diff --git a/notebooks/sparse/quickstart.ipynb b/notebooks/sparse/quickstart.ipynb
index f061b0d6bfca..c9bf47cded6d 100644
--- a/notebooks/sparse/quickstart.ipynb
+++ b/notebooks/sparse/quickstart.ipynb
@@ -1208,15 +1208,8 @@
     {
       "cell_type": "markdown",
       "source": [
-        "## Exercise"
-      ],
-      "metadata": {
-        "id": "1iBNlJVYz3zi"
-      }
-    },
-    {
-      "cell_type": "markdown",
-      "source": [
+        "## Exercise \\#1\n",
+        "\n",
         "*Let's test what you've learned. Feel free to [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/dmlc/dgl/blob/master/notebooks/sparse/quickstart.ipynb).*\n",
         "\n",
         "Given a sparse symmetrical adjacency matrix $A$, calculate its symmetrically normalized adjacency matrix: $$norm = \\bar{D}^{-\\frac{1}{2}}\\bar{A}\\bar{D}^{-\\frac{1}{2}}$$\n",
@@ -1224,7 +1217,7 @@
         "Where $\\bar{A} = A + I$, $I$ is the identity matrix, and $\\bar{D}$ is the diagonal node degree matrix of $\\bar{A}$."
       ],
       "metadata": {
-        "id": "yDQ4Kmr_08St"
+        "id": "1iBNlJVYz3zi"
       }
     },
     {
@@ -1254,7 +1247,7 @@
     {
       "cell_type": "markdown",
       "source": [
-        "## Exercise\n",
+        "## Exercise \\#2\n",
         "\n",
         "Let's implement a simplified version of the Graph Attention Network (GAT) layer.\n",
         "\n",
@@ -1324,4 +1317,4 @@
       "outputs": []
     }
   ]
-}
+}
\ No newline at end of file