diff --git a/required_packages.py b/required_packages.py
index cefd122969..fbb6305291 100644
--- a/required_packages.py
+++ b/required_packages.py
@@ -26,7 +26,6 @@
     'cloudpickle>=1.3',
     'gast>=0.3.2',  # For autobatching
     'dm-tree',  # For NumPy/JAX backends (hence, also for prefer_static)
-    'typing-extensions<4.6.0',  # TODO(b/284106340): Remove this pin
 ]
 
 if __name__ == '__main__':
diff --git a/tensorflow_probability/python/experimental/linalg/linear_operator_psd_kernel_test.py b/tensorflow_probability/python/experimental/linalg/linear_operator_psd_kernel_test.py
index c3e45be183..9a4aefcf13 100644
--- a/tensorflow_probability/python/experimental/linalg/linear_operator_psd_kernel_test.py
+++ b/tensorflow_probability/python/experimental/linalg/linear_operator_psd_kernel_test.py
@@ -271,15 +271,17 @@ def test_matmul_grad_xla_kernelparams(self):
     feature_dim = 3
 
     def kernel_fn(eq_params, poly_params):
-      return (exponentiated_quadratic.ExponentiatedQuadratic(**eq_params) *
-              polynomial.Polynomial(**poly_params))
+      return (exponentiated_quadratic.ExponentiatedQuadratic(*eq_params) *
+              polynomial.Polynomial(bias_amplitude=poly_params[0],
+                                    shift=poly_params[1]))
 
+    # TODO(b/284106340): Return this to a dictionary.
     kernel_args = (
-        dict(length_scale=tf.random.uniform([], .5, 1.5, dtype=tf.float64),
-             amplitude=tf.random.uniform([], 1.5, 2.5, dtype=tf.float64)),
-        dict(bias_amplitude=tf.random.uniform([feature_dim], .5, 1.5,
-                                              dtype=tf.float64),
-             shift=tf.random.normal([feature_dim], dtype=tf.float64)))
+        (tf.random.uniform([], 1.5, 2.5, dtype=tf.float64),  # amplitude
+         tf.random.uniform([], .5, 1.5, dtype=tf.float64)),  # length_scale
+        (tf.random.uniform([feature_dim], .5, 1.5,  # bias_amplitude
+                           dtype=tf.float64),
+         tf.random.normal([feature_dim], dtype=tf.float64)))  # shift
 
     x1 = tf.random.normal([5, feature_dim], dtype=tf.float64)
     x2 = tf.random.normal([7, feature_dim], dtype=tf.float64)