diff --git a/datar/__init__.py b/datar/__init__.py
index 6975478b..b9af686d 100644
--- a/datar/__init__.py
+++ b/datar/__init__.py
@@ -30,7 +30,7 @@
 )
 
 __all__ = ("f", "get_versions")
-__version__ = "0.6.0"
+__version__ = "0.6.1"
 
 
 def get_versions(prnt: bool = True) -> _VersionsTuple:
diff --git a/datar/base/arithmetic.py b/datar/base/arithmetic.py
index 0592cf56..740e3dce 100644
--- a/datar/base/arithmetic.py
+++ b/datar/base/arithmetic.py
@@ -90,7 +90,7 @@ def prod(x: "NDFrame", na_rm: bool = True) -> "NDFrame":
     (TibbleGrouped, GroupBy),
     "prod",
     pre=lambda x, na_rm=True: _warn_na_rm(
-        "sum", na_rm, "Use f.x.prod(min_count=...) to control NA produces."
+        "prod", na_rm, "Use f.x.prod(min_count=...) to control NA produces."
     )
     or (x, (), {}),
 )
diff --git a/datar/core/broadcast.py b/datar/core/broadcast.py
index c3a921bc..83ee6a1d 100644
--- a/datar/core/broadcast.py
+++ b/datar/core/broadcast.py
@@ -684,11 +684,10 @@ def _(value: SeriesGroupBy, name: str) -> Tibble:
 @init_tibble_from.register(DataFrameGroupBy)
 def _(value: Union[DataFrame, DataFrameGroupBy], name: str) -> Tibble:
     from ..tibble import as_tibble
-
     result = regcall(as_tibble, value)
+
     if name:
-        if result is value:
-            result = value.copy()
+        result = result.copy()
         result.columns = [f"{name}${col}" for col in result.columns]
     return result
 
diff --git a/datar/core/factory.py b/datar/core/factory.py
index 4bb1eaa8..02c47439 100644
--- a/datar/core/factory.py
+++ b/datar/core/factory.py
@@ -81,7 +81,7 @@ def _preprocess_args(sign: "Signature", data_args, args, kwargs):
             bound.arguments[arg] = args_df
         elif arg == "__args_raw":
             bound.arguments[arg] = args_raw
-        elif arg in args_df:
+        elif arg in args_df or args_df.columns.str.startswith(f"{arg}$").any():
             bound.arguments[arg] = args_df[arg]
         elif sign.parameters[arg].kind == sign.parameters[arg].VAR_POSITIONAL:
             star_args = bound.arguments[arg]
@@ -345,7 +345,6 @@ def func_factory(
 
     def _pipda_func(__x, *args, **kwargs):
         bound = _preprocess_args(sign, data_args, (__x, *args), kwargs)
-
         out = dispatched(*bound.args, **bound.kwargs)
         if (
             kind == "transform"
diff --git a/datar/core/tibble.py b/datar/core/tibble.py
index 31d5a23a..4ded325e 100644
--- a/datar/core/tibble.py
+++ b/datar/core/tibble.py
@@ -12,7 +12,7 @@
 
 from .collections import Collection
 from .contexts import Context
-from .utils import name_of, regcall
+from .utils import apply_dtypes, name_of, regcall
 from .names import repair_names
 
 
@@ -86,19 +86,7 @@ def from_pairs(
         if _dtypes is True:
             return out.convert_dtypes()
 
-        if not isinstance(_dtypes, dict):
-            dtypes = zip(out.columns, [_dtypes] * out.shape[1])
-        else:
-            dtypes = _dtypes.items()
-
-        for column, dtype in dtypes:
-            if column in out:
-                out[column] = out[column].astype(dtype)
-            else:
-                for col in out:
-                    if col.startswith(f"{column}$"):
-                        out[col] = out[col].astype(dtype)
-
+        apply_dtypes(out, _dtypes)
         return out
 
     @classmethod
@@ -160,7 +148,7 @@ def __setitem__(self, key, value):
             else:
                 for col in value.columns:
                     colname = f"{key}${col}"
-                    super().__setitem__(colname, value[col])
+                    super().__setitem__(colname, value[col].copy())
 
         else:
             super().__setitem__(key, value)
@@ -268,7 +256,8 @@ def __getitem__(self, key):
         result = super().__getitem__(key)
         if isinstance(result, Series):
             return self._datar["grouped"][key]
-
+        if isinstance(result, DataFrame):
+            return TibbleGrouped(result, copy=False, meta=self._datar)
         return result  # pragma: no cover
 
     def __setitem__(self, key, value):
@@ -403,7 +392,8 @@ def __getitem__(self, key):
         result = super().__getitem__(key)
         if isinstance(result, SeriesGroupBy):
             result.is_rowwise = True
-
+        elif isinstance(result, DataFrame):
+            return reconstruct_tibble(self, result)
         return result
 
     def copy(self, deep: bool = True) -> "TibbleRowwise":
diff --git a/datar/dplyr/across.py b/datar/dplyr/across.py
index 3ea93064..3427270c 100644
--- a/datar/dplyr/across.py
+++ b/datar/dplyr/across.py
@@ -10,7 +10,7 @@
 from pipda.utils import functype
 
 from ..core.broadcast import add_to_tibble
-from ..core.tibble import Tibble
+from ..core.tibble import Tibble, reconstruct_tibble
 from ..core.utils import vars_select, regcall
 from ..core.middlewares import CurColumn
 from ..core.contexts import Context
@@ -226,7 +226,7 @@ def c_across(
         _cols = regcall(everything, _data)
 
     _cols = vars_select(_data.columns, _cols)
-    return _data.iloc[:, _cols]
+    return reconstruct_tibble(_data, _data.iloc[:, _cols])
 
 
 @register_func(
diff --git a/datar/tibble/tibble.py b/datar/tibble/tibble.py
index 386251fa..7a9d4bfd 100644
--- a/datar/tibble/tibble.py
+++ b/datar/tibble/tibble.py
@@ -163,7 +163,7 @@ def tibble_row(
 @register_verb(context=Context.EVAL)
 def as_tibble(df: Any) -> Tibble:
     """Convert a pandas DataFrame object to Tibble object"""
-    return Tibble(df)
+    return Tibble(df, copy=False)
 
 
 @as_tibble.register(DataFrameGroupBy)
diff --git a/docs/CHANGELOG.md b/docs/CHANGELOG.md
index 6d1ac2d6..3d60cdcd 100644
--- a/docs/CHANGELOG.md
+++ b/docs/CHANGELOG.md
@@ -1,3 +1,8 @@
+## 0.6.1
+
+- 🐛 Fix `rep(df, n)` producing a nested df
+- 🐛 Fix `TibbleGrouped.__getitem__()` not keeping grouping structures
+
 ## 0.6.0
 
 ### General
diff --git a/pyproject.toml b/pyproject.toml
index 117b1fa5..e6325319 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "datar"
-version = "0.6.0"
+version = "0.6.1"
 description = "Port of dplyr and other related R packages in python, using pipda."
 authors = ["pwwang <pwwang@pwwang.com>"]
 readme = "README.md"
diff --git a/tests/core/test_factory.py b/tests/core/test_factory.py
index deb69cf4..9ed4ea81 100644
--- a/tests/core/test_factory.py
+++ b/tests/core/test_factory.py
@@ -47,7 +47,7 @@ def double(x):
     x = DataFrame({"a": [3, 4]})
     out = double(x)
     assert isinstance(out, DataFrame)
-    assert_iterable_equal(out["x$a"], [6, 8])
+    assert_iterable_equal(out.a, [6, 8])
 
     # default on seriesgroupby
     x = Series([1, 2, 1, 2]).groupby([1, 1, 2, 2])
@@ -60,7 +60,7 @@ def double(x):
     x = tibble(x=[1, 2, 1, 2], g=[1, 1, 2, 2]).group_by("g")
     out = double(x)
     # grouping variables not included
-    assert_iterable_equal(out.x, [2, 4, 2, 4])
+    assert_iterable_equal(out.x.obj, [2, 4, 2, 4])
 
     x = tibble(x=[1, 2, 1, 2], g=[1, 1, 2, 2]).rowwise("g")
     out = double(x)
diff --git a/tests/dplyr/test_mutate.py b/tests/dplyr/test_mutate.py
index fa20fc7d..7ab39b89 100644
--- a/tests/dplyr/test_mutate.py
+++ b/tests/dplyr/test_mutate.py
@@ -145,11 +145,11 @@ def test_handles_data_frame_columns():
     assert_tibble_equal(res["new_col"], tibble(x=[1, 2, 3]))
 
     res = mutate(group_by(df, f.a), new_col=tibble(x=f.a))
-    assert_tibble_equal(res["new_col"], tibble(x=[1, 2, 3]))
+    assert_iterable_equal(res["new_col"].x.obj, [1, 2, 3])
 
     rf = rowwise(df, f.a)
     res = mutate(rf, new_col=tibble(x=f.a))
-    assert_tibble_equal(res["new_col"], tibble(x=[1, 2, 3]))
+    assert_tibble_equal(res["new_col"], tibble(x=[1, 2, 3]) >> rowwise())
 
 
 def test_unnamed_data_frames_are_automatically_unspliced():