diff --git a/apps/common/util/split_model.py b/apps/common/util/split_model.py
index 104ce2cec..6e0a5f8eb 100644
--- a/apps/common/util/split_model.py
+++ b/apps/common/util/split_model.py
@@ -26,7 +26,7 @@ def get_level_block(text, level_content_list, level_content_index, cursor):
         level_content_list) else None
     start_index = text.index(start_content, cursor)
     end_index = text.index(next_content, start_index + 1) if next_content is not None else len(text)
-    return text[start_index:end_index].replace(level_content_list[level_content_index]['content'], ""), end_index
+    return text[start_index:end_index].lstrip(level_content_list[level_content_index]['content']), end_index
 
 
 def to_tree_obj(content, state='title'):
@@ -355,9 +355,9 @@ class SplitModel:
 
 
 default_split_pattern = {
-    'md': [re.compile('(?<!#)# (?!#).*'), re.compile('(?<!#)## (?!#).*'), re.compile("(?<!#)### (?!#).*"),
-           re.compile("(?<!#)####(?!#).*"), re.compile("(?<!#)#####(?!#).*"),
-           re.compile("(?<!#)######(?!#).*")],
+    'md': [re.compile('(?<=^)# .*|(?<=\\n)# .*'), re.compile('(?<!#)## (?!#).*'), re.compile("(?<!#)### (?!#).*"),
+           re.compile("(?<!#)#### (?!#).*"), re.compile("(?<!#)##### (?!#).*"),
+           re.compile("(?<!#)###### (?!#).*")],
     'default': [re.compile("(?<!\n)\n\n.+")]
 }
 
diff --git a/apps/dataset/serializers/document_serializers.py b/apps/dataset/serializers/document_serializers.py
index 3670d33eb..4e79cf062 100644
--- a/apps/dataset/serializers/document_serializers.py
+++ b/apps/dataset/serializers/document_serializers.py
@@ -490,9 +490,10 @@ class DocumentSerializers(ApiMixin, serializers.Serializer):
     class SplitPattern(ApiMixin, serializers.Serializer):
         @staticmethod
         def list():
-            return [{'key': "#", 'value': '# .*'}, {'key': '##', 'value': '(?<!#)## (?!#).*'},
-                    {'key': '###', 'value': "(?<!#)### (?!#).*"}, {'key': '####', 'value': "(?<!#)####(?!#).*"},
-                    {'key': '#####', 'value': "(?<!#)#####(?!#).*"}, {'key': '######', 'value': "(?<!#)######(?!#).*"},
+            return [{'key': "#", 'value': '(?<=^)# .*|(?<=\\n)# .*'}, {'key': '##', 'value': '(?<!#)## (?!#).*'},
+                    {'key': '###', 'value': "(?<!#)### (?!#).*"}, {'key': '####', 'value': "(?<!#)#### (?!#).*"},
+                    {'key': '#####', 'value': "(?<!#)##### (?!#).*"},
+                    {'key': '######', 'value': "(?<!#)###### (?!#).*"},
                     {'key': '-', 'value': '(?<! )- .*'},
                     {'key': '空格', 'value': '(?<!\\s)\\s(?!\\s)'},
                     {'key': '分号', 'value': '(?<!；)；(?!；)'}, {'key': '逗号', 'value': '(?<!，)，(?!，)'},