diff --git a/docs/Brief-Introduction/Usage.rst b/docs/Brief-Introduction/Usage.rst
index 5067980..fd6e76c 100644
--- a/docs/Brief-Introduction/Usage.rst
+++ b/docs/Brief-Introduction/Usage.rst
@@ -42,7 +42,7 @@ In the MNIST Add example, the machine learning model looks like
 
 .. code:: python
 
-    cls = LeNet5(num_classes=len(kb.pseudo_label_list))
+    cls = LeNet5(num_classes=10)
     device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
     criterion = torch.nn.CrossEntropyLoss()
     optimizer = torch.optim.Adam(cls.parameters(), lr=0.001, betas=(0.9, 0.99))
diff --git a/docs/Examples/MNISTAdd.rst b/docs/Examples/MNISTAdd.rst
index 2390fbc..741a195 100644
--- a/docs/Examples/MNISTAdd.rst
+++ b/docs/Examples/MNISTAdd.rst
@@ -1,5 +1,19 @@
 MNIST Add
 ==================
 
-.. contents:: Table of Contents
+MNIST Add was first introduced in [1] and the inputs of this task are pairs of MNIST images and the outputs are their sums. The dataset looks like this:
 
+.. image:: ../img/image_1.png
+   :width: 350px
+   :align: center
+
+|
+
+The ``gt_pseudo_label`` is only used to test the performance of the machine learning model and is not used in the training phase.
+
+In the Abductive Learning framework, the inference process is as follows:
+
+.. image:: ../img/image_2.png
+   :width: 700px
+
+[1] Robin Manhaeve, Sebastijan Dumancic, Angelika Kimmig, Thomas Demeester, and Luc De Raedt. Deepproblog: Neural probabilistic logic programming. In Advances in Neural Information Processing Systems 31 (NeurIPS), pages 3749-3759.2018.
\ No newline at end of file