diff --git a/modelscope/models/nlp/bart/text_error_correction.py b/modelscope/models/nlp/bart/text_error_correction.py
index 97c3a7a9..09b867c4 100644
--- a/modelscope/models/nlp/bart/text_error_correction.py
+++ b/modelscope/models/nlp/bart/text_error_correction.py
@@ -82,5 +82,5 @@ class BartForTextErrorCorrection(TorchModel):
         batch_preds = []
         for i in range(batch_size):
             # get 1-best List[Tensor]
-            batch_preds.append(translations[i][0]['tokens'])
+            batch_preds.append(translations[i][0]['tokens'].tolist())
         return TextErrorCorrectionOutput(predictions=batch_preds)
diff --git a/modelscope/outputs/nlp_outputs.py b/modelscope/outputs/nlp_outputs.py
index ed42cb5a..747f5bd3 100644
--- a/modelscope/outputs/nlp_outputs.py
+++ b/modelscope/outputs/nlp_outputs.py
@@ -326,7 +326,7 @@ class TextErrorCorrectionOutput(ModelOutputBase):
     """The output class for information extraction models.
     """
 
-    predictions: np.ndarray = None
+    predictions: List = None
 
 
 @dataclass
diff --git a/modelscope/pipelines/nlp/text_error_correction_pipeline.py b/modelscope/pipelines/nlp/text_error_correction_pipeline.py
index dc4bc40a..9fa5a2a8 100644
--- a/modelscope/pipelines/nlp/text_error_correction_pipeline.py
+++ b/modelscope/pipelines/nlp/text_error_correction_pipeline.py
@@ -80,7 +80,8 @@ class TextErrorCorrectionPipeline(Pipeline):
 
         sc_tensor = inputs['predictions']
         if isinstance(sc_tensor, list):
-            sc_tensor = sc_tensor[0]
+            if isinstance(sc_tensor[0], list):
+                sc_tensor = sc_tensor[0]
         sc_sent = self.vocab.string(
             sc_tensor, extra_symbols_to_ignore={self.vocab.pad()})
         sc_sent = (sc_sent + ' ').replace('##', '').rstrip()
diff --git a/tests/pipelines/test_text_error_correction.py b/tests/pipelines/test_text_error_correction.py
index b4bf5be9..b2a5fd4d 100644
--- a/tests/pipelines/test_text_error_correction.py
+++ b/tests/pipelines/test_text_error_correction.py
@@ -41,12 +41,16 @@ class TextErrorCorrectionTest(unittest.TestCase):
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_run_with_model_name_batch(self):
-        run_kwargs = {'batch_size': 2}
         pipeline_ins = pipeline(
             task=Tasks.text_error_correction, model=self.model_id)
-        print(
-            'batch: ',
-            pipeline_ins([self.input, self.input_2, self.input_3], run_kwargs))
+        sents = [
+            self.input, self.input_2, self.input_3, self.input_4,
+            self.input_law
+        ]
+        rs1 = pipeline_ins(sents, batch_size=2)
+        rs2 = pipeline_ins(sents)
+        print('batch: ', rs1, rs2)
+        self.assertEqual(rs1, rs2)
 
     @unittest.skipUnless(test_level() >= 1, 'skip test in current test level')
     def test_run_with_model_from_modelhub(self):