diff --git a/docs/source/conf.py b/docs/source/conf.py
index 48c6e836c..f9cbcc18e 100644
--- a/docs/source/conf.py
+++ b/docs/source/conf.py
@@ -95,7 +95,8 @@
 # You can specify multiple suffix as a list of string:
 #
 # source_suffix = ['.rst', '.md']
-source_suffix = [".rst", ".ipynb"]
+# NOTE: `.rst` is the default suffix of sphinx, and nbsphinx will
+# automatically add support for `.ipynb` suffix.
 
 # do not execute cells
 nbsphinx_execute = "never"
diff --git a/notebooks/source/bayesian_hierarchical_stacking.ipynb b/notebooks/source/bayesian_hierarchical_stacking.ipynb
index 40ba794a2..0b4fb8673 100644
--- a/notebooks/source/bayesian_hierarchical_stacking.ipynb
+++ b/notebooks/source/bayesian_hierarchical_stacking.ipynb
@@ -737,7 +737,7 @@
     "\n",
     "Such a matrix $W$ would be required to have each column sum to $1$. Hence, we calculate each row $W_i$ of $W$ as:\n",
     "\n",
-    "$$ W_i = \\text{softmax}(X\\text{_stacking}_i \\cdot \\beta), $$\n",
+    "$$ W_i = \\text{softmax}(X\\_\\text{stacking}_i \\cdot \\beta), $$\n",
     "\n",
     "where $\\beta$ is a matrix whose values we seek to determine. For the discrete features, $\\beta$ is given a hierarchical structure over the possible inputs. Continuous features, on the other hand, get no hierarchical structure in this case study and just vary according to the input values.\n",
     "\n",
diff --git a/numpyro/distributions/directional.py b/numpyro/distributions/directional.py
index 6fb19d7ff..cfe91318e 100644
--- a/numpyro/distributions/directional.py
+++ b/numpyro/distributions/directional.py
@@ -154,10 +154,11 @@ def variance(self):
 
 
 class SineSkewed(Distribution):
-    """Sine-skewing [1] is a procedure for producing a distribution that breaks pointwise symmetry on a torus
+    r"""Sine-skewing [1] is a procedure for producing a distribution that breaks pointwise symmetry on a torus
     distribution. The new distribution is called the Sine Skewed X distribution, where X is the name of the (symmetric)
     base distribution. Torus distributions are distributions with support on products of circles
-    (i.e., ⨂^d S^1 where S^1=[-pi,pi) ). So, a 0-torus is a point, the 1-torus is a circle,
+    (i.e., :math:`\otimes S^1` where :math:`S^1 = [-pi,pi)`).
+    So, a 0-torus is a point, the 1-torus is a circle,
     and the 2-torus is commonly associated with the donut shape.
 
     The sine skewed X distribution is parameterized by a weight parameter for each dimension of the event of X.
@@ -284,7 +285,8 @@ def mean(self):
 
 
 class SineBivariateVonMises(Distribution):
-    r"""Unimodal distribution of two dependent angles on the 2-torus (S^1 ⨂ S^1) given by
+    r"""Unimodal distribution of two dependent angles on the 2-torus
+    (:math:`S^1 \otimes S^1`) given by
 
     .. math::
         C^{-1}\exp(\kappa_1\cos(x_1-\mu_1) + \kappa_2\cos(x_2 -\mu_2) + \rho\sin(x_1 - \mu_1)\sin(x_2 - \mu_2))