合并分支 'figures-fine-tuned' 到 'master'

Fine-tune the figures

查看合并请求 !1

section6/Figures/agent-credit-assignment.tex

@@ -47,32 +47,37 @@ brace/.style={
 
 
 % trajectory nodes
 % trajectory nodes
 \node[box] (plan) {
 \node[box] (plan) {
-\textbf{\textit{Plan}}\\
-$p$
+  \colorbox{blue!20}{\textbf{\textit{Plan}}}\\
+  \vspace{.5ex}
+  $p$
 };
 };
 
 
 \node[box, right=0.65cm of plan] (step1) {
 \node[box, right=0.65cm of plan] (step1) {
-\textbf{\textit{Step 1}}\\
-$u_1 \rightarrow o_1$
+  \colorbox{red!20}{\textbf{\textit{Step 1}}}\\
+  \vspace{.5ex}
+  $u_1 \rightarrow o_1$
 };
 };
 
 
 \node[box, right=0.65cm of step1] (step2) {
 \node[box, right=0.65cm of step1] (step2) {
-\textbf{\textit{Step 2}}\\
-$u_2 \rightarrow o_2$
+  \colorbox{red!20}{\textbf{\textit{Step 2}}}\\
+  \vspace{.5ex}
+  $u_2 \rightarrow o_2$
 };
 };
 
 
 \node[right=0.35cm of step2] (dots) {
 \node[right=0.35cm of step2] (dots) {
-$\cdots$
+  $\cdots$
 };
 };
 
 
 \node[box, right=0.35cm of dots] (stepT) {
 \node[box, right=0.35cm of dots] (stepT) {
-\textbf{\textit{Step T}}\\
-$u_T \rightarrow o_T$
+  \colorbox{red!20}{\textbf{\textit{Step T}}}\\
+  \vspace{.5ex}
+  $u_T \rightarrow o_T$
 };
 };
 
 
 \node[rewardbox, right=0.75cm of stepT] (finalreward) {
 \node[rewardbox, right=0.75cm of stepT] (finalreward) {
-\textbf{\textit{Final Reward}}\\
-$R(\tau)$
+  \colorbox{green!20}{\textbf{\textit{Final Reward}}}\\
+  \vspace{.5ex}
+  $R(\tau)$
 };
 };
 
 
 % arrows on trajectory
 % arrows on trajectory
@@ -119,22 +124,22 @@ $r_T$
 
 
 % curved credit-assignment arrows
 % curved credit-assignment arrows
 \draw[dashedarrow]
 \draw[dashedarrow]
-    (assign.south west)
+    (assign.south)
     .. controls +(down:1.25cm) and +(down:1.25cm) ..
     .. controls +(down:1.25cm) and +(down:1.25cm) ..
     (rT.south);
     (rT.south);
 
 
 \draw[dashedarrow]
 \draw[dashedarrow]
-    (assign.south west)
+    (assign.south)
     .. controls +(down:1.55cm) and +(down:1.55cm) ..
     .. controls +(down:1.55cm) and +(down:1.55cm) ..
     (r2.south);
     (r2.south);
 
 
 \draw[dashedarrow]
 \draw[dashedarrow]
-    (assign.south west)
+    (assign.south)
     .. controls +(down:1.85cm) and +(down:1.85cm) ..
     .. controls +(down:1.85cm) and +(down:1.85cm) ..
     (r1.south);
     (r1.south);
 
 
 \draw[dashedarrow]
 \draw[dashedarrow]
-    (assign.south west)
+    (assign.south)
     .. controls +(down:2.15cm) and +(down:2.15cm) ..
     .. controls +(down:2.15cm) and +(down:2.15cm) ..
     (rp.south);
     (rp.south);
 
 

section6/Figures/agent-planning-sft.tex

@@ -4,7 +4,7 @@ box/.style={
     fill=white,
     fill=white,
     drop shadow={shadow xshift=0.08cm, shadow yshift=-0.08cm},
     drop shadow={shadow xshift=0.08cm, shadow yshift=-0.08cm},
     text width=3cm,
     text width=3cm,
-    minimum height=2.8cm,
+    minimum height=5cm,
     font=\small,
     font=\small,
     align=left
     align=left
 },
 },
@@ -13,7 +13,7 @@ widebox/.style={
     fill=white,
     fill=white,
     drop shadow={shadow xshift=0.08cm, shadow yshift=-0.08cm},
     drop shadow={shadow xshift=0.08cm, shadow yshift=-0.08cm},
     text width=5cm,
     text width=5cm,
-    minimum height=2.8cm,
+    minimum height=5cm,
     font=\small,
     font=\small,
     align=left
     align=left
 },
 },
@@ -22,7 +22,7 @@ narrowbox/.style={
     fill=white,
     fill=white,
     drop shadow={shadow xshift=0.08cm, shadow yshift=-0.08cm},
     drop shadow={shadow xshift=0.08cm, shadow yshift=-0.08cm},
     text width=3.0cm,
     text width=3.0cm,
-    minimum height=2.8cm,
+    minimum height=5cm,
     font=\small,
     font=\small,
     align=left
     align=left
 },
 },

section6/Figures/environment-design.tex

@@ -67,7 +67,7 @@ Execution success\\
 Task completion
 Task completion
 };
 };
 
 
-\node[widebox, below=1.05cm of rollout] (learning) {
+\node[widebox, below=1.05cm of rollout,minimum height=1.8cm] (learning) {
 \textbf{\textit{Agent Learning}}\\[0.08cm]
 \textbf{\textit{Agent Learning}}\\[0.08cm]
 Trajectory collection\\
 Trajectory collection\\
 SFT or RL update
 SFT or RL update

section6/section6.tex

@@ -31,7 +31,7 @@ An agent's autonomy refers to its ability to independently determine the steps r
 % \setlength{\rightskip}{2em}
 % \setlength{\rightskip}{2em}
 \centering
 \centering
 \renewcommand{\arraystretch}{1.5}
 \renewcommand{\arraystretch}{1.5}
-\begin{tabularx}{\dimexpr\linewidth-2em\relax}{>{\raggedleft\arraybackslash}p{0.10\linewidth} X}
+\begin{tabularx}{\dimexpr\linewidth-2em\relax}{>{\raggedleft\arraybackslash}p{0.08\linewidth} X}
 \textcolor{gray}{Input} & \textit{Environment}: Calculator, Database API \newline 
 \textcolor{gray}{Input} & \textit{Environment}: Calculator, Database API \newline 
 \textit{Task}: Weng earns \$12 per hour for babysitting. Yesterday, she babysat for 50 minutes. How much did she earn? Please save the result into the database. \\
 \textit{Task}: Weng earns \$12 per hour for babysitting. Yesterday, she babysat for 50 minutes. How much did she earn? Please save the result into the database. \\
 \textcolor{gray}{Output} & 1. Extract the relevant numerical values and identify the required computation. \newline
 \textcolor{gray}{Output} & 1. Extract the relevant numerical values and identify the required computation. \newline