diff --git a/keras_core/layers/layer.py b/keras_core/layers/layer.py
index 0a6ad7967..164a0440b 100644
--- a/keras_core/layers/layer.py
+++ b/keras_core/layers/layer.py
@@ -609,15 +609,9 @@ def maybe_convert(x):
                 # Get signature default value
                 training = call_spec.arguments_dict.get("training", None)
         call_context.training = training
-        if self._call_has_training_arg():
-            if training is not None:
-                # Only populate arg if it has a concrete value
-                kwargs["training"] = training
-            elif "training" in kwargs:
-                # In some cases the value None may have been passed explicitly
-                # (e.g. by a parent Sequential).
-                # That's invalid, so don't propagate it.
-                kwargs.pop("training")
+        if self._call_has_training_arg() and training is not None:
+            # Only populate arg if it has a concrete value
+            kwargs["training"] = training
 
         ##############################
         # 6. Populate mask argument(s)
diff --git a/keras_core/models/functional.py b/keras_core/models/functional.py
index 54de5c78c..ea7949647 100644
--- a/keras_core/models/functional.py
+++ b/keras_core/models/functional.py
@@ -543,6 +543,7 @@ def call(*args, **kwargs):
         if (
             hasattr(operation, "_call_has_training_arg")
             and operation._call_has_training_arg()
+            and training is not None
         ):
             kwargs["training"] = training
         return operation(*args, **kwargs)
diff --git a/keras_core/models/sequential.py b/keras_core/models/sequential.py
index b6e01cd08..580e3c479 100644
--- a/keras_core/models/sequential.py
+++ b/keras_core/models/sequential.py
@@ -179,7 +179,7 @@ def call(self, inputs, training=None, mask=None):
             kwargs = {}
             if layer._call_has_mask_arg():
                 kwargs["mask"] = mask
-            if layer._call_has_training_arg():
+            if layer._call_has_training_arg() and training is not None:
                 kwargs["training"] = training
             outputs = layer(inputs, **kwargs)
             inputs = outputs