diff --git a/doc/tutorials/feature_interaction_constraint.rst b/doc/tutorials/feature_interaction_constraint.rst
index 205524025..3b08320cb 100644
--- a/doc/tutorials/feature_interaction_constraint.rst
+++ b/doc/tutorials/feature_interaction_constraint.rst
@@ -44,12 +44,9 @@ is the product of interaction between :math:`x_1`, :math:`x_7`, and
Source(source, format='png').render('../_static/feature_interaction_illustration1', view=False)
Source(source, format='svg').render('../_static/feature_interaction_illustration1', view=False)
-.. raw:: html
-
-
-
-
+.. figure:: ../_static/feature_interaction_illustration1.svg
+ :align: center
+ :figwidth: 80 %
When the tree depth is larger than one, many variables interact on
the sole basis of minimizing training loss, and the resulting decision tree may
@@ -130,14 +127,12 @@ first and second constraints (``[0, 1]``, ``[2, 3, 4]``).
Source(source, format='png').render('../_static/feature_interaction_illustration3', view=False)
Source(source, format='svg').render('../_static/feature_interaction_illustration3', view=False)
-.. raw:: html
-
-
-
-
+.. figure:: ../_static/feature_interaction_illustration2.svg
+ :align: left
+
+.. figure:: ../_static/feature_interaction_illustration3.svg
+
****************************************************
Enforcing Feature Interaction Constraints in XGBoost