diff --git a/test/test_utils.py b/test/test_utils.py
index 21f9bf529..d8a2518b0 100644
--- a/test/test_utils.py
+++ b/test/test_utils.py
@@ -211,7 +211,7 @@ def test_remove_duplicates_1dim(key, dim):
         assert (output_tensordict == expected_output).all()
 
 
-@pytest.mark.parametrize("dim", (0, 1, -1, -2))
+@pytest.mark.parametrize("dim", (0, 1, 2, -1, -2, -3))
 def test_remove_duplicates_2dims(dim):
     key = "tensor1"
     input_tensordict = TensorDict(
@@ -222,26 +222,32 @@ def test_remove_duplicates_2dims(dim):
         batch_size=[4, 4],
     )
 
-    output_tensordict = remove_duplicates(input_tensordict, key, dim)
+    if dim in (2, -3):
+        with pytest.raises(
+            ValueError,
+            match=f"The specified dimension '{dim}' is invalid for a TensorDict with batch size .*.",
+        ):
+            remove_duplicates(input_tensordict, key, dim)
 
-    if dim in (0, -2):
-        expected_output = TensorDict(
-            {
-                "tensor1": torch.ones(1, 4),
-                "tensor2": torch.ones(1, 4),
-            },
-            batch_size=[1, 4],
-        )
     else:
-        expected_output = TensorDict(
-            {
-                "tensor1": torch.ones(4, 1),
-                "tensor2": torch.ones(4, 1),
-            },
-            batch_size=[4, 1],
-        )
-
-    assert (output_tensordict == expected_output).all()
+        output_tensordict = remove_duplicates(input_tensordict, key, dim)
+        if dim in (0, -2):
+            expected_output = TensorDict(
+                {
+                    "tensor1": torch.ones(1, 4),
+                    "tensor2": torch.ones(1, 4),
+                },
+                batch_size=[1, 4],
+            )
+        else:
+            expected_output = TensorDict(
+                {
+                    "tensor1": torch.ones(4, 1),
+                    "tensor2": torch.ones(4, 1),
+                },
+                batch_size=[4, 1],
+            )
+        assert (output_tensordict == expected_output).all()
 
 
 if __name__ == "__main__":