diff --git a/doubleml/double_ml.py b/doubleml/double_ml.py
index 22a066cc..f47273c5 100644
--- a/doubleml/double_ml.py
+++ b/doubleml/double_ml.py
@@ -895,7 +895,7 @@ def tune(
             raise TypeError(f"return_tune_res must be True or False. Got {str(return_tune_res)}.")
 
         if tune_on_folds:
-            tuning_res = [[None] * self.n_rep] * self._dml_data.n_treat
+            tuning_res = [[None] * self.n_rep for _ in range(self._dml_data.n_treat)]
         else:
             tuning_res = [None] * self._dml_data.n_treat
 
@@ -921,7 +921,7 @@ def tune(
                         n_iter_randomized_search,
                     )
 
-                    tuning_res[i_rep][i_d] = res
+                    tuning_res[i_d][i_rep] = res
                     nuisance_params.append(res["params"])
 
                 if set_as_params:
diff --git a/doubleml/tests/test_tune_on_folds.py b/doubleml/tests/test_tune_on_folds.py
new file mode 100644
index 00000000..a805f845
--- /dev/null
+++ b/doubleml/tests/test_tune_on_folds.py
@@ -0,0 +1,30 @@
+import numpy as np
+import pytest
+from sklearn.linear_model import Lasso
+
+import doubleml as dml
+from doubleml.plm.datasets import make_plr_CCDDHNR2018
+
+
+@pytest.mark.ci
+@pytest.mark.parametrize("n_rep, n_treat", [(1, 1), (2, 1), (1, 2), (3, 2)])
+def test_tune_on_folds_result_per_treatment_and_repetition(n_rep, n_treat):
+    np.random.seed(3141)
+    df = make_plr_CCDDHNR2018(n_obs=200, dim_x=5, return_type="DataFrame")
+    d_cols = ["d", "X1"][:n_treat]
+    x_cols = [c for c in df.columns if c.startswith("X") and c not in d_cols]
+    dml_data = dml.DoubleMLData(df, "y", d_cols, x_cols)
+    dml_plr = dml.DoubleMLPLR(dml_data, Lasso(), Lasso(), n_folds=2, n_rep=n_rep)
+
+    grid = {"ml_l": {"alpha": [0.05, 0.1]}, "ml_m": {"alpha": [0.05, 0.1]}}
+    tune_res = dml_plr.tune(grid, tune_on_folds=True, n_folds_tune=2, return_tune_res=True)
+
+    assert len(tune_res) == n_treat
+    for i_d in range(n_treat):
+        assert len(tune_res[i_d]) == n_rep
+        assert len({id(res) for res in tune_res[i_d]}) == n_rep
+        for i_rep in range(n_rep):
+            # the results of each treatment and repetition are the parameters that were set
+            params = dml_plr.params["ml_l"][d_cols[i_d]][i_rep]
+            assert tune_res[i_d][i_rep]["params"]["ml_l"] == params
+    dml_plr.fit()