\documentclass[letterpaper]{article}
\usepackage{graphicx}
\usepackage{booktabs}
\begin{document}
\s...
\documentclass[letterpaper]{article}
\usepackage{graphicx}
\usepackage{booktabs}
\begin{document}
\section{Section1}
\section{Section2}
\subsection{Subsection}
\begin{table}[]
\caption{N-gram}
\label{tab:ngram}
\begin{tabular}{@{}llllllll@{}}
\toprule
Method & B-1 & B-2 & B-3 & B-4 & M & R & C \\ \midrule
Baseline & 70.3 & 57.2 & 43.5 & 33.1 & 25.2 & 54.3 & 107.2 \\
B & 71.8 & 59.5 & 45.0 & 35.8 & 26.7 & 57.3 & 110.2 \\
M & 75.6 & 58.3 & 32.4 & 16.2 & 27.2 & 57.1 & 20.4 \\
R & 58.2 & 38.3 & 22.7 & 4.5 & 21.3 & 58.7 & 18.7 \\
C & 73.2 & 58.3 & 45.6 & 32.4 & 25.8 & 53.3 & 108.3 \\
\textbf{Ours} & 75.3 & 60.2 & 47.9 & 36.2 & 26.5 & 57.1 & 112.1 \\ \bottomrule
\end{tabular}
\end{table}
Firstly, We compare the experimental results of our reward model with using automatic metrics as reword in Table \ref{tab:ngram}. It can be seen that our model can defeat all models with automatic evaluation metrics as reward, giving evidence of implicit learning metrics performing better then automatic metrics.
\end{document}
一周热门 更多>