\
          \midrule
          LLaVA-1.5-7b$^\heartsuit$ & - & - & - & - & - & - & - & - \\
          LLaVA-1.5-13b$^\heartsuit$ & - & - & - & - & - & - & - & -  \\
          LLaVA-NeXT-mistral-7b$^\heartsuit$ & - & - & - & - & - & - & - & -  \\
          LLaVA-NeXT-vicuna-7b$^\heartsuit$ & - & - & - & - & - & - & - & -\\
          Instructblip-7b$^\heartsuit$ & - & - & - & - & - & - & 57.4 & 85.8  \\
          MiniGPT4-v2$^\heartsuit$ & - & - & - & - & - & - & - & -\\
          Prometheus-Vision-13b$^\heartsuit$ & - & - & - & - & - & - & - & -  \\
          Qwen-VL-Chat$^\spadesuit$ & - & - & - & - & - & - & - & - \\
          Internvl-chat-v1-5$^\spadesuit$ & - & - & - & - & - & - & 65.3 & 83.5  \\
          Idefics2-8b$^\spadesuit$ & - & - & - & - & - & - & 52.7 & 77.6 \\
          \midrule
          GPT-4-vision$^\clubsuit$ & - & - & - & - & - & - & 80.4 & 93.2  \\
          GPT-4o$^\clubsuit$ & - & - & - & - & - & - & 82.5 & 92.8 \\
          Gemini Ultra$^\clubsuit$ & - & - & - & - & - & - & 75.3 & 88.6 \\
          Claude 3 Opus$^\clubsuit$ & - & - & - & - & - & - & 65.6 & 85.0  \\
    \bottomrule
    \end{tabular}%
    }
    \label{exp:numeric_likert}
\end{table}