~

AAAI2021/paper.tex

@@ -179,7 +179,7 @@
   attracted extensive attention from a broad range of communities. Existing
   studies achieve high compositionality through \emph{deliberately handcrafted}
   inductions (e.g., addtional rewards, constructed 
-  loss functions and ease-of-teaching) in multi-agent learning, which are unnatural. 
+  loss functions, or ease-of-teaching) in multi-agent learning, which are unnatural. 
   Yet, few studies investigate the emergence of symbolic language with high
   compositionality \emph{naturally}, i.e., without deliberately handcrafted
   inductions. 

AAAI2021/tex/experiments.tex

@@ -91,7 +91,6 @@ the high compositionality has statistical significance related to agent
 capacity.
 
 
-
 %\subsection{Breakdown}
 %\label{ssec:language}
 

AAAI2021/tex/introduction.tex

@@ -48,9 +48,33 @@ vocabulary can express almost infinite concepts.}
   \label{fig:induction}
   \end{figure}
 
+
+
+\begin{table*}[t]
+  \centering
+  \small
+  \caption{Handcrafted inductions in related works.}
+  \label{tab:rel}
+  \begin{tabular}{llllll}
+    \toprule
+    Works & Handcrafted induction & Compositionality\\ 
+    \midrule
+    \cite{kirby2015compression}&Expressivity and compressibility&Qualitative, Speaker\\
+    \cite{kottur-etal-2017-natural}&Listener's memory&Qualitative, Speaker\\
+    \cite{choi2018compositional}&Maximum message length&Qualitative, Speaker+Listener\\
+    \cite{lazaridou2018emergence}&Structure of input data&Quantitative, Speaker\\
+    \cite{evtimova2018emergent}&Multi-modal scenarios&Quantitative, Speaker\\
+    \cite{li2019ease}&Population size, resetting all listeners&Quantitative, Speaker\\
+    \cite{chaabouni-etal-2019-word}&Word-order constraints&Qualitative, Speaker\\
+    \cite{chaabouni2020compositionality}&Easier to decode&Quantitative, Speaker\\
+    \textbf{Ours} & \textbf{None} & \textbf{Quantitative, Speaker+Listener} \\
+    \bottomrule
+    \end{tabular}
+  \end{table*}
+
 Prior studies focus on achieving high compositional symbolic language 
-through \emph{deliberately handcrafted} inductions, e.g., small vocabulary
-sizes~\cite{}, memoryless~\cite{}, addtional rewards~\cite{}, constructed loss functions~\cite{}, and
+through \emph{deliberately handcrafted} inductions, e.g., memoryless~\cite{},
+addtional rewards~\cite{}, constructed loss functions~\cite{}, and 
 ease-of-teaching~\cite{}. \note{Such optimization methodologies are driven by the challenges to generate high compositional symbolic without induction in existing multi-agent environment.}
 Figure~\ref{fig:induction} reports the compositionality when training two agents
 in the widely-used listener-speaker referential game for emerging 100 symbolic
@@ -180,6 +204,7 @@ can generate a higher compositional symbolic language with a higher probability.
 %%\endsection
 
 
+
 In this paper, we made the following contributions:
 \begin{itemize}[topsep=0pt,itemsep=0cm]
 \item To our best knowledge, we are the first work to successfully achieve 

AAAI2021/tex/relatedwork.tex

 \section{Related works}
 \label{sec:relatedwork}
 
-\begin{table*}[b]
-  \centering
-  \small
-  \caption{Handcrafted inductions in related works.}
-  \label{tab:rel}
-  \begin{tabular}{llllll}
-    \toprule
-    Works & Handcrafted induction & Compositionality\\ 
-    \midrule
-    \cite{kirby2015compression}&Expressivity and compressibility&Qualitative, Speaker\\
-    \cite{kottur-etal-2017-natural}&Listener's memory&Qualitative, Speaker\\
-    \cite{choi2018compositional}&Maximum message length&Qualitative, Speaker+Listener\\
-    \cite{lazaridou2018emergence}&Structure of input data&Quantitative, Speaker\\
-    \cite{evtimova2018emergent}&Multi-modal scenarios&Quantitative, Speaker\\
-    \cite{li2019ease}&Population size, resetting all listeners&Quantitative, Speaker\\
-    \cite{chaabouni-etal-2019-word}&Word-order constraints&Qualitative, Speaker\\
-    \cite{chaabouni2020compositionality}&Easier to decode&Quantitative, Speaker\\
-    \textbf{Ours} & \textbf{None} & \textbf{Quantitative, Speaker+Listener} \\
-    \bottomrule
-    \end{tabular}
-  \end{table*}
-
 
 %external environmental factors
 Previous works focus on the external environmental factors that impact the

AAAI2021/tex/theory.tex

-\begin{figure}[h]
+
+\section{ Symbolic Language Producing}
+\label{sec:thory}
+
+
+\begin{figure}[t]
   \centering  \includegraphics[width=\columnwidth]{fig/Figure2_The_referential_game_environment.pdf}
   \caption{The referential game in this paper.}
   \label{fig:game}
 \end{figure}
 
+\begin{figure*}[t]
+  \centering
+  \includegraphics[width=1.8\columnwidth]{fig/Figure3_The_architecture_of_agents.pdf}
+  \caption{The architecture of agents. \emph{Left:} speaker. \emph{Right:} listener.}
+  \label{fig:agents}
+\end{figure*}
 
-\section{ Symbolic Language Producing }
-\label{sec:thory}
 
 Before going to the detail of the training algorithms, we first introduce the environment, gaming rules, and agent architecture for enabling the emergence of symbolic language. 
 
@@ -43,12 +52,6 @@ Please note that since $t$ and $\hat{t}$ have different length, we say
 $t=\hat{t}$ if $t$ expresses the same meaning as $\hat{t}$, e.g., ``red circle''. 
 
 
-\begin{figure*}[t]
-  \centering
-  \includegraphics[width=1.8\columnwidth]{fig/Figure3_The_architecture_of_agents.pdf}
-  \caption{The architecture of agents. \emph{Left:} speaker. \emph{Right:} listener.}
-  \label{fig:agents}
-\end{figure*}