diff --git a/src/lerobot/policies/groot/modeling_groot.py b/src/lerobot/policies/groot/modeling_groot.py
index cbfe647f7..f0e1c08d3 100644
--- a/src/lerobot/policies/groot/modeling_groot.py
+++ b/src/lerobot/policies/groot/modeling_groot.py
@@ -468,6 +468,14 @@ class GrootPolicy(PreTrainedPolicy):
     @torch.no_grad()
     def select_action(self, batch: dict[str, Tensor]) -> Tensor:
         """Select single action from action queue."""
+        if getattr(self.config, "use_relative_actions", False):
+            raise NotImplementedError(
+                "GrootPolicy.select_action does not support relative-action policies because cached "
+                "relative chunk actions can be decoded against newer observation states. Use "
+                "predict_action_chunk and postprocess the full chunk before queuing actions, or use "
+                "the RTC/chunked rollout inference path."
+            )
+
         self.eval()
 
         if len(self._action_queue) == 0:
diff --git a/src/lerobot/policies/groot/processor_groot.py b/src/lerobot/policies/groot/processor_groot.py
index f37e2dbfd..6c68aa693 100644
--- a/src/lerobot/policies/groot/processor_groot.py
+++ b/src/lerobot/policies/groot/processor_groot.py
@@ -2158,6 +2158,12 @@ class GrootN17ActionDecodeStep(ProcessorStep):
             return transition
 
         action_np = action.detach().cpu().float().numpy()
+        if self.use_relative_action and action_np.ndim != 3:
+            raise NotImplementedError(
+                "GrootN17ActionDecodeStep cannot decode native relative actions one step at a time. "
+                "Decode the full action chunk returned by predict_action_chunk while the matching "
+                "GrootN17PackInputsStep state is still cached, then queue the decoded absolute actions."
+            )
         # The sync action queue postprocesses popped actions as (B, D); decode
         # them as single-step (B, 1, D) chunks and squeeze the horizon back at
         # the end so both ranks share the chunk decode logic below.
diff --git a/tests/policies/groot/test_groot_n1_7.py b/tests/policies/groot/test_groot_n1_7.py
index 2fbf3c11a..4a45393f8 100644
--- a/tests/policies/groot/test_groot_n1_7.py
+++ b/tests/policies/groot/test_groot_n1_7.py
@@ -1324,6 +1324,55 @@ def test_groot_n1_7_action_decode_truncates_to_valid_horizon_for_relative_stats(
     torch.testing.assert_close(decoded[..., 5], torch.full((1, 16), 5.0))
 
 
+def test_groot_n1_7_action_decode_rejects_stepwise_native_relative_actions():
+    raw_stats = {
+        "state": {
+            "single_arm": _stats([0.0] * 5),
+            "gripper": _stats([0.0]),
+        },
+        "action": {
+            "single_arm": _stats([0.0] * 5),
+            "gripper": _stats([0.0]),
+        },
+        "relative_action": {
+            "single_arm": _stats([0.0] * 5),
+        },
+    }
+    modality_config = {
+        "state": {
+            "modality_keys": ["single_arm", "gripper"],
+        },
+        "action": {
+            "modality_keys": ["single_arm", "gripper"],
+            "action_configs": [
+                {"rep": "RELATIVE", "type": "NON_EEF", "format": "DEFAULT", "state_key": None},
+                {"rep": "ABSOLUTE", "type": "NON_EEF", "format": "DEFAULT", "state_key": None},
+            ],
+        },
+    }
+    pack_step = GrootN17PackInputsStep(
+        raw_stats=raw_stats,
+        modality_config=modality_config,
+        normalize_min_max=False,
+    )
+    pack_step(
+        {
+            TransitionKey.OBSERVATION: {OBS_STATE: torch.zeros(1, 6)},
+            TransitionKey.COMPLEMENTARY_DATA: {},
+        }
+    )
+    decode_step = GrootN17ActionDecodeStep(
+        env_action_dim=6,
+        raw_stats=raw_stats,
+        modality_config=modality_config,
+        use_relative_action=True,
+        pack_step=pack_step,
+    )
+
+    with pytest.raises(NotImplementedError, match="cannot decode native relative actions one step at a time"):
+        decode_step({TransitionKey.ACTION: torch.zeros(1, 6)})
+
+
 def test_groot_n1_7_action_decode_requires_gripper_key_for_libero_transform():
     step = GrootN17ActionDecodeStep(
         env_action_dim=1,
@@ -2324,6 +2373,14 @@ def test_groot_n1_7_libero_execution_horizon_uses_core_eight_action_cadence(tmp_
     assert infer_groot_n1_7_action_execution_horizon(model_path, "libero_sim") == 8
 
 
+def test_groot_select_action_rejects_relative_action_policies():
+    policy = object.__new__(GrootPolicy)
+    object.__setattr__(policy, "config", SimpleNamespace(use_relative_actions=True))
+
+    with pytest.raises(NotImplementedError, match="select_action does not support relative-action policies"):
+        policy.select_action({})
+
+
 def test_groot_n1_7_select_action_uses_checkpoint_valid_horizon(tmp_path, monkeypatch):
     from lerobot.policies.groot.groot_n1_7 import GR00TN17