diff --git a/returnn/tf/layers/basic.py b/returnn/tf/layers/basic.py
index c386175284..fe2230f0d9 100644
--- a/returnn/tf/layers/basic.py
+++ b/returnn/tf/layers/basic.py
@@ -7246,7 +7246,7 @@ class VariableLayer(LayerBase):
   """
   layer_class = "variable"
 
-  def __init__(self, shape, dtype="float32", add_batch_axis=True, add_time_axis=False, trainable=True,
+  def __init__(self, shape, dtype="float32", add_batch_axis=False, add_time_axis=False, trainable=True,
                init=0,
                **kwargs):
     """
@@ -7298,7 +7298,7 @@ def transform_config_dict(cls, d, network, get_layer):
 
   @classmethod
   def get_out_data_from_opts(cls, name, network,
-                             shape, dtype="float32", add_batch_axis=True, add_time_axis=False, **kwargs):
+                             shape, dtype="float32", add_batch_axis=False, add_time_axis=False, **kwargs):
     """
     :param str name:
     :param returnn.tf.network.TFNetwork network:
diff --git a/tests/test_TFNetworkLayer.py b/tests/test_TFNetworkLayer.py
index 1833d74aba..f5b37abdbd 100644
--- a/tests/test_TFNetworkLayer.py
+++ b/tests/test_TFNetworkLayer.py
@@ -2473,8 +2473,8 @@ def test_TileLayer():
 def test_ScatterNdLayer_RangeLayer_RangeInAxisLayer():
   from returnn.tf.util.data import BatchDim, DimensionTag, ImplicitDynSizeDim
   n_batch, n_time, n_ts, n_in, n_out = 2, 3, 6, 7, 11
-  time_dim = DimensionTag(kind=DimensionTag.Types.Spatial, description="T")
-  feat_dim = DimensionTag(kind=DimensionTag.Types.Feature, description="F", dimension=n_in)
+  time_dim = DimensionTag(kind=DimensionTag.Types.Spatial, description="time")
+  feat_dim = DimensionTag(kind=DimensionTag.Types.Feature, description="in-feature", dimension=n_in)
   ts_dim = DimensionTag(kind=DimensionTag.Types.Spatial, description="ts", dimension=n_ts)
   rnd = numpy.random.RandomState(42)
   config = Config({
@@ -2489,9 +2489,9 @@ def test_ScatterNdLayer_RangeLayer_RangeInAxisLayer():
     "add_t": {
       "class": "combine", "kind": "add", "from": ["t", "range"],
       "out_shape": {time_dim, ts_dim, ImplicitDynSizeDim(BatchDim)}},  # (T,Ts)
-    "t_rel_var": {"class": "variable", "shape": (n_ts, n_out), "init": "glorot_uniform"},  # (B,Ts,D)
-    "output": {"class": "scatter_nd", "from": "t_rel_var", "position": "add_t", "position_axis": -1,
-               "output_dim_via_time_from": "data", "filter_invalid_indices": True}
+    "t_rel_var": {"class": "variable", "shape": (ts_dim, n_out), "init": "glorot_uniform"},  # (Ts,D)
+    "output": {"class": "scatter_nd", "from": "t_rel_var", "position": "add_t", "position_axis": ts_dim,
+               "output_dim_via_time_from": "data", "filter_invalid_indices": True}  # (T,T,D)
   }
   with make_scope() as session:
     network = TFNetwork(config=config, train_flag=True)
@@ -2506,7 +2506,7 @@ def test_ScatterNdLayer_RangeLayer_RangeInAxisLayer():
     out_layer = network.get_default_output_layer()
     assert isinstance(out_layer, ScatterNdLayer)
     assert out_layer.output.shape == (None, None, 11)
-    assert out_layer.output.feature_dim_axis_or_unspecified is NotSpecified and out_layer.output.feature_dim_axis == 3
+    assert out_layer.output.feature_dim_axis_or_unspecified is NotSpecified and out_layer.output.feature_dim_axis == 2
     assert out_layer.output.time_dim_axis == 0
 
     session.run(tf_compat.v1.variables_initializer(tf_compat.v1.global_variables() + [network.global_train_step]))