diff --git a/keras/backend/common/variables_test.py b/keras/backend/common/variables_test.py
index 42bd4d45524..a188f954bef 100644
--- a/keras/backend/common/variables_test.py
+++ b/keras/backend/common/variables_test.py
@@ -238,7 +238,7 @@ def test_variable_numpy(self):
         self.assertAllClose(v.numpy(), np.array([1, 2, 3]))
 
     @pytest.mark.skipif(
-        backend.backend() != "tf",
+        backend.backend() != "tensorflow",
         reason="Tests for MirroredVariable under tf backend",
     )
     def test_variable_numpy_scalar(self):
diff --git a/keras/layers/preprocessing/discretization.py b/keras/layers/preprocessing/discretization.py
index 005109c745b..ca5eb255a5e 100644
--- a/keras/layers/preprocessing/discretization.py
+++ b/keras/layers/preprocessing/discretization.py
@@ -235,6 +235,8 @@ def call(self, inputs):
             dtype=self.compute_dtype,
             backend_module=self.backend,
         )
+        if self.sparse:
+            return tf.sparse.from_dense(outputs)
         return outputs
 
     def get_config(self):
diff --git a/keras/layers/preprocessing/discretization_test.py b/keras/layers/preprocessing/discretization_test.py
index 490ff9fa4da..8677282de7f 100644
--- a/keras/layers/preprocessing/discretization_test.py
+++ b/keras/layers/preprocessing/discretization_test.py
@@ -1,6 +1,7 @@
 import os
 
 import numpy as np
+import pytest
 from absl.testing import parameterized
 from tensorflow import data as tf_data
 
@@ -11,7 +12,7 @@
 from keras.saving import saving_api
 
 
-class DicretizationTest(testing.TestCase, parameterized.TestCase):
+class DiscretizationTest(testing.TestCase, parameterized.TestCase):
     def test_discretization_basics(self):
         self.run_layer_test(
             layers.Discretization,
@@ -125,6 +126,55 @@ def test_saving(self):
         model = saving_api.load_model(fpath)
         self.assertAllClose(layer(ref_input), ref_output)
 
-    def test_sparse_inputs(self):
-        # TODO
-        pass
+    @parameterized.parameters(
+        [
+            (
+                "one_hot",
+                [[-1.0, 0.2, 0.7, 1.2]],
+                [
+                    [
+                        [1.0, 0.0, 0.0, 0.0],
+                        [0.0, 1.0, 0.0, 0.0],
+                        [0.0, 0.0, 1.0, 0.0],
+                        [0.0, 0.0, 0.0, 1.0],
+                    ]
+                ],
+            ),
+            (
+                "multi_hot",
+                [[[-1.0], [0.2], [0.7], [1.2]]],
+                [
+                    [
+                        [1.0, 0.0, 0.0, 0.0],
+                        [0.0, 1.0, 0.0, 0.0],
+                        [0.0, 0.0, 1.0, 0.0],
+                        [0.0, 0.0, 0.0, 1.0],
+                    ]
+                ],
+            ),
+            (
+                "count",
+                [[-1.0], [0.2], [0.7], [1.2]],
+                [
+                    [1.0, 0.0, 0.0, 0.0],
+                    [0.0, 1.0, 0.0, 0.0],
+                    [0.0, 0.0, 1.0, 0.0],
+                    [0.0, 0.0, 0.0, 1.0],
+                ],
+            ),
+        ]
+    )
+    @pytest.mark.skipif(
+        backend.backend() != "tensorflow",
+        reason="Sparse tensor only works in TensorFlow",
+    )
+    def test_sparse_output(self, output_mode, input_array, expected_output):
+        from keras.utils.module_utils import tensorflow as tf
+
+        x = np.array(input_array)
+        layer = layers.Discretization(
+            bin_boundaries=[0.0, 0.5, 1.0], sparse=True, output_mode=output_mode
+        )
+        output = layer(x)
+        self.assertTrue(isinstance(output, tf.SparseTensor))
+        self.assertAllClose(output, np.array(expected_output))