diff --git a/dpsynth/local_mode/vectorized_transformations.py b/dpsynth/local_mode/vectorized_transformations.py
index 5144c081..830d7bc4 100644
--- a/dpsynth/local_mode/vectorized_transformations.py
+++ b/dpsynth/local_mode/vectorized_transformations.py
@@ -218,8 +218,10 @@ def undiscretize(
   else:
     raise ValueError(f'Unsupported interval_handling: {handling}')
 
-  if attribute_domain.dtype == 'int' and attribute_domain.clip_to_range:
-    result = np.ceil(result).astype(int)
+  if attribute_domain.dtype == 'int':
+    result = np.ceil(result)
+    if attribute_domain.clip_to_range or not np.isnan(sentinel):
+      result = result.astype(int)
   return result
 
 
diff --git a/tests/local_mode/vectorized_transformations_test.py b/tests/local_mode/vectorized_transformations_test.py
index 6f7fb17b..32d1d55f 100644
--- a/tests/local_mode/vectorized_transformations_test.py
+++ b/tests/local_mode/vectorized_transformations_test.py
@@ -298,6 +298,27 @@ def test_custom_sentinel_sample(self):
     )
     self.assertEqual(result[0], -1)
 
+  @parameterized.parameters(('midpoint', None), ('sample', -1))
+  def test_integer_dtype_no_clip(self, interval_handling, sentinel):
+    rng = np.random.default_rng(0)
+    attr = domain.NumericalAttribute(
+        min_value=0,
+        max_value=3,
+        dtype='int',
+        clip_to_range=False,
+        sentinel=sentinel,
+        interval_handling=interval_handling,
+    )
+    result = vectorized_transformations.undiscretize(
+        np.array([0, 1, 2, 3, 4]), np.array([0.0, 1.0, 2.0]), attr, rng=rng
+    )
+    np.testing.assert_array_equal(result[1:], [0, 1, 2, 3])
+    if sentinel is None:
+      self.assertTrue(np.isnan(result[0]))
+    else:
+      self.assertTrue(np.issubdtype(result.dtype, np.integer))
+      self.assertEqual(result[0], sentinel)
+
 
 class MergeRareValuesTest(absltest.TestCase):