diff --git a/examples/benchmarks/TFT/data_formatters/qlib_Alpha158.py b/examples/benchmarks/TFT/data_formatters/qlib_Alpha158.py
index da3d14343..44a9284f7 100644
--- a/examples/benchmarks/TFT/data_formatters/qlib_Alpha158.py
+++ b/examples/benchmarks/TFT/data_formatters/qlib_Alpha158.py
@@ -194,10 +194,10 @@ class Alpha158Formatter(GenericDataFormatter):
         """Returns fixed model parameters for experiments."""
 
         fixed_params = {
-            "total_time_steps": 16 + 6,
-            "num_encoder_steps": 16,
+            "total_time_steps": 6 + 6,
+            "num_encoder_steps": 6,
             "num_epochs": 100,
-            "early_stopping_patience": 5,
+            "early_stopping_patience": 10,
             "multiprocessing_workers": 5,
         }
 
@@ -207,11 +207,11 @@ class Alpha158Formatter(GenericDataFormatter):
         """Returns default optimised model parameters."""
 
         model_params = {
-            "dropout_rate": 0.3,
-            "hidden_layer_size": 160,
-            "learning_rate": 0.001,
-            "minibatch_size": 64,
-            "max_gradient_norm": 0.01,
+            "dropout_rate": 0.4,
+            "hidden_layer_size": 16,
+            "learning_rate": 0.0001,
+            "minibatch_size": 128,
+            "max_gradient_norm": 0.0135,
             "num_heads": 1,
             "stack_size": 1,
         }
diff --git a/examples/benchmarks/TFT/libs/tft_model.py b/examples/benchmarks/TFT/libs/tft_model.py
index 658bae60f..3e6e4346e 100644
--- a/examples/benchmarks/TFT/libs/tft_model.py
+++ b/examples/benchmarks/TFT/libs/tft_model.py
@@ -721,12 +721,7 @@ class TemporalFusionTransformer(object):
         encoder_steps = self.num_encoder_steps
 
         # Inputs.
-        all_inputs = tf.keras.layers.Input(
-            shape=(
-                time_steps,
-                combined_input_size,
-            )
-        )
+        all_inputs = tf.keras.layers.Input(shape=(time_steps, combined_input_size,))
 
         unknown_inputs, known_combined_layer, obs_inputs, static_inputs = self.get_tft_embeddings(all_inputs)
 
@@ -866,10 +861,7 @@ class TemporalFusionTransformer(object):
             """Returns LSTM cell initialized with default parameters."""
             if self.use_cudnn:
                 lstm = tf.keras.layers.CuDNNLSTM(
-                    self.hidden_layer_size,
-                    return_sequences=True,
-                    return_state=return_state,
-                    stateful=False,
+                    self.hidden_layer_size, return_sequences=True, return_state=return_state, stateful=False,
                 )
             else:
                 lstm = tf.keras.layers.LSTM(
diff --git a/examples/benchmarks/TFT/tft.py b/examples/benchmarks/TFT/tft.py
index 3387a5947..388ec7f14 100644
--- a/examples/benchmarks/TFT/tft.py
+++ b/examples/benchmarks/TFT/tft.py
@@ -82,7 +82,7 @@ def process_predicted(df, col_name):
 
     """
     df_res = df.copy()
-    df_res = df_res.rename(columns={"forecast_time": "datetime", "identifier": "instrument", "t+5": col_name})
+    df_res = df_res.rename(columns={"forecast_time": "datetime", "identifier": "instrument", "t+4": col_name})
     df_res = df_res.set_index(["datetime", "instrument"]).sort_index()
     df_res = df_res[[col_name]]
     return df_res
@@ -232,7 +232,9 @@ class TFTModel(ModelFT):
             p90_forecast = self.data_formatter.format_predictions(output_map["p90"])
             tf.keras.backend.set_session(default_keras_session)
 
-        predict = format_score(p90_forecast, "pred", 0)  # self.label_shift
+        predict50 = format_score(p50_forecast, "pred", 1)
+        predict90 = format_score(p90_forecast, "pred", 1)
+        predict = (predict50 + predict90) / 2  # self.label_shift
         # ===========================Predicting Process===========================
         return predict