diff --git a/slime/rollout/rm_hub/deepscaler.py b/slime/rollout/rm_hub/deepscaler.py
index 39d4de383a..428c12cca2 100644
--- a/slime/rollout/rm_hub/deepscaler.py
+++ b/slime/rollout/rm_hub/deepscaler.py
@@ -2,6 +2,12 @@
 
 
 def get_deepscaler_rule_based_reward(response, label):
+    # Guard against a missing response, mirroring the gpqa / f1 reward
+    # functions in this package. async_rm passes sample.response straight
+    # through, so a None response would otherwise raise a TypeError here
+    # instead of scoring 0.
+    if not response:
+        return 0
     if "</think>" in response:
         model_solution = response.split("</think>")[-1]
     elif "###Response" in response:
diff --git a/tests/test_rm_deepscaler.py b/tests/test_rm_deepscaler.py
index 0883a60622..828b8df37d 100644
--- a/tests/test_rm_deepscaler.py
+++ b/tests/test_rm_deepscaler.py
@@ -80,6 +80,16 @@ def test_label_with_boxed_marker_is_extracted_too():
     assert get_deepscaler_rule_based_reward(r"</think>\boxed{42}", r"\boxed{42}") == 1
 
 
+@pytest.mark.unit
+def test_missing_response_returns_zero():
+    """A None/empty response scores 0 instead of raising. ``async_rm``
+    feeds ``sample.response`` straight in, and the sibling gpqa / f1 reward
+    functions already guard this; without the guard ``None`` raised
+    ``TypeError: argument of type 'NoneType' is not iterable``."""
+    assert get_deepscaler_rule_based_reward(None, "42") == 0
+    assert get_deepscaler_rule_based_reward("", "42") == 0
+
+
 @pytest.mark.unit
 def test_wrong_answer_returns_zero():
     """Sanity-check the negative side of the contract."""