diff --git a/src/datasets/features/features.py b/src/datasets/features/features.py
index fe747bd7f4a..896a1c57643 100644
--- a/src/datasets/features/features.py
+++ b/src/datasets/features/features.py
@@ -1195,7 +1195,10 @@ class Sequence:
             which are converted to `dict` of lists of sub-features for compatibility with TFDS.
 
     """
-
+    @property
+    def dtype(self) -> str:
+        return "list"
+    
     def __new__(cls, feature=None, length=-1, **kwargs):
         # useful to still get isinstance(Sequence(Value("int64")), Sequence)
         if (
@@ -1235,7 +1238,10 @@ def __repr__(self):
             return f"{type(self).__name__}({self.feature}, length={self.length})"
         else:
             return f"{type(self).__name__}({self.feature})"
-
+        
+    @property
+    def dtype(self) -> str:
+        return "list"
 
 @dataclass(repr=False)
 class LargeList:
@@ -1256,7 +1262,10 @@ class LargeList:
 
     def __repr__(self):
         return f"{type(self).__name__}({self.feature})"
-
+    
+    @property
+    def dtype(self) -> str:
+        return "list"
 
 FeatureType = Union[
     dict,
diff --git a/src/datasets/iterable_dataset.py b/src/datasets/iterable_dataset.py
index 4555d97b487..881b726a75e 100644
--- a/src/datasets/iterable_dataset.py
+++ b/src/datasets/iterable_dataset.py
@@ -3404,7 +3404,7 @@ def filter(
             ex_iterable = FormattedExamplesIterable(
                 ex_iterable,
                 formatting=self._formatting,
-                features=None if ex_iterable.is_typed else self._info.features,
+                features=ex_iterable.features if ex_iterable.is_typed else self._info.features,
                 token_per_repo_id=self._token_per_repo_id,
             )