diff --git a/beginner_source/maskedtensor_safe_softmax.rst b/beginner_source/maskedtensor_safe_softmax.rst
new file mode 100644
index 00000000000..0f34b479caf
--- /dev/null
+++ b/beginner_source/maskedtensor_safe_softmax.rst
@@ -0,0 +1,33 @@
+Safe Softmax
+------------
+
+One of the issues that frequently comes up is the necessity for a safe softmax -- that is, if there is an entire
+batch that is "masked out" or consists entirely of padding (which, in the softmax case, translates to being set `-inf`),
+then this will result in NaNs, which can leading to training divergence. For more detail on why this functionality
+is necessary, please find refer to 
+`Issue 55056 - Feature Request for Safe Softmax <https://github.com/pytorch/pytorch/issues/55056>`__.
+
+Luckily, :class:`MaskedTensor` has solved this issue:
+
+    >>> data = torch.randn(3, 3)
+    >>> mask = torch.tensor([[True, False, False], [True, False, True], [False, False, False]])
+    >>> x = data.masked_fill(~mask, float('-inf'))
+    >>> mt = masked_tensor(data, mask)
+
+PyTorch result:
+
+    >>> x.softmax(0)
+    tensor([[0.3548,    nan, 0.0000],
+            [0.6452,    nan, 1.0000],
+            [0.0000,    nan, 0.0000]])
+
+:class:`MaskedTensor` result:
+
+    >>> mt.softmax(0)
+    MaskedTensor(
+      [
+        [  0.3548,       --,       --],
+        [  0.6452,       --,   1.0000],
+        [      --,       --,       --]
+      ]
+    )
diff --git a/index.rst b/index.rst
index df01394eded..bb7a019c83b 100644
--- a/index.rst
+++ b/index.rst
@@ -813,6 +813,7 @@ Additional Resources
    beginner/maskedtensor_overview
    beginner/maskedtensor_sparsity
    beginner/maskedtensor_distinguish_gradient
+   beginner/maskedtensor_safe_softmax
 
 
 .. toctree::